Hospital Medicine

From the Department of Medicine, Medical College of Georgia at the Augusta University-University of Georgia Medical Partnership, Athens, GA (Syed H. Ali, Syed Hyder, and Dr. Murrow), and the Department of Cardiology, Piedmont Heart Institute, Piedmont Athens Regional, Athens, GA (Dr. Murrow and Mrs. Davis).

Abstract

Objectives: The aim of this study was to describe the characteristics and in-hospital outcomes of patients with acute ST-segment elevation myocardial infarction (STEMI) during the early COVID-19 pandemic at Piedmont Athens Regional (PAR), a 330-bed tertiary referral center in Northeast Georgia. 

Methods: A retrospective study was conducted at PAR to evaluate patients with acute STEMI admitted over an 8-week period during the initial COVID-19 outbreak. This study group was compared to patients admitted during the corresponding period in 2019. The primary endpoint of this study was defined as a composite of sustained ventricular arrhythmia, congestive heart failure (CHF) with pulmonary congestion, and/or in-hospital mortality. 

Results: This study cohort was composed of 64 patients with acute STEMI; 30 patients (46.9%) were hospitalized during the COVID-19 pandemic. Patients with STEMI in both the COVID-19 and control groups had similar comorbidities, Killip classification score, and clinical presentations. The median (interquartile range) time from symptom onset to reperfusion (total ischemic time) increased from 99.5 minutes (84.8-132) in 2019 to 149 minutes (96.3-231.8; P = .032) in 2020. Hospitalization during the COVID-19 period was associated with an increased risk for combined in-hospital outcome (odds ratio, 3.96; P = .046). 

Conclusion: Patients with STEMI admitted during the first wave of the COVID-19 outbreak experienced longer total ischemic time and increased risk for combined in-hospital outcomes compared to patients admitted during the corresponding period in 2019. 

Keywords: myocardial infarction, acute coronary syndrome, hospitalization, outcomes.

The emergence of the SARS-Cov-2 virus in December 2019 caused a worldwide shift in resource allocation and the restructuring of health care systems within the span of a few months. With the rapid spread of infection, the World Health Organization officially declared a pandemic in March 2020. The pandemic led to the deferral and cancellation of in-person patient visits, routine diagnostic studies, and nonessential surgeries and procedures. This response occurred secondary to a joint effort to reduce transmission via stay-at-home mandates and appropriate social distancing.¹ 

Alongside the reduction in elective procedures and health care visits, significant reductions in hospitalization rates due to decreases in acute ST-segment elevation myocardial infarction (STEMI) and catheterization laboratory utilization have been reported in many studies from around the world.^2-7 Comprehensive data demonstrating the impact of the COVID-19 pandemic on acute STEMI patient characteristics, clinical presentation, and in-hospital outcomes are lacking. Although patients with previously diagnosed cardiovascular disease are more likely to encounter worse outcomes in the setting of COVID-19, there may also be an indirect impact of the pandemic on high-risk patients, including those without the infection.⁸ Several theories have been hypothesized to explain this phenomenon. One theory postulates that the fear of contracting the virus during hospitalization is great enough to prevent patients from seeking care.² Another theory suggests that the increased utilization of telemedicine prevents exacerbation of chronic conditions and the need for hospitalization.⁹ Contrary to this trend, previous studies have shown an increased incidence of acute STEMI following stressful events such as natural disasters.¹⁰ 

The aim of this study was to describe trends pertaining to clinical characteristics and in-hospital outcomes of patients with acute STEMI during the early COVID-19 pandemic at Piedmont Athens Regional (PAR), a 330-bed tertiary referral center in Northeast Georgia. 

Methods

A retrospective cohort study was conducted at PAR to evaluate patients with STEMI admitted to the cardiovascular intensive care unit over an 8-week period (March 5 to May 5, 2020) during the COVID-19 outbreak. COVID-19 was declared a national emergency on March 13, 2020, in the United States. The institutional review board at PAR approved the study; the need for individual consent was waived under the condition that participant data would undergo de-identification and be strictly safeguarded. 

Data Collection

Because there are seasonal variations in cardiovascular admissions, patient data from a control period (March 9 to May 9, 2019) were obtained to compare with data from the 2020 period. The number of patients with the diagnosis of acute STEMI during the COVID-19 period was recorded. Demographic data, clinical characteristics, and primary angiographic findings were gathered for all patients. Time from symptom onset to hospital admission and time from hospital admission to reperfusion (defined as door-to-balloon time) were documented for each patient. Killip classification was used to assess patients’ clinical status on admission. Length of stay was determined as days from hospital admission to discharge or death (if occurring during the same hospitalization).

Adverse in-hospital complications were also recorded. These were selected based on inclusion of the following categories of acute STEMI complications: ischemic, mechanical, arrhythmic, embolic, and inflammatory. The following complications occurred in our patient cohort: sustained ventricular arrhythmia, congestive heart failure (CHF) defined as congestion requiring intravenous diuretics, re-infarction, mechanical complications (free-wall rupture, ventricular septal defect, or mitral regurgitation), second- or third-degree atrioventricular block, atrial fibrillation, stroke, mechanical ventilation, major bleeding, pericarditis, cardiogenic shock, cardiac arrest, and in-hospital mortality. The primary outcome of this study was defined as a composite of sustained ventricular arrhythmia, CHF with congestion requiring intravenous diuretics, and/or in-hospital mortality. Ventricular arrythmia and CHF were included in the composite outcome because they are defined as the 2 most common causes of sudden cardiac death following acute STEMI.^11,12

Statistical Analysis

Normally distributed continuous variables and categorical variables were compared using the paired t-test. A 2-sided P value <.05 was considered to be statistically significant. Mean admission rates for acute STEMI hospitalizations were determined by dividing the number of admissions by the number of days in each time period. The daily rate of COVID-19 cases per 100,000 individuals was obtained from the Centers for Disease Control and Prevention COVID-19 database. All data analyses were performed using Microsoft Excel. 

Results

The study cohort consisted of 64 patients, of whom 30 (46.9%) were hospitalized between March 5 and May 5, 2020, and 34 (53.1%) who were admitted during the analogous time period in 2019. This reflected a 6% decrease in STEMI admissions at PAR in the COVID-19 cohort. 

Acute STEMI Hospitalization Rates and COVID-19 Incidence

The mean daily acute STEMI admission rate was 0.50 during the study period compared to 0.57 during the control period. During the study period in 2020 in the state of Georgia, the daily rate of newly confirmed COVID-19 cases ranged from 0.194 per 100,000 on March 5 to 8.778 per 100,000 on May 5. Results of COVID-19 testing were available for 9 STEMI patients, and of these 0 tests were positive. 

Baseline Characteristics

Baseline characteristics of the acute STEMI cohorts are presented in Table 1. Approximately 75% were male; median (interquartile range [IQR]) age was 60 (51-72) years. There were no significant differences in age and gender between the study periods. Three-quarters of patients had a history of hypertension, and 87.5% had a history of dyslipidemia. There was no significant difference in baseline comorbidity profiles between the 2 study periods; therefore, our sample populations shared similar characteristics.

Clinical Presentation

Significant differences were observed regarding the time intervals of STEMI patients in the COVID-19 period and the control period (Table 2). Median time from symptom onset to hospital admission (patient delay) was extended from 57.5 minutes (IQR, 40.3-106) in 2019 to 93 minutes (IQR, 48.8-132) in 2020; however, this difference was not statistically significant (P = .697). Median time from hospital admission to reperfusion (system delay) was prolonged from 45 minutes (IQR, 28-61) in 2019 to 78 minutes (IQR, 50-110) in 2020 (P < .001). Overall time from symptom onset to reperfusion (total ischemic time) increased from 99.5 minutes (IQR, 84.8-132) in 2019 to 149 minutes (IQR, 96.3-231.8) in 2020 (P = .032). 

Regarding mode of transportation, 23.5% of patients in 2019 were walk-in admissions to the emergency department. During the COVID-19 period, walk-in admissions decreased to 6.7% (P = .065). There were no significant differences between emergency medical service, transfer, or in-patient admissions for STEMI cases between the 2 study periods. 

Killip classification scores were calculated for all patients on admission; 90.6% of patients were classified as Killip Class 1. There was no significant difference between hemodynamic presentations during the COVID-19 period compared to the control period. 

Angiographic Data

Overall, 53 (82.8%) patients admitted with acute STEMI underwent coronary angiography during their hospital stay. The proportion of patients who underwent primary reperfusion was greater in the control period than in the COVID-19 period (85.3% vs 80%; P = .582). Angiographic characteristics and findings were similar between the 2 study groups (Table 2).

In-Hospital Outcomes

In-hospital outcome data were available for all patients. As shown in Table 3, hospitalization during the COVID-19 period was independently associated with an increased risk for combined in-hospital outcome (odds ratio, 3.96; P = .046). The rate of in-hospital mortality was greater in the COVID-19 period (P = .013). We found no significant difference when comparing secondary outcomes from admissions during the COVID-19 period and the control period in 2019. For the 5 patients who died during the study period, the primary diagnosis at death was acute STEMI complicated by CHF (3 patients) or cardiogenic shock (2 patients).

This single-center retrospective study at PAR looks at the impact of COVID-19 on hospitalizations for acute STEMI during the initial peak of the pandemic. The key findings of this study show a significant increase in ischemic time parameters (symptom onset to reperfusion, hospital admission to reperfusion), in-hospital mortality, and combined in-hospital outcomes.

There was a 49.5-minute increase in total ischemic time noted in this study (P = .032). Though there was a numerical increase in time of symptom onset to hospital admission by 23.5 minutes, this difference was not statistically significant (P = .697). However, this study observed a statistically significant 33-minute increase in ischemic time from hospital admission to reperfusion (P < .001). Multiple studies globally have found a similar increase in total ischemic times, including those conducted in China and Europe.^13-15 Every level of potential delay must be considered, including pre-hospital, triage and emergency department, and/or reperfusion team. Pre-hospital sources of delays that have been suggested include “stay-at-home” orders and apprehension to seek medical care due to concern about contracting the virus or overwhelming the health care facilities. There was a clinically significant 4-fold decrease in the number of walk-in acute STEMI cases in the study period. In 2019, there were 8 walk-in cases compared to 2 cases in 2020 (P = .065). However, this change was not statistically significant. In-hospital/systemic sources of delays have been mentioned in other studies; they include increased time taken to rule out COVID-19 (nasopharyngeal swab/chest x-ray) and increased time due to the need for intensive gowning and gloving procedures by staff. It was difficult to objectively determine the sources of system delay by the reperfusion team due to a lack of quantitative data.

In the current study, we found a significant increase in in-hospital mortality during the COVID-19 period compared to a parallel time frame in 2019. This finding is contrary to a multicenter study from Spain that reported no difference in in-hospital outcomes or mortality rates among all acute coronary syndrome cases.¹⁶ The worsening outcomes and prognosis may simply be a result of increased ischemic time; however, the virus that causes COVID-19 itself may play a role as well. Studies have found that SARS-Cov-2 infection places patients at greater risk for cardiovascular conditions such as hypercoagulability, myocarditis, and arrhythmias.¹⁷ In our study, however, there were no acute STEMI patients who tested positive for COVID-19. Therefore, we cannot discuss the impact of increased thrombus burden in patients with COVID-19. Piedmont Healthcare published a STEMI treatment protocol in May 2020 that advised increased use of tissue plasminogen activator (tPA) in COVID-19-positive cases; during the study period, however, there were no occasions when tPA use was deemed appropriate based on clinical judgment.

Our findings align with previous studies that describe an increase in combined in-hospital adverse outcomes during the COVID-19 era. Previous studies detected a higher rate of complications in the COVID-19 cohort, but in the current study, the adverse in-hospital course is unrelated to underlying infection.^18,19 This study reports a higher incidence of major in-hospital outcomes, including a 65% increase in the rate of combined in-hospital outcomes, which is similar to a multicenter study conducted in Israel.¹⁹ There was a 2.3-fold numerical increase in sustained ventricular arrhythmias and a 2.5-fold numerical increase in the incidence of cardiac arrest in the study period. This phenomenon was observed despite a similar rate of reperfusion procedures in both groups. 

Acute STEMI is a highly fatal condition with an incidence of 8.5 in 10,000 annually in the United States. While studies across the world have shown a 25% to 40% reduction in the rate of hospitalized acute coronary syndrome cases during the COVID-19 pandemic, the decrease from 34 to 30 STEMI admissions at PAR is not statistically significant.²⁰ Possible reasons for the reduction globally include increased out-of-hospital mortality and decreased incidence of acute STEMI across the general population as a result of improved access to telemedicine or decreased levels of life stressors.²⁰  

In summary, there was an increase in ischemic time to reperfusion, in-hospital mortality, and combined in-hospital outcomes for acute STEMI patients at PAR during the COVID period.  

Limitations

This study has several limitations. This is a single-center study, so the sample size is small and may not be generalizable to a larger population. This is a retrospective observational study, so causation cannot be inferred. This study analyzed ischemic time parameters as average rates over time rather than in an interrupted time series. Post-reperfusion outcomes were limited to hospital stay. Post-hospital follow-up would provide a better picture of the effects of STEMI intervention. There is no account of patients who died out-of-hospital secondary to acute STEMI. COVID-19 testing was not introduced until midway in our study period. Therefore, we cannot rule out the possibility of the SARS-Cov-2 virus inciting acute STEMI and subsequently leading to worse outcomes and poor prognosis. 

Conclusions

This study provides an analysis of the incidence, characteristics, and clinical outcomes of patients presenting with acute STEMI during the early period of the COVID-19 pandemic. In-hospital mortality and ischemic time to reperfusion increased while combined in-hospital outcomes worsened. 

Acknowledgment: The authors thank Piedmont Athens Regional IRB for approving this project and allowing access to patient data.

Corresponding author: Syed H. Ali; Department of Medicine, Medical College of Georgia at the Augusta University-University of Georgia Medical Partnership, 30606, Athens, GA; syedha.ali@gmail.com

Disclosures: None reported.

doi:10.12788/jcom.0085

From the Department of Medicine, Medical College of Georgia at the Augusta University-University of Georgia Medical Partnership, Athens, GA (Syed H. Ali, Syed Hyder, and Dr. Murrow), and the Department of Cardiology, Piedmont Heart Institute, Piedmont Athens Regional, Athens, GA (Dr. Murrow and Mrs. Davis).

Abstract

Objectives: The aim of this study was to describe the characteristics and in-hospital outcomes of patients with acute ST-segment elevation myocardial infarction (STEMI) during the early COVID-19 pandemic at Piedmont Athens Regional (PAR), a 330-bed tertiary referral center in Northeast Georgia. 

Methods: A retrospective study was conducted at PAR to evaluate patients with acute STEMI admitted over an 8-week period during the initial COVID-19 outbreak. This study group was compared to patients admitted during the corresponding period in 2019. The primary endpoint of this study was defined as a composite of sustained ventricular arrhythmia, congestive heart failure (CHF) with pulmonary congestion, and/or in-hospital mortality. 

Results: This study cohort was composed of 64 patients with acute STEMI; 30 patients (46.9%) were hospitalized during the COVID-19 pandemic. Patients with STEMI in both the COVID-19 and control groups had similar comorbidities, Killip classification score, and clinical presentations. The median (interquartile range) time from symptom onset to reperfusion (total ischemic time) increased from 99.5 minutes (84.8-132) in 2019 to 149 minutes (96.3-231.8; P = .032) in 2020. Hospitalization during the COVID-19 period was associated with an increased risk for combined in-hospital outcome (odds ratio, 3.96; P = .046). 

Conclusion: Patients with STEMI admitted during the first wave of the COVID-19 outbreak experienced longer total ischemic time and increased risk for combined in-hospital outcomes compared to patients admitted during the corresponding period in 2019. 

Keywords: myocardial infarction, acute coronary syndrome, hospitalization, outcomes.

The emergence of the SARS-Cov-2 virus in December 2019 caused a worldwide shift in resource allocation and the restructuring of health care systems within the span of a few months. With the rapid spread of infection, the World Health Organization officially declared a pandemic in March 2020. The pandemic led to the deferral and cancellation of in-person patient visits, routine diagnostic studies, and nonessential surgeries and procedures. This response occurred secondary to a joint effort to reduce transmission via stay-at-home mandates and appropriate social distancing.¹ 

Alongside the reduction in elective procedures and health care visits, significant reductions in hospitalization rates due to decreases in acute ST-segment elevation myocardial infarction (STEMI) and catheterization laboratory utilization have been reported in many studies from around the world.^2-7 Comprehensive data demonstrating the impact of the COVID-19 pandemic on acute STEMI patient characteristics, clinical presentation, and in-hospital outcomes are lacking. Although patients with previously diagnosed cardiovascular disease are more likely to encounter worse outcomes in the setting of COVID-19, there may also be an indirect impact of the pandemic on high-risk patients, including those without the infection.⁸ Several theories have been hypothesized to explain this phenomenon. One theory postulates that the fear of contracting the virus during hospitalization is great enough to prevent patients from seeking care.² Another theory suggests that the increased utilization of telemedicine prevents exacerbation of chronic conditions and the need for hospitalization.⁹ Contrary to this trend, previous studies have shown an increased incidence of acute STEMI following stressful events such as natural disasters.¹⁰ 

The aim of this study was to describe trends pertaining to clinical characteristics and in-hospital outcomes of patients with acute STEMI during the early COVID-19 pandemic at Piedmont Athens Regional (PAR), a 330-bed tertiary referral center in Northeast Georgia. 

Methods

A retrospective cohort study was conducted at PAR to evaluate patients with STEMI admitted to the cardiovascular intensive care unit over an 8-week period (March 5 to May 5, 2020) during the COVID-19 outbreak. COVID-19 was declared a national emergency on March 13, 2020, in the United States. The institutional review board at PAR approved the study; the need for individual consent was waived under the condition that participant data would undergo de-identification and be strictly safeguarded. 

Data Collection

Because there are seasonal variations in cardiovascular admissions, patient data from a control period (March 9 to May 9, 2019) were obtained to compare with data from the 2020 period. The number of patients with the diagnosis of acute STEMI during the COVID-19 period was recorded. Demographic data, clinical characteristics, and primary angiographic findings were gathered for all patients. Time from symptom onset to hospital admission and time from hospital admission to reperfusion (defined as door-to-balloon time) were documented for each patient. Killip classification was used to assess patients’ clinical status on admission. Length of stay was determined as days from hospital admission to discharge or death (if occurring during the same hospitalization).

Adverse in-hospital complications were also recorded. These were selected based on inclusion of the following categories of acute STEMI complications: ischemic, mechanical, arrhythmic, embolic, and inflammatory. The following complications occurred in our patient cohort: sustained ventricular arrhythmia, congestive heart failure (CHF) defined as congestion requiring intravenous diuretics, re-infarction, mechanical complications (free-wall rupture, ventricular septal defect, or mitral regurgitation), second- or third-degree atrioventricular block, atrial fibrillation, stroke, mechanical ventilation, major bleeding, pericarditis, cardiogenic shock, cardiac arrest, and in-hospital mortality. The primary outcome of this study was defined as a composite of sustained ventricular arrhythmia, CHF with congestion requiring intravenous diuretics, and/or in-hospital mortality. Ventricular arrythmia and CHF were included in the composite outcome because they are defined as the 2 most common causes of sudden cardiac death following acute STEMI.^11,12

Statistical Analysis

Normally distributed continuous variables and categorical variables were compared using the paired t-test. A 2-sided P value <.05 was considered to be statistically significant. Mean admission rates for acute STEMI hospitalizations were determined by dividing the number of admissions by the number of days in each time period. The daily rate of COVID-19 cases per 100,000 individuals was obtained from the Centers for Disease Control and Prevention COVID-19 database. All data analyses were performed using Microsoft Excel. 

Results

The study cohort consisted of 64 patients, of whom 30 (46.9%) were hospitalized between March 5 and May 5, 2020, and 34 (53.1%) who were admitted during the analogous time period in 2019. This reflected a 6% decrease in STEMI admissions at PAR in the COVID-19 cohort. 

Acute STEMI Hospitalization Rates and COVID-19 Incidence

The mean daily acute STEMI admission rate was 0.50 during the study period compared to 0.57 during the control period. During the study period in 2020 in the state of Georgia, the daily rate of newly confirmed COVID-19 cases ranged from 0.194 per 100,000 on March 5 to 8.778 per 100,000 on May 5. Results of COVID-19 testing were available for 9 STEMI patients, and of these 0 tests were positive. 

Baseline Characteristics

Baseline characteristics of the acute STEMI cohorts are presented in Table 1. Approximately 75% were male; median (interquartile range [IQR]) age was 60 (51-72) years. There were no significant differences in age and gender between the study periods. Three-quarters of patients had a history of hypertension, and 87.5% had a history of dyslipidemia. There was no significant difference in baseline comorbidity profiles between the 2 study periods; therefore, our sample populations shared similar characteristics.

Clinical Presentation

Significant differences were observed regarding the time intervals of STEMI patients in the COVID-19 period and the control period (Table 2). Median time from symptom onset to hospital admission (patient delay) was extended from 57.5 minutes (IQR, 40.3-106) in 2019 to 93 minutes (IQR, 48.8-132) in 2020; however, this difference was not statistically significant (P = .697). Median time from hospital admission to reperfusion (system delay) was prolonged from 45 minutes (IQR, 28-61) in 2019 to 78 minutes (IQR, 50-110) in 2020 (P < .001). Overall time from symptom onset to reperfusion (total ischemic time) increased from 99.5 minutes (IQR, 84.8-132) in 2019 to 149 minutes (IQR, 96.3-231.8) in 2020 (P = .032). 

Regarding mode of transportation, 23.5% of patients in 2019 were walk-in admissions to the emergency department. During the COVID-19 period, walk-in admissions decreased to 6.7% (P = .065). There were no significant differences between emergency medical service, transfer, or in-patient admissions for STEMI cases between the 2 study periods. 

Killip classification scores were calculated for all patients on admission; 90.6% of patients were classified as Killip Class 1. There was no significant difference between hemodynamic presentations during the COVID-19 period compared to the control period. 

Angiographic Data

Overall, 53 (82.8%) patients admitted with acute STEMI underwent coronary angiography during their hospital stay. The proportion of patients who underwent primary reperfusion was greater in the control period than in the COVID-19 period (85.3% vs 80%; P = .582). Angiographic characteristics and findings were similar between the 2 study groups (Table 2).

In-Hospital Outcomes

In-hospital outcome data were available for all patients. As shown in Table 3, hospitalization during the COVID-19 period was independently associated with an increased risk for combined in-hospital outcome (odds ratio, 3.96; P = .046). The rate of in-hospital mortality was greater in the COVID-19 period (P = .013). We found no significant difference when comparing secondary outcomes from admissions during the COVID-19 period and the control period in 2019. For the 5 patients who died during the study period, the primary diagnosis at death was acute STEMI complicated by CHF (3 patients) or cardiogenic shock (2 patients).

This single-center retrospective study at PAR looks at the impact of COVID-19 on hospitalizations for acute STEMI during the initial peak of the pandemic. The key findings of this study show a significant increase in ischemic time parameters (symptom onset to reperfusion, hospital admission to reperfusion), in-hospital mortality, and combined in-hospital outcomes.

There was a 49.5-minute increase in total ischemic time noted in this study (P = .032). Though there was a numerical increase in time of symptom onset to hospital admission by 23.5 minutes, this difference was not statistically significant (P = .697). However, this study observed a statistically significant 33-minute increase in ischemic time from hospital admission to reperfusion (P < .001). Multiple studies globally have found a similar increase in total ischemic times, including those conducted in China and Europe.^13-15 Every level of potential delay must be considered, including pre-hospital, triage and emergency department, and/or reperfusion team. Pre-hospital sources of delays that have been suggested include “stay-at-home” orders and apprehension to seek medical care due to concern about contracting the virus or overwhelming the health care facilities. There was a clinically significant 4-fold decrease in the number of walk-in acute STEMI cases in the study period. In 2019, there were 8 walk-in cases compared to 2 cases in 2020 (P = .065). However, this change was not statistically significant. In-hospital/systemic sources of delays have been mentioned in other studies; they include increased time taken to rule out COVID-19 (nasopharyngeal swab/chest x-ray) and increased time due to the need for intensive gowning and gloving procedures by staff. It was difficult to objectively determine the sources of system delay by the reperfusion team due to a lack of quantitative data.

In the current study, we found a significant increase in in-hospital mortality during the COVID-19 period compared to a parallel time frame in 2019. This finding is contrary to a multicenter study from Spain that reported no difference in in-hospital outcomes or mortality rates among all acute coronary syndrome cases.¹⁶ The worsening outcomes and prognosis may simply be a result of increased ischemic time; however, the virus that causes COVID-19 itself may play a role as well. Studies have found that SARS-Cov-2 infection places patients at greater risk for cardiovascular conditions such as hypercoagulability, myocarditis, and arrhythmias.¹⁷ In our study, however, there were no acute STEMI patients who tested positive for COVID-19. Therefore, we cannot discuss the impact of increased thrombus burden in patients with COVID-19. Piedmont Healthcare published a STEMI treatment protocol in May 2020 that advised increased use of tissue plasminogen activator (tPA) in COVID-19-positive cases; during the study period, however, there were no occasions when tPA use was deemed appropriate based on clinical judgment.

Our findings align with previous studies that describe an increase in combined in-hospital adverse outcomes during the COVID-19 era. Previous studies detected a higher rate of complications in the COVID-19 cohort, but in the current study, the adverse in-hospital course is unrelated to underlying infection.^18,19 This study reports a higher incidence of major in-hospital outcomes, including a 65% increase in the rate of combined in-hospital outcomes, which is similar to a multicenter study conducted in Israel.¹⁹ There was a 2.3-fold numerical increase in sustained ventricular arrhythmias and a 2.5-fold numerical increase in the incidence of cardiac arrest in the study period. This phenomenon was observed despite a similar rate of reperfusion procedures in both groups. 

Acute STEMI is a highly fatal condition with an incidence of 8.5 in 10,000 annually in the United States. While studies across the world have shown a 25% to 40% reduction in the rate of hospitalized acute coronary syndrome cases during the COVID-19 pandemic, the decrease from 34 to 30 STEMI admissions at PAR is not statistically significant.²⁰ Possible reasons for the reduction globally include increased out-of-hospital mortality and decreased incidence of acute STEMI across the general population as a result of improved access to telemedicine or decreased levels of life stressors.²⁰  

In summary, there was an increase in ischemic time to reperfusion, in-hospital mortality, and combined in-hospital outcomes for acute STEMI patients at PAR during the COVID period.  

Limitations

This study has several limitations. This is a single-center study, so the sample size is small and may not be generalizable to a larger population. This is a retrospective observational study, so causation cannot be inferred. This study analyzed ischemic time parameters as average rates over time rather than in an interrupted time series. Post-reperfusion outcomes were limited to hospital stay. Post-hospital follow-up would provide a better picture of the effects of STEMI intervention. There is no account of patients who died out-of-hospital secondary to acute STEMI. COVID-19 testing was not introduced until midway in our study period. Therefore, we cannot rule out the possibility of the SARS-Cov-2 virus inciting acute STEMI and subsequently leading to worse outcomes and poor prognosis. 

Conclusions

This study provides an analysis of the incidence, characteristics, and clinical outcomes of patients presenting with acute STEMI during the early period of the COVID-19 pandemic. In-hospital mortality and ischemic time to reperfusion increased while combined in-hospital outcomes worsened. 

Acknowledgment: The authors thank Piedmont Athens Regional IRB for approving this project and allowing access to patient data.

Corresponding author: Syed H. Ali; Department of Medicine, Medical College of Georgia at the Augusta University-University of Georgia Medical Partnership, 30606, Athens, GA; syedha.ali@gmail.com

Disclosures: None reported.

doi:10.12788/jcom.0085

From the Department of Medicine, Medical College of Georgia at the Augusta University-University of Georgia Medical Partnership, Athens, GA (Syed H. Ali, Syed Hyder, and Dr. Murrow), and the Department of Cardiology, Piedmont Heart Institute, Piedmont Athens Regional, Athens, GA (Dr. Murrow and Mrs. Davis).

Abstract

Objectives: The aim of this study was to describe the characteristics and in-hospital outcomes of patients with acute ST-segment elevation myocardial infarction (STEMI) during the early COVID-19 pandemic at Piedmont Athens Regional (PAR), a 330-bed tertiary referral center in Northeast Georgia. 

Methods: A retrospective study was conducted at PAR to evaluate patients with acute STEMI admitted over an 8-week period during the initial COVID-19 outbreak. This study group was compared to patients admitted during the corresponding period in 2019. The primary endpoint of this study was defined as a composite of sustained ventricular arrhythmia, congestive heart failure (CHF) with pulmonary congestion, and/or in-hospital mortality. 

Results: This study cohort was composed of 64 patients with acute STEMI; 30 patients (46.9%) were hospitalized during the COVID-19 pandemic. Patients with STEMI in both the COVID-19 and control groups had similar comorbidities, Killip classification score, and clinical presentations. The median (interquartile range) time from symptom onset to reperfusion (total ischemic time) increased from 99.5 minutes (84.8-132) in 2019 to 149 minutes (96.3-231.8; P = .032) in 2020. Hospitalization during the COVID-19 period was associated with an increased risk for combined in-hospital outcome (odds ratio, 3.96; P = .046). 

Conclusion: Patients with STEMI admitted during the first wave of the COVID-19 outbreak experienced longer total ischemic time and increased risk for combined in-hospital outcomes compared to patients admitted during the corresponding period in 2019. 

Keywords: myocardial infarction, acute coronary syndrome, hospitalization, outcomes.

The emergence of the SARS-Cov-2 virus in December 2019 caused a worldwide shift in resource allocation and the restructuring of health care systems within the span of a few months. With the rapid spread of infection, the World Health Organization officially declared a pandemic in March 2020. The pandemic led to the deferral and cancellation of in-person patient visits, routine diagnostic studies, and nonessential surgeries and procedures. This response occurred secondary to a joint effort to reduce transmission via stay-at-home mandates and appropriate social distancing.¹ 

Alongside the reduction in elective procedures and health care visits, significant reductions in hospitalization rates due to decreases in acute ST-segment elevation myocardial infarction (STEMI) and catheterization laboratory utilization have been reported in many studies from around the world.^2-7 Comprehensive data demonstrating the impact of the COVID-19 pandemic on acute STEMI patient characteristics, clinical presentation, and in-hospital outcomes are lacking. Although patients with previously diagnosed cardiovascular disease are more likely to encounter worse outcomes in the setting of COVID-19, there may also be an indirect impact of the pandemic on high-risk patients, including those without the infection.⁸ Several theories have been hypothesized to explain this phenomenon. One theory postulates that the fear of contracting the virus during hospitalization is great enough to prevent patients from seeking care.² Another theory suggests that the increased utilization of telemedicine prevents exacerbation of chronic conditions and the need for hospitalization.⁹ Contrary to this trend, previous studies have shown an increased incidence of acute STEMI following stressful events such as natural disasters.¹⁰ 

The aim of this study was to describe trends pertaining to clinical characteristics and in-hospital outcomes of patients with acute STEMI during the early COVID-19 pandemic at Piedmont Athens Regional (PAR), a 330-bed tertiary referral center in Northeast Georgia. 

Methods

A retrospective cohort study was conducted at PAR to evaluate patients with STEMI admitted to the cardiovascular intensive care unit over an 8-week period (March 5 to May 5, 2020) during the COVID-19 outbreak. COVID-19 was declared a national emergency on March 13, 2020, in the United States. The institutional review board at PAR approved the study; the need for individual consent was waived under the condition that participant data would undergo de-identification and be strictly safeguarded. 

Data Collection

Because there are seasonal variations in cardiovascular admissions, patient data from a control period (March 9 to May 9, 2019) were obtained to compare with data from the 2020 period. The number of patients with the diagnosis of acute STEMI during the COVID-19 period was recorded. Demographic data, clinical characteristics, and primary angiographic findings were gathered for all patients. Time from symptom onset to hospital admission and time from hospital admission to reperfusion (defined as door-to-balloon time) were documented for each patient. Killip classification was used to assess patients’ clinical status on admission. Length of stay was determined as days from hospital admission to discharge or death (if occurring during the same hospitalization).

Adverse in-hospital complications were also recorded. These were selected based on inclusion of the following categories of acute STEMI complications: ischemic, mechanical, arrhythmic, embolic, and inflammatory. The following complications occurred in our patient cohort: sustained ventricular arrhythmia, congestive heart failure (CHF) defined as congestion requiring intravenous diuretics, re-infarction, mechanical complications (free-wall rupture, ventricular septal defect, or mitral regurgitation), second- or third-degree atrioventricular block, atrial fibrillation, stroke, mechanical ventilation, major bleeding, pericarditis, cardiogenic shock, cardiac arrest, and in-hospital mortality. The primary outcome of this study was defined as a composite of sustained ventricular arrhythmia, CHF with congestion requiring intravenous diuretics, and/or in-hospital mortality. Ventricular arrythmia and CHF were included in the composite outcome because they are defined as the 2 most common causes of sudden cardiac death following acute STEMI.^11,12

Statistical Analysis

Normally distributed continuous variables and categorical variables were compared using the paired t-test. A 2-sided P value <.05 was considered to be statistically significant. Mean admission rates for acute STEMI hospitalizations were determined by dividing the number of admissions by the number of days in each time period. The daily rate of COVID-19 cases per 100,000 individuals was obtained from the Centers for Disease Control and Prevention COVID-19 database. All data analyses were performed using Microsoft Excel. 

Results

The study cohort consisted of 64 patients, of whom 30 (46.9%) were hospitalized between March 5 and May 5, 2020, and 34 (53.1%) who were admitted during the analogous time period in 2019. This reflected a 6% decrease in STEMI admissions at PAR in the COVID-19 cohort. 

Acute STEMI Hospitalization Rates and COVID-19 Incidence

The mean daily acute STEMI admission rate was 0.50 during the study period compared to 0.57 during the control period. During the study period in 2020 in the state of Georgia, the daily rate of newly confirmed COVID-19 cases ranged from 0.194 per 100,000 on March 5 to 8.778 per 100,000 on May 5. Results of COVID-19 testing were available for 9 STEMI patients, and of these 0 tests were positive. 

Baseline Characteristics

Baseline characteristics of the acute STEMI cohorts are presented in Table 1. Approximately 75% were male; median (interquartile range [IQR]) age was 60 (51-72) years. There were no significant differences in age and gender between the study periods. Three-quarters of patients had a history of hypertension, and 87.5% had a history of dyslipidemia. There was no significant difference in baseline comorbidity profiles between the 2 study periods; therefore, our sample populations shared similar characteristics.

Clinical Presentation

Significant differences were observed regarding the time intervals of STEMI patients in the COVID-19 period and the control period (Table 2). Median time from symptom onset to hospital admission (patient delay) was extended from 57.5 minutes (IQR, 40.3-106) in 2019 to 93 minutes (IQR, 48.8-132) in 2020; however, this difference was not statistically significant (P = .697). Median time from hospital admission to reperfusion (system delay) was prolonged from 45 minutes (IQR, 28-61) in 2019 to 78 minutes (IQR, 50-110) in 2020 (P < .001). Overall time from symptom onset to reperfusion (total ischemic time) increased from 99.5 minutes (IQR, 84.8-132) in 2019 to 149 minutes (IQR, 96.3-231.8) in 2020 (P = .032). 

Regarding mode of transportation, 23.5% of patients in 2019 were walk-in admissions to the emergency department. During the COVID-19 period, walk-in admissions decreased to 6.7% (P = .065). There were no significant differences between emergency medical service, transfer, or in-patient admissions for STEMI cases between the 2 study periods. 

Killip classification scores were calculated for all patients on admission; 90.6% of patients were classified as Killip Class 1. There was no significant difference between hemodynamic presentations during the COVID-19 period compared to the control period. 

Angiographic Data

Overall, 53 (82.8%) patients admitted with acute STEMI underwent coronary angiography during their hospital stay. The proportion of patients who underwent primary reperfusion was greater in the control period than in the COVID-19 period (85.3% vs 80%; P = .582). Angiographic characteristics and findings were similar between the 2 study groups (Table 2).

In-Hospital Outcomes

In-hospital outcome data were available for all patients. As shown in Table 3, hospitalization during the COVID-19 period was independently associated with an increased risk for combined in-hospital outcome (odds ratio, 3.96; P = .046). The rate of in-hospital mortality was greater in the COVID-19 period (P = .013). We found no significant difference when comparing secondary outcomes from admissions during the COVID-19 period and the control period in 2019. For the 5 patients who died during the study period, the primary diagnosis at death was acute STEMI complicated by CHF (3 patients) or cardiogenic shock (2 patients).

This single-center retrospective study at PAR looks at the impact of COVID-19 on hospitalizations for acute STEMI during the initial peak of the pandemic. The key findings of this study show a significant increase in ischemic time parameters (symptom onset to reperfusion, hospital admission to reperfusion), in-hospital mortality, and combined in-hospital outcomes.

There was a 49.5-minute increase in total ischemic time noted in this study (P = .032). Though there was a numerical increase in time of symptom onset to hospital admission by 23.5 minutes, this difference was not statistically significant (P = .697). However, this study observed a statistically significant 33-minute increase in ischemic time from hospital admission to reperfusion (P < .001). Multiple studies globally have found a similar increase in total ischemic times, including those conducted in China and Europe.^13-15 Every level of potential delay must be considered, including pre-hospital, triage and emergency department, and/or reperfusion team. Pre-hospital sources of delays that have been suggested include “stay-at-home” orders and apprehension to seek medical care due to concern about contracting the virus or overwhelming the health care facilities. There was a clinically significant 4-fold decrease in the number of walk-in acute STEMI cases in the study period. In 2019, there were 8 walk-in cases compared to 2 cases in 2020 (P = .065). However, this change was not statistically significant. In-hospital/systemic sources of delays have been mentioned in other studies; they include increased time taken to rule out COVID-19 (nasopharyngeal swab/chest x-ray) and increased time due to the need for intensive gowning and gloving procedures by staff. It was difficult to objectively determine the sources of system delay by the reperfusion team due to a lack of quantitative data.

In the current study, we found a significant increase in in-hospital mortality during the COVID-19 period compared to a parallel time frame in 2019. This finding is contrary to a multicenter study from Spain that reported no difference in in-hospital outcomes or mortality rates among all acute coronary syndrome cases.¹⁶ The worsening outcomes and prognosis may simply be a result of increased ischemic time; however, the virus that causes COVID-19 itself may play a role as well. Studies have found that SARS-Cov-2 infection places patients at greater risk for cardiovascular conditions such as hypercoagulability, myocarditis, and arrhythmias.¹⁷ In our study, however, there were no acute STEMI patients who tested positive for COVID-19. Therefore, we cannot discuss the impact of increased thrombus burden in patients with COVID-19. Piedmont Healthcare published a STEMI treatment protocol in May 2020 that advised increased use of tissue plasminogen activator (tPA) in COVID-19-positive cases; during the study period, however, there were no occasions when tPA use was deemed appropriate based on clinical judgment.

Our findings align with previous studies that describe an increase in combined in-hospital adverse outcomes during the COVID-19 era. Previous studies detected a higher rate of complications in the COVID-19 cohort, but in the current study, the adverse in-hospital course is unrelated to underlying infection.^18,19 This study reports a higher incidence of major in-hospital outcomes, including a 65% increase in the rate of combined in-hospital outcomes, which is similar to a multicenter study conducted in Israel.¹⁹ There was a 2.3-fold numerical increase in sustained ventricular arrhythmias and a 2.5-fold numerical increase in the incidence of cardiac arrest in the study period. This phenomenon was observed despite a similar rate of reperfusion procedures in both groups. 

Acute STEMI is a highly fatal condition with an incidence of 8.5 in 10,000 annually in the United States. While studies across the world have shown a 25% to 40% reduction in the rate of hospitalized acute coronary syndrome cases during the COVID-19 pandemic, the decrease from 34 to 30 STEMI admissions at PAR is not statistically significant.²⁰ Possible reasons for the reduction globally include increased out-of-hospital mortality and decreased incidence of acute STEMI across the general population as a result of improved access to telemedicine or decreased levels of life stressors.²⁰  

In summary, there was an increase in ischemic time to reperfusion, in-hospital mortality, and combined in-hospital outcomes for acute STEMI patients at PAR during the COVID period.  

Limitations

This study has several limitations. This is a single-center study, so the sample size is small and may not be generalizable to a larger population. This is a retrospective observational study, so causation cannot be inferred. This study analyzed ischemic time parameters as average rates over time rather than in an interrupted time series. Post-reperfusion outcomes were limited to hospital stay. Post-hospital follow-up would provide a better picture of the effects of STEMI intervention. There is no account of patients who died out-of-hospital secondary to acute STEMI. COVID-19 testing was not introduced until midway in our study period. Therefore, we cannot rule out the possibility of the SARS-Cov-2 virus inciting acute STEMI and subsequently leading to worse outcomes and poor prognosis. 

Conclusions

This study provides an analysis of the incidence, characteristics, and clinical outcomes of patients presenting with acute STEMI during the early period of the COVID-19 pandemic. In-hospital mortality and ischemic time to reperfusion increased while combined in-hospital outcomes worsened. 

Acknowledgment: The authors thank Piedmont Athens Regional IRB for approving this project and allowing access to patient data.

Corresponding author: Syed H. Ali; Department of Medicine, Medical College of Georgia at the Augusta University-University of Georgia Medical Partnership, 30606, Athens, GA; syedha.ali@gmail.com

Disclosures: None reported.

doi:10.12788/jcom.0085

From Lahey Hospital and Medical Center, Burlington, MA (Drs. Liesching and Lei), and Tufts University School of Medicine, Boston, MA (Dr. Liesching)

ABSTRACT

Objective: To compare the utilization of oxygen therapies and clinical outcomes of patients admitted for COVID-19 during the second surge of the pandemic to that of patients admitted during the first surge.

Design: Observational study using a registry database.

Setting: Three hospitals (791 inpatient beds and 76 intensive care unit [ICU] beds) within the Beth Israel Lahey Health system in Massachusetts.

Participants: We included 3183 patients with COVID-19 admitted to hospitals.

Measurements: Baseline data included demographics and comorbidities. Treatments included low-flow supplemental oxygen (2-6 L/min), high-flow oxygen via nasal cannula, and invasive mechanical ventilation. Outcomes included ICU admission, length of stay, ventilator days, and mortality.

Results: A total of 3183 patients were included: 1586 during the first surge and 1597 during the second surge. Compared to the first surge, patients admitted during the second surge had a similar rate of receiving low-flow supplemental oxygen (65.8% vs 64.1%, P = .3), a higher rate of receiving high-flow nasal cannula (15.4% vs 10.8%, P = .0001), and a lower ventilation rate (5.6% vs 9.7%, P < .0001). The outcomes during the second surge were better than those during the first surge: lower ICU admission rate (8.1% vs 12.7%, P < .0001), shorter length of hospital stay (5 vs 6 days, P < .0001), fewer ventilator days (10 vs 16, P = .01), and lower mortality (8.3% vs 19.2%, P < .0001). Among ventilated patients, those who received high-flow nasal cannula had lower mortality.

Conclusion: Compared to the first surge of the COVID-19 pandemic, patients admitted during the second surge had similar likelihood of receiving low-flow supplemental oxygen, were more likely to receive high-flow nasal cannula, were less likely to be ventilated, and had better outcomes.

Keywords: supplemental oxygen, high-flow nasal cannula, ventilator.

The respiratory system receives the major impact of SARS-CoV-2 virus, and hypoxemia has been the predominant diagnosis for patients hospitalized with COVID-19.^1,2 During the initial stage of the pandemic, oxygen therapies and mechanical ventilation were the only choices for these patients.^3-6 Standard-of-care treatment for patients with COVID-19 during the initial surge included oxygen therapies and mechanical ventilation for hypoxemia and medications for comorbidities and COVID-19–associated sequelae, such as multi-organ dysfunction and failure. A report from New York during the first surge (May 2020) showed that among 5700 hospitalized patients with COVID-19, 27.8% received supplemental oxygen and 12.2% received invasive mechanical ventilation.⁷ High-flow nasal cannula (HFNC) oxygen delivery has been utilized widely throughout the pandemic due to its superiority over other noninvasive respiratory support techniques.^8-12 Mechanical ventilation is always necessary for critically ill patients with acute respiratory distress syndrome. However, ventilator scarcity has become a bottleneck in caring for severely ill patients with COVID-19 during the pandemic.¹³

The clinical outcomes of hospitalized COVID-19 patients include a high intubation rate, long length of hospital and intensive care unit (ICU) stay, and high mortality.^14,15 As the pandemic evolved, new medications, including remdesivir, hydroxychloroquine, lopinavir, or interferon β-1a, were used in addition to the standard of care, but these did not result in significantly different mortality from standard of care.¹⁶ Steroids are becoming foundational to the treatment of severe COVID-19 pneumonia, but evidence from high-quality randomized controlled clinical trials is lacking.¹⁷ 

During the first surge from March to May 2020, Massachusetts had the third highest number of COVID-19 cases among states in the United States.¹⁸ In early 2021, COVID-19 cases were climbing close to the peak of the second surge in Massachusetts. In this study, we compared utilization of low-flow supplemental oxygen, HFNC, and mechanical ventilation and clinical outcomes of patients admitted to 3 hospitals in Massachusetts during the second surge of the pandemic to that of patients admitted during the first surge.

Methods

Setting

Beth Israel Lahey Health is a system of academic and teaching hospitals with primary care and specialty care providers. We included 3 centers within the Beth Israel Lahey Health system in Massachusetts: Lahey Hospital and Medical Center, with 335 inpatient hospital beds and 52 critical care beds; Beverly Hospital, with 227 beds and 14 critical care beds; and Winchester Hospital, with 229 beds and 10 ICU beds.

Participants

We included patients admitted to the 3 hospitals with COVID-19 as a primary or secondary diagnosis during the first surge of the pandemic (March 1, 2020 to June 15, 2020) and the second surge (November 15, 2020 to January 27, 2021). The timeframe of the first surge was defined as the window between the start date and the end date of data collection. During the time window of the first surge, 1586 patients were included. The start time of the second surge was defined as the date when the data collection was restarted; the end date was set when the number of patients (1597) accumulated was close to the number of patients in the first surge (1586), so that the two groups had similar sample size.

Study Design

A data registry of COVID-19 patients was created by our institution, and the data were prospectively collected starting in March 2020. We retrospectively extracted data on the following from the registry database for this observational study: demographics and baseline comorbidities; the use of low-flow supplemental oxygen, HFNC, and invasive mechanical ventilator; and ICU admission, length of hospital stay, length of ICU stay, and hospital discharge disposition. Start and end times for each oxygen therapy were not entered in the registry. Data about other oxygen therapies, such as noninvasive positive-pressure ventilation, were not collected in the registry database, and therefore were not included in the analysis.

Statistical Analysis

Continuous variables (eg, age) were tested for data distribution normality using the Shapiro-Wilk test. Normally distributed data were tested using unpaired t-tests and displayed as mean (SD). The skewed data were tested using the Wilcoxon rank sum test and displayed as median (interquartile range [IQR]). The categorical variables were compared using chi-square test. Comparisons with P ≤ .05 were considered significantly different. Statistical analysis for this study was generated using Statistical Analysis Software (SAS), version 9.4 for Windows (SAS Institute Inc.).

Results

Baseline Characteristics

We included 3183 patients: 1586 admitted during the first surge and 1597 admitted during the second surge. Baseline characteristics of patients with COVID-19 admitted during the first and second surges are shown in Table 1. Patients admitted during the second surge were older (73 years vs 71 years, P = .01) and had higher rates of hypertension (64.8% vs 59.6%, P = .003) and asthma (12.9% vs 10.7%, P = .049) but a lower rate of interstitial lung disease (3.3% vs 7.7%, P < .001). Sequential organ failure assessment scores at admission and the rates of other comorbidities were not significantly different between the 2 surges.

The number of patients who were hospitalized and received low-flow supplemental oxygen, and/or HFNC, and/or ventilator in the first surge and the second surge is shown in the Figure. Of all patients included, 2067 (64.9%) received low-flow supplemental oxygen; of these, 374 (18.1%) subsequently received HFNC, and 85 (22.7%) of these subsequently received mechanical ventilation. Of all 3183 patients, 417 (13.1%) received HFNC; 43 of these patients received HFNC without receiving low-flow supplemental oxygen, and 98 (23.5%) subsequently received mechanical ventilation. Out of all 3183 patients, 244 (7.7%) received mechanical ventilation; 98 (40.2%) of these received HFNC while the remaining 146 (59.8%) did not. At the beginning of the first surge, the ratio of patients who received invasive mechanical ventilation to patients who received HFNC was close to 1:1 (10/10); the ratio decreased to 6:10 in May and June 2020. At the beginning of the second surge, the ratio was 8:10 and then decreased to 3:10 in December 2020 and January 2021.

As shown in Table 2, the proportion of patients who received low-flow supplemental oxygen during the second surge was similar to that during the first surge (65.8% vs 64.1%, P = .3). Patients admitted during the second surge were more likely to receive HFNC than patients admitted during the first surge (15.4% vs 10.8%, P = .0001). Patients admitted during the second surge were less likely to be ventilated than the patients admitted during the first surge (5.6% vs 9.7%, P < .0001).

Clinical Outcomes

As shown in Table 3, second surge outcomes were much better than first surge outcomes: the ICU admission rate was lower (8.1% vs 12.7%, P < .0001); patients were more likely to be discharged to home (60.2% vs 47.4%, P < .0001), had a shorter length of hospital stay (5 vs 6 days, P < .0001), and had fewer ventilator days (10 vs 16, P = .01); and mortality was lower (8.3% vs 19.2%, P < .0001). There was a trend that length of ICU stay was shorter during the second surge than during the first surge (7 days vs 9 days, P = .09).

As noted (Figure), the ratio of patients who received invasive mechanical ventilation to patients who received HFNC was decreasing during both the first surge and the second surge. To further analyze the relation between ventilator and HFNC, we performed a subgroup analysis for 244 ventilated patients during both surges to compare outcomes between patients who received HFNC and those who did not receive HFNC (Table 4). Ninety-eight (40%) patients received HFNC. Ventilated patients who received HFNC had lower mortality than those patients who did not receive HFNC (31.6% vs 48%, P = .01), but had a longer length of hospital stay (29 days vs 14 days, P < .0001), longer length of ICU stay (17 days vs 9 days, P < .0001), and a higher number of ventilator days (16 vs 11, P = .001).

Our study compared the baseline patient characteristics; utilization of low-flow supplemental oxygen therapy, HFNC, and mechanical ventilation; and clinical outcomes between the first surge (n = 1586) and the second surge (n = 1597) of the COVID-19 pandemic. During both surges, about two-thirds of admitted patients received low-flow supplemental oxygen. A higher proportion of the admitted patients received HFNC during the second surge than during the first surge, while the intubation rate was lower during the second surge than during the first surge.

Reported low-flow supplemental oxygen use ranged from 28% to 63% depending on the cohort characteristics and location during the first surge.^6,7,19 A report from New York during the first surge (March 1 to April 4, 2020) showed that among 5700 hospitalized patients with COVID-19, 27.8% received low-flow supplemental oxygen.⁷ HFNC is recommended in guidelines on management of patients with acute respiratory failure due to COVID-19.²⁰ In our study, HFNC was utilized in a higher proportion of patients admitted for COVID-19 during the second surge (15.5% vs 10.8%, P = .0001). During the early pandemic period in Wuhan, China, 11% to 21% of admitted COVID-19 patients received HFNC.^21,22 Utilization of HFNC in New York during the first surge (March to May 2020) varied from 5% to 14.3% of patients admitted with COVID-19.^23,24 Our subgroup analysis of the ventilated patients showed that patients who received HFNC had lower mortality than those who did not (31.6% vs 48.0%, P = .011). Comparably, a report from Paris, France, showed that among patients admitted to ICUs for acute hypoxemic respiratory failure, those who received HFNC had lower mortality at day 60 than those who did not (21% vs 31%, P = .052).²⁵ Our recent analysis showed that patients treated with HFNC prior to mechanical ventilation had lower mortality than those treated with only conventional oxygen (30% vs 52%, P = .05).²⁶ In this subgroup analysis, we could not determine if HFNC treatment was administered before or after ventilation because HFNC was entered as dichotomous data (“Yes” or “No”) in the registry database. We merely showed the beneficial effect of HFNC on reducing mortality for ventilated COVID-19 patients, but did not mean to focus on how and when to apply HFNC.

We observed that the patients admitted during the second surge were less likely to be ventilated than the patients admitted during the first surge (5.6% vs 9.7%, P < .0001). During the first surge in New York, among 5700 patients admitted with COVID-19, 12.2% received invasive mechanical ventilation.⁷ In another report, also from New York during the first surge, 26.1% of 2015 hospitalized COVID-19 patients received mechanical ventilation.²⁷ In our study, the ventilation rate of 9.7% during the first surge was lower.

Outcomes during the second surge were better than during the first surge, including ICU admission rate, hospital and ICU length of stay, ventilator days, and mortality. The mortality was 19.2% during the first surge vs 8.3% during the second surge (P < .0001). The mortality of 19.2% was lower than the 30.6% mortality reported for 2015 hospitalized COVID-19 patients in New York during the first surge.²⁷ A retrospective study showed that early administration of remdesivir was associated with reduced ICU admission, ventilation use, and mortality.²⁸ The RECOVERY clinical trial showed that dexamethasone improved mortality for COVID-19 patients who received respiratory support, but not for patients who did not receive any respiratory support.²⁹ Perhaps some, if not all, of the improvement in ICU admission and mortality during the second surge was attributed to the new medications, such as antivirals and steroids.

The length of hospital stay for patients with moderate to severe COVID-19 varied from 4 to 53 days at different locations of the world, as shown in a meta-analysis by Rees and colleagues.³⁰ Our results showing a length of stay of 6 days during the first surge and 5 days during the second surge fell into the shorter end of this range. In a retrospective analysis of 1643 adults with severe COVID-19 admitted to hospitals in New York City between March 9, 2020 and April 23, 2020, median hospital length of stay was 7 (IQR, 3-14) days.³¹ For the ventilated patients in our study, the length of stay of 14 days (did not receive HFNC) and 29 days (received HFNC) was much longer. This longer length of stay might be attributed to the patients in our study being older and having more severe comorbidities.

The main purpose of this study was to compare the oxygen therapies and outcomes between 2 surges. It is difficult to associate the clinical outcomes with the oxygen therapies because new therapies and medications were available after the first surge. It was not possible to adjust the outcomes with confounders (other therapies and medications) because the registry data did not include the new therapies and medications.

A strength of this study was that we included a large, balanced number of patients in the first surge and the second surge. We did not plan the sample size in both groups as we could not predict the number of admissions. We set the end date of data collection for analysis as the time when the number of patients admitted during the second surge was similar to the number of patients admitted during the first surge. A limitation was that the registry database was created by the institution and was not designed solely for this study. The data for oxygen therapies were limited to low-flow supplemental oxygen, HFNC, and invasive mechanical ventilation; data for noninvasive ventilation were not included.

Conclusion

At our centers, during the second surge of COVID-19 pandemic, patients hospitalized with COVID-19 infection were more likely to receive HFNC but less likely to be ventilated. Compared to the first surge, the hospitalized patients with COVID-19 infection had a lower ICU admission rate, shorter length of hospital stay, fewer ventilator days, and lower mortality. For ventilated patients, those who received HFNC had lower mortality than those who did not.

Corresponding author: Timothy N. Liesching, MD, 41 Mall Road, Burlington, MA 01805; Timothy.N.Liesching@lahey.org

Disclosures: None reported.

doi:10.12788/jcom.0086

From Lahey Hospital and Medical Center, Burlington, MA (Drs. Liesching and Lei), and Tufts University School of Medicine, Boston, MA (Dr. Liesching)

ABSTRACT

Objective: To compare the utilization of oxygen therapies and clinical outcomes of patients admitted for COVID-19 during the second surge of the pandemic to that of patients admitted during the first surge.

Design: Observational study using a registry database.

Setting: Three hospitals (791 inpatient beds and 76 intensive care unit [ICU] beds) within the Beth Israel Lahey Health system in Massachusetts.

Participants: We included 3183 patients with COVID-19 admitted to hospitals.

Measurements: Baseline data included demographics and comorbidities. Treatments included low-flow supplemental oxygen (2-6 L/min), high-flow oxygen via nasal cannula, and invasive mechanical ventilation. Outcomes included ICU admission, length of stay, ventilator days, and mortality.

Results: A total of 3183 patients were included: 1586 during the first surge and 1597 during the second surge. Compared to the first surge, patients admitted during the second surge had a similar rate of receiving low-flow supplemental oxygen (65.8% vs 64.1%, P = .3), a higher rate of receiving high-flow nasal cannula (15.4% vs 10.8%, P = .0001), and a lower ventilation rate (5.6% vs 9.7%, P < .0001). The outcomes during the second surge were better than those during the first surge: lower ICU admission rate (8.1% vs 12.7%, P < .0001), shorter length of hospital stay (5 vs 6 days, P < .0001), fewer ventilator days (10 vs 16, P = .01), and lower mortality (8.3% vs 19.2%, P < .0001). Among ventilated patients, those who received high-flow nasal cannula had lower mortality.

Conclusion: Compared to the first surge of the COVID-19 pandemic, patients admitted during the second surge had similar likelihood of receiving low-flow supplemental oxygen, were more likely to receive high-flow nasal cannula, were less likely to be ventilated, and had better outcomes.

Keywords: supplemental oxygen, high-flow nasal cannula, ventilator.

The respiratory system receives the major impact of SARS-CoV-2 virus, and hypoxemia has been the predominant diagnosis for patients hospitalized with COVID-19.^1,2 During the initial stage of the pandemic, oxygen therapies and mechanical ventilation were the only choices for these patients.^3-6 Standard-of-care treatment for patients with COVID-19 during the initial surge included oxygen therapies and mechanical ventilation for hypoxemia and medications for comorbidities and COVID-19–associated sequelae, such as multi-organ dysfunction and failure. A report from New York during the first surge (May 2020) showed that among 5700 hospitalized patients with COVID-19, 27.8% received supplemental oxygen and 12.2% received invasive mechanical ventilation.⁷ High-flow nasal cannula (HFNC) oxygen delivery has been utilized widely throughout the pandemic due to its superiority over other noninvasive respiratory support techniques.^8-12 Mechanical ventilation is always necessary for critically ill patients with acute respiratory distress syndrome. However, ventilator scarcity has become a bottleneck in caring for severely ill patients with COVID-19 during the pandemic.¹³

The clinical outcomes of hospitalized COVID-19 patients include a high intubation rate, long length of hospital and intensive care unit (ICU) stay, and high mortality.^14,15 As the pandemic evolved, new medications, including remdesivir, hydroxychloroquine, lopinavir, or interferon β-1a, were used in addition to the standard of care, but these did not result in significantly different mortality from standard of care.¹⁶ Steroids are becoming foundational to the treatment of severe COVID-19 pneumonia, but evidence from high-quality randomized controlled clinical trials is lacking.¹⁷ 

During the first surge from March to May 2020, Massachusetts had the third highest number of COVID-19 cases among states in the United States.¹⁸ In early 2021, COVID-19 cases were climbing close to the peak of the second surge in Massachusetts. In this study, we compared utilization of low-flow supplemental oxygen, HFNC, and mechanical ventilation and clinical outcomes of patients admitted to 3 hospitals in Massachusetts during the second surge of the pandemic to that of patients admitted during the first surge.

Methods

Setting

Beth Israel Lahey Health is a system of academic and teaching hospitals with primary care and specialty care providers. We included 3 centers within the Beth Israel Lahey Health system in Massachusetts: Lahey Hospital and Medical Center, with 335 inpatient hospital beds and 52 critical care beds; Beverly Hospital, with 227 beds and 14 critical care beds; and Winchester Hospital, with 229 beds and 10 ICU beds.

Participants

We included patients admitted to the 3 hospitals with COVID-19 as a primary or secondary diagnosis during the first surge of the pandemic (March 1, 2020 to June 15, 2020) and the second surge (November 15, 2020 to January 27, 2021). The timeframe of the first surge was defined as the window between the start date and the end date of data collection. During the time window of the first surge, 1586 patients were included. The start time of the second surge was defined as the date when the data collection was restarted; the end date was set when the number of patients (1597) accumulated was close to the number of patients in the first surge (1586), so that the two groups had similar sample size.

Study Design

A data registry of COVID-19 patients was created by our institution, and the data were prospectively collected starting in March 2020. We retrospectively extracted data on the following from the registry database for this observational study: demographics and baseline comorbidities; the use of low-flow supplemental oxygen, HFNC, and invasive mechanical ventilator; and ICU admission, length of hospital stay, length of ICU stay, and hospital discharge disposition. Start and end times for each oxygen therapy were not entered in the registry. Data about other oxygen therapies, such as noninvasive positive-pressure ventilation, were not collected in the registry database, and therefore were not included in the analysis.

Statistical Analysis

Continuous variables (eg, age) were tested for data distribution normality using the Shapiro-Wilk test. Normally distributed data were tested using unpaired t-tests and displayed as mean (SD). The skewed data were tested using the Wilcoxon rank sum test and displayed as median (interquartile range [IQR]). The categorical variables were compared using chi-square test. Comparisons with P ≤ .05 were considered significantly different. Statistical analysis for this study was generated using Statistical Analysis Software (SAS), version 9.4 for Windows (SAS Institute Inc.).

Results

Baseline Characteristics

We included 3183 patients: 1586 admitted during the first surge and 1597 admitted during the second surge. Baseline characteristics of patients with COVID-19 admitted during the first and second surges are shown in Table 1. Patients admitted during the second surge were older (73 years vs 71 years, P = .01) and had higher rates of hypertension (64.8% vs 59.6%, P = .003) and asthma (12.9% vs 10.7%, P = .049) but a lower rate of interstitial lung disease (3.3% vs 7.7%, P < .001). Sequential organ failure assessment scores at admission and the rates of other comorbidities were not significantly different between the 2 surges.

The number of patients who were hospitalized and received low-flow supplemental oxygen, and/or HFNC, and/or ventilator in the first surge and the second surge is shown in the Figure. Of all patients included, 2067 (64.9%) received low-flow supplemental oxygen; of these, 374 (18.1%) subsequently received HFNC, and 85 (22.7%) of these subsequently received mechanical ventilation. Of all 3183 patients, 417 (13.1%) received HFNC; 43 of these patients received HFNC without receiving low-flow supplemental oxygen, and 98 (23.5%) subsequently received mechanical ventilation. Out of all 3183 patients, 244 (7.7%) received mechanical ventilation; 98 (40.2%) of these received HFNC while the remaining 146 (59.8%) did not. At the beginning of the first surge, the ratio of patients who received invasive mechanical ventilation to patients who received HFNC was close to 1:1 (10/10); the ratio decreased to 6:10 in May and June 2020. At the beginning of the second surge, the ratio was 8:10 and then decreased to 3:10 in December 2020 and January 2021.

As shown in Table 2, the proportion of patients who received low-flow supplemental oxygen during the second surge was similar to that during the first surge (65.8% vs 64.1%, P = .3). Patients admitted during the second surge were more likely to receive HFNC than patients admitted during the first surge (15.4% vs 10.8%, P = .0001). Patients admitted during the second surge were less likely to be ventilated than the patients admitted during the first surge (5.6% vs 9.7%, P < .0001).

Clinical Outcomes

As shown in Table 3, second surge outcomes were much better than first surge outcomes: the ICU admission rate was lower (8.1% vs 12.7%, P < .0001); patients were more likely to be discharged to home (60.2% vs 47.4%, P < .0001), had a shorter length of hospital stay (5 vs 6 days, P < .0001), and had fewer ventilator days (10 vs 16, P = .01); and mortality was lower (8.3% vs 19.2%, P < .0001). There was a trend that length of ICU stay was shorter during the second surge than during the first surge (7 days vs 9 days, P = .09).

As noted (Figure), the ratio of patients who received invasive mechanical ventilation to patients who received HFNC was decreasing during both the first surge and the second surge. To further analyze the relation between ventilator and HFNC, we performed a subgroup analysis for 244 ventilated patients during both surges to compare outcomes between patients who received HFNC and those who did not receive HFNC (Table 4). Ninety-eight (40%) patients received HFNC. Ventilated patients who received HFNC had lower mortality than those patients who did not receive HFNC (31.6% vs 48%, P = .01), but had a longer length of hospital stay (29 days vs 14 days, P < .0001), longer length of ICU stay (17 days vs 9 days, P < .0001), and a higher number of ventilator days (16 vs 11, P = .001).

Our study compared the baseline patient characteristics; utilization of low-flow supplemental oxygen therapy, HFNC, and mechanical ventilation; and clinical outcomes between the first surge (n = 1586) and the second surge (n = 1597) of the COVID-19 pandemic. During both surges, about two-thirds of admitted patients received low-flow supplemental oxygen. A higher proportion of the admitted patients received HFNC during the second surge than during the first surge, while the intubation rate was lower during the second surge than during the first surge.

Reported low-flow supplemental oxygen use ranged from 28% to 63% depending on the cohort characteristics and location during the first surge.^6,7,19 A report from New York during the first surge (March 1 to April 4, 2020) showed that among 5700 hospitalized patients with COVID-19, 27.8% received low-flow supplemental oxygen.⁷ HFNC is recommended in guidelines on management of patients with acute respiratory failure due to COVID-19.²⁰ In our study, HFNC was utilized in a higher proportion of patients admitted for COVID-19 during the second surge (15.5% vs 10.8%, P = .0001). During the early pandemic period in Wuhan, China, 11% to 21% of admitted COVID-19 patients received HFNC.^21,22 Utilization of HFNC in New York during the first surge (March to May 2020) varied from 5% to 14.3% of patients admitted with COVID-19.^23,24 Our subgroup analysis of the ventilated patients showed that patients who received HFNC had lower mortality than those who did not (31.6% vs 48.0%, P = .011). Comparably, a report from Paris, France, showed that among patients admitted to ICUs for acute hypoxemic respiratory failure, those who received HFNC had lower mortality at day 60 than those who did not (21% vs 31%, P = .052).²⁵ Our recent analysis showed that patients treated with HFNC prior to mechanical ventilation had lower mortality than those treated with only conventional oxygen (30% vs 52%, P = .05).²⁶ In this subgroup analysis, we could not determine if HFNC treatment was administered before or after ventilation because HFNC was entered as dichotomous data (“Yes” or “No”) in the registry database. We merely showed the beneficial effect of HFNC on reducing mortality for ventilated COVID-19 patients, but did not mean to focus on how and when to apply HFNC.

We observed that the patients admitted during the second surge were less likely to be ventilated than the patients admitted during the first surge (5.6% vs 9.7%, P < .0001). During the first surge in New York, among 5700 patients admitted with COVID-19, 12.2% received invasive mechanical ventilation.⁷ In another report, also from New York during the first surge, 26.1% of 2015 hospitalized COVID-19 patients received mechanical ventilation.²⁷ In our study, the ventilation rate of 9.7% during the first surge was lower.

Outcomes during the second surge were better than during the first surge, including ICU admission rate, hospital and ICU length of stay, ventilator days, and mortality. The mortality was 19.2% during the first surge vs 8.3% during the second surge (P < .0001). The mortality of 19.2% was lower than the 30.6% mortality reported for 2015 hospitalized COVID-19 patients in New York during the first surge.²⁷ A retrospective study showed that early administration of remdesivir was associated with reduced ICU admission, ventilation use, and mortality.²⁸ The RECOVERY clinical trial showed that dexamethasone improved mortality for COVID-19 patients who received respiratory support, but not for patients who did not receive any respiratory support.²⁹ Perhaps some, if not all, of the improvement in ICU admission and mortality during the second surge was attributed to the new medications, such as antivirals and steroids.

The length of hospital stay for patients with moderate to severe COVID-19 varied from 4 to 53 days at different locations of the world, as shown in a meta-analysis by Rees and colleagues.³⁰ Our results showing a length of stay of 6 days during the first surge and 5 days during the second surge fell into the shorter end of this range. In a retrospective analysis of 1643 adults with severe COVID-19 admitted to hospitals in New York City between March 9, 2020 and April 23, 2020, median hospital length of stay was 7 (IQR, 3-14) days.³¹ For the ventilated patients in our study, the length of stay of 14 days (did not receive HFNC) and 29 days (received HFNC) was much longer. This longer length of stay might be attributed to the patients in our study being older and having more severe comorbidities.

The main purpose of this study was to compare the oxygen therapies and outcomes between 2 surges. It is difficult to associate the clinical outcomes with the oxygen therapies because new therapies and medications were available after the first surge. It was not possible to adjust the outcomes with confounders (other therapies and medications) because the registry data did not include the new therapies and medications.

A strength of this study was that we included a large, balanced number of patients in the first surge and the second surge. We did not plan the sample size in both groups as we could not predict the number of admissions. We set the end date of data collection for analysis as the time when the number of patients admitted during the second surge was similar to the number of patients admitted during the first surge. A limitation was that the registry database was created by the institution and was not designed solely for this study. The data for oxygen therapies were limited to low-flow supplemental oxygen, HFNC, and invasive mechanical ventilation; data for noninvasive ventilation were not included.

Conclusion

At our centers, during the second surge of COVID-19 pandemic, patients hospitalized with COVID-19 infection were more likely to receive HFNC but less likely to be ventilated. Compared to the first surge, the hospitalized patients with COVID-19 infection had a lower ICU admission rate, shorter length of hospital stay, fewer ventilator days, and lower mortality. For ventilated patients, those who received HFNC had lower mortality than those who did not.

Corresponding author: Timothy N. Liesching, MD, 41 Mall Road, Burlington, MA 01805; Timothy.N.Liesching@lahey.org

Disclosures: None reported.

doi:10.12788/jcom.0086

From Lahey Hospital and Medical Center, Burlington, MA (Drs. Liesching and Lei), and Tufts University School of Medicine, Boston, MA (Dr. Liesching)

ABSTRACT

Objective: To compare the utilization of oxygen therapies and clinical outcomes of patients admitted for COVID-19 during the second surge of the pandemic to that of patients admitted during the first surge.

Design: Observational study using a registry database.

Setting: Three hospitals (791 inpatient beds and 76 intensive care unit [ICU] beds) within the Beth Israel Lahey Health system in Massachusetts.

Participants: We included 3183 patients with COVID-19 admitted to hospitals.

Measurements: Baseline data included demographics and comorbidities. Treatments included low-flow supplemental oxygen (2-6 L/min), high-flow oxygen via nasal cannula, and invasive mechanical ventilation. Outcomes included ICU admission, length of stay, ventilator days, and mortality.

Results: A total of 3183 patients were included: 1586 during the first surge and 1597 during the second surge. Compared to the first surge, patients admitted during the second surge had a similar rate of receiving low-flow supplemental oxygen (65.8% vs 64.1%, P = .3), a higher rate of receiving high-flow nasal cannula (15.4% vs 10.8%, P = .0001), and a lower ventilation rate (5.6% vs 9.7%, P < .0001). The outcomes during the second surge were better than those during the first surge: lower ICU admission rate (8.1% vs 12.7%, P < .0001), shorter length of hospital stay (5 vs 6 days, P < .0001), fewer ventilator days (10 vs 16, P = .01), and lower mortality (8.3% vs 19.2%, P < .0001). Among ventilated patients, those who received high-flow nasal cannula had lower mortality.

Conclusion: Compared to the first surge of the COVID-19 pandemic, patients admitted during the second surge had similar likelihood of receiving low-flow supplemental oxygen, were more likely to receive high-flow nasal cannula, were less likely to be ventilated, and had better outcomes.

Keywords: supplemental oxygen, high-flow nasal cannula, ventilator.

The respiratory system receives the major impact of SARS-CoV-2 virus, and hypoxemia has been the predominant diagnosis for patients hospitalized with COVID-19.^1,2 During the initial stage of the pandemic, oxygen therapies and mechanical ventilation were the only choices for these patients.^3-6 Standard-of-care treatment for patients with COVID-19 during the initial surge included oxygen therapies and mechanical ventilation for hypoxemia and medications for comorbidities and COVID-19–associated sequelae, such as multi-organ dysfunction and failure. A report from New York during the first surge (May 2020) showed that among 5700 hospitalized patients with COVID-19, 27.8% received supplemental oxygen and 12.2% received invasive mechanical ventilation.⁷ High-flow nasal cannula (HFNC) oxygen delivery has been utilized widely throughout the pandemic due to its superiority over other noninvasive respiratory support techniques.^8-12 Mechanical ventilation is always necessary for critically ill patients with acute respiratory distress syndrome. However, ventilator scarcity has become a bottleneck in caring for severely ill patients with COVID-19 during the pandemic.¹³

The clinical outcomes of hospitalized COVID-19 patients include a high intubation rate, long length of hospital and intensive care unit (ICU) stay, and high mortality.^14,15 As the pandemic evolved, new medications, including remdesivir, hydroxychloroquine, lopinavir, or interferon β-1a, were used in addition to the standard of care, but these did not result in significantly different mortality from standard of care.¹⁶ Steroids are becoming foundational to the treatment of severe COVID-19 pneumonia, but evidence from high-quality randomized controlled clinical trials is lacking.¹⁷ 

During the first surge from March to May 2020, Massachusetts had the third highest number of COVID-19 cases among states in the United States.¹⁸ In early 2021, COVID-19 cases were climbing close to the peak of the second surge in Massachusetts. In this study, we compared utilization of low-flow supplemental oxygen, HFNC, and mechanical ventilation and clinical outcomes of patients admitted to 3 hospitals in Massachusetts during the second surge of the pandemic to that of patients admitted during the first surge.

Methods

Setting

Beth Israel Lahey Health is a system of academic and teaching hospitals with primary care and specialty care providers. We included 3 centers within the Beth Israel Lahey Health system in Massachusetts: Lahey Hospital and Medical Center, with 335 inpatient hospital beds and 52 critical care beds; Beverly Hospital, with 227 beds and 14 critical care beds; and Winchester Hospital, with 229 beds and 10 ICU beds.

Participants

We included patients admitted to the 3 hospitals with COVID-19 as a primary or secondary diagnosis during the first surge of the pandemic (March 1, 2020 to June 15, 2020) and the second surge (November 15, 2020 to January 27, 2021). The timeframe of the first surge was defined as the window between the start date and the end date of data collection. During the time window of the first surge, 1586 patients were included. The start time of the second surge was defined as the date when the data collection was restarted; the end date was set when the number of patients (1597) accumulated was close to the number of patients in the first surge (1586), so that the two groups had similar sample size.

Study Design

A data registry of COVID-19 patients was created by our institution, and the data were prospectively collected starting in March 2020. We retrospectively extracted data on the following from the registry database for this observational study: demographics and baseline comorbidities; the use of low-flow supplemental oxygen, HFNC, and invasive mechanical ventilator; and ICU admission, length of hospital stay, length of ICU stay, and hospital discharge disposition. Start and end times for each oxygen therapy were not entered in the registry. Data about other oxygen therapies, such as noninvasive positive-pressure ventilation, were not collected in the registry database, and therefore were not included in the analysis.

Statistical Analysis

Continuous variables (eg, age) were tested for data distribution normality using the Shapiro-Wilk test. Normally distributed data were tested using unpaired t-tests and displayed as mean (SD). The skewed data were tested using the Wilcoxon rank sum test and displayed as median (interquartile range [IQR]). The categorical variables were compared using chi-square test. Comparisons with P ≤ .05 were considered significantly different. Statistical analysis for this study was generated using Statistical Analysis Software (SAS), version 9.4 for Windows (SAS Institute Inc.).

Results

Baseline Characteristics

We included 3183 patients: 1586 admitted during the first surge and 1597 admitted during the second surge. Baseline characteristics of patients with COVID-19 admitted during the first and second surges are shown in Table 1. Patients admitted during the second surge were older (73 years vs 71 years, P = .01) and had higher rates of hypertension (64.8% vs 59.6%, P = .003) and asthma (12.9% vs 10.7%, P = .049) but a lower rate of interstitial lung disease (3.3% vs 7.7%, P < .001). Sequential organ failure assessment scores at admission and the rates of other comorbidities were not significantly different between the 2 surges.

The number of patients who were hospitalized and received low-flow supplemental oxygen, and/or HFNC, and/or ventilator in the first surge and the second surge is shown in the Figure. Of all patients included, 2067 (64.9%) received low-flow supplemental oxygen; of these, 374 (18.1%) subsequently received HFNC, and 85 (22.7%) of these subsequently received mechanical ventilation. Of all 3183 patients, 417 (13.1%) received HFNC; 43 of these patients received HFNC without receiving low-flow supplemental oxygen, and 98 (23.5%) subsequently received mechanical ventilation. Out of all 3183 patients, 244 (7.7%) received mechanical ventilation; 98 (40.2%) of these received HFNC while the remaining 146 (59.8%) did not. At the beginning of the first surge, the ratio of patients who received invasive mechanical ventilation to patients who received HFNC was close to 1:1 (10/10); the ratio decreased to 6:10 in May and June 2020. At the beginning of the second surge, the ratio was 8:10 and then decreased to 3:10 in December 2020 and January 2021.

As shown in Table 2, the proportion of patients who received low-flow supplemental oxygen during the second surge was similar to that during the first surge (65.8% vs 64.1%, P = .3). Patients admitted during the second surge were more likely to receive HFNC than patients admitted during the first surge (15.4% vs 10.8%, P = .0001). Patients admitted during the second surge were less likely to be ventilated than the patients admitted during the first surge (5.6% vs 9.7%, P < .0001).

Clinical Outcomes

As shown in Table 3, second surge outcomes were much better than first surge outcomes: the ICU admission rate was lower (8.1% vs 12.7%, P < .0001); patients were more likely to be discharged to home (60.2% vs 47.4%, P < .0001), had a shorter length of hospital stay (5 vs 6 days, P < .0001), and had fewer ventilator days (10 vs 16, P = .01); and mortality was lower (8.3% vs 19.2%, P < .0001). There was a trend that length of ICU stay was shorter during the second surge than during the first surge (7 days vs 9 days, P = .09).

As noted (Figure), the ratio of patients who received invasive mechanical ventilation to patients who received HFNC was decreasing during both the first surge and the second surge. To further analyze the relation between ventilator and HFNC, we performed a subgroup analysis for 244 ventilated patients during both surges to compare outcomes between patients who received HFNC and those who did not receive HFNC (Table 4). Ninety-eight (40%) patients received HFNC. Ventilated patients who received HFNC had lower mortality than those patients who did not receive HFNC (31.6% vs 48%, P = .01), but had a longer length of hospital stay (29 days vs 14 days, P < .0001), longer length of ICU stay (17 days vs 9 days, P < .0001), and a higher number of ventilator days (16 vs 11, P = .001).

Our study compared the baseline patient characteristics; utilization of low-flow supplemental oxygen therapy, HFNC, and mechanical ventilation; and clinical outcomes between the first surge (n = 1586) and the second surge (n = 1597) of the COVID-19 pandemic. During both surges, about two-thirds of admitted patients received low-flow supplemental oxygen. A higher proportion of the admitted patients received HFNC during the second surge than during the first surge, while the intubation rate was lower during the second surge than during the first surge.

Reported low-flow supplemental oxygen use ranged from 28% to 63% depending on the cohort characteristics and location during the first surge.^6,7,19 A report from New York during the first surge (March 1 to April 4, 2020) showed that among 5700 hospitalized patients with COVID-19, 27.8% received low-flow supplemental oxygen.⁷ HFNC is recommended in guidelines on management of patients with acute respiratory failure due to COVID-19.²⁰ In our study, HFNC was utilized in a higher proportion of patients admitted for COVID-19 during the second surge (15.5% vs 10.8%, P = .0001). During the early pandemic period in Wuhan, China, 11% to 21% of admitted COVID-19 patients received HFNC.^21,22 Utilization of HFNC in New York during the first surge (March to May 2020) varied from 5% to 14.3% of patients admitted with COVID-19.^23,24 Our subgroup analysis of the ventilated patients showed that patients who received HFNC had lower mortality than those who did not (31.6% vs 48.0%, P = .011). Comparably, a report from Paris, France, showed that among patients admitted to ICUs for acute hypoxemic respiratory failure, those who received HFNC had lower mortality at day 60 than those who did not (21% vs 31%, P = .052).²⁵ Our recent analysis showed that patients treated with HFNC prior to mechanical ventilation had lower mortality than those treated with only conventional oxygen (30% vs 52%, P = .05).²⁶ In this subgroup analysis, we could not determine if HFNC treatment was administered before or after ventilation because HFNC was entered as dichotomous data (“Yes” or “No”) in the registry database. We merely showed the beneficial effect of HFNC on reducing mortality for ventilated COVID-19 patients, but did not mean to focus on how and when to apply HFNC.

We observed that the patients admitted during the second surge were less likely to be ventilated than the patients admitted during the first surge (5.6% vs 9.7%, P < .0001). During the first surge in New York, among 5700 patients admitted with COVID-19, 12.2% received invasive mechanical ventilation.⁷ In another report, also from New York during the first surge, 26.1% of 2015 hospitalized COVID-19 patients received mechanical ventilation.²⁷ In our study, the ventilation rate of 9.7% during the first surge was lower.

Outcomes during the second surge were better than during the first surge, including ICU admission rate, hospital and ICU length of stay, ventilator days, and mortality. The mortality was 19.2% during the first surge vs 8.3% during the second surge (P < .0001). The mortality of 19.2% was lower than the 30.6% mortality reported for 2015 hospitalized COVID-19 patients in New York during the first surge.²⁷ A retrospective study showed that early administration of remdesivir was associated with reduced ICU admission, ventilation use, and mortality.²⁸ The RECOVERY clinical trial showed that dexamethasone improved mortality for COVID-19 patients who received respiratory support, but not for patients who did not receive any respiratory support.²⁹ Perhaps some, if not all, of the improvement in ICU admission and mortality during the second surge was attributed to the new medications, such as antivirals and steroids.

The length of hospital stay for patients with moderate to severe COVID-19 varied from 4 to 53 days at different locations of the world, as shown in a meta-analysis by Rees and colleagues.³⁰ Our results showing a length of stay of 6 days during the first surge and 5 days during the second surge fell into the shorter end of this range. In a retrospective analysis of 1643 adults with severe COVID-19 admitted to hospitals in New York City between March 9, 2020 and April 23, 2020, median hospital length of stay was 7 (IQR, 3-14) days.³¹ For the ventilated patients in our study, the length of stay of 14 days (did not receive HFNC) and 29 days (received HFNC) was much longer. This longer length of stay might be attributed to the patients in our study being older and having more severe comorbidities.

The main purpose of this study was to compare the oxygen therapies and outcomes between 2 surges. It is difficult to associate the clinical outcomes with the oxygen therapies because new therapies and medications were available after the first surge. It was not possible to adjust the outcomes with confounders (other therapies and medications) because the registry data did not include the new therapies and medications.

A strength of this study was that we included a large, balanced number of patients in the first surge and the second surge. We did not plan the sample size in both groups as we could not predict the number of admissions. We set the end date of data collection for analysis as the time when the number of patients admitted during the second surge was similar to the number of patients admitted during the first surge. A limitation was that the registry database was created by the institution and was not designed solely for this study. The data for oxygen therapies were limited to low-flow supplemental oxygen, HFNC, and invasive mechanical ventilation; data for noninvasive ventilation were not included.

Conclusion

At our centers, during the second surge of COVID-19 pandemic, patients hospitalized with COVID-19 infection were more likely to receive HFNC but less likely to be ventilated. Compared to the first surge, the hospitalized patients with COVID-19 infection had a lower ICU admission rate, shorter length of hospital stay, fewer ventilator days, and lower mortality. For ventilated patients, those who received HFNC had lower mortality than those who did not.

Corresponding author: Timothy N. Liesching, MD, 41 Mall Road, Burlington, MA 01805; Timothy.N.Liesching@lahey.org

Disclosures: None reported.

doi:10.12788/jcom.0086

Study Overview

Objective: This study evaluated whether a clinical-decision-support (CDS) tool that utilizes a real-time algorithm incorporating patient vital sign data can identify hospitalized patients who can forgo overnight vital sign checks and thus reduce delirium incidence.

Design: This was a parallel randomized clinical trial of adult inpatients admitted to the general medical service of a tertiary care academic medical center in the United States. The trial intervention consisted of a CDS notification in the electronic health record (EHR) that informed the physician if a patient had a high likelihood of nighttime vital signs within the reference ranges based on a logistic regression model of real-time patient data input. This notification provided the physician an opportunity to discontinue nighttime vital sign checks, dismiss the notification for 1 hour, or dismiss the notification until the next day.

Setting and participants: This clinical trial was conducted at the University of California, San Francisco Medical Center from March 11 to November 24, 2019. Participants included physicians who served on the primary team (eg, attending, resident) of 1699 patients on the general medical service who were outside of the intensive care unit (ICU). The hospital encounters were randomized (allocation ratio of 1:1) to sleep promotion vitals CDS (SPV CDS) intervention or usual care.

Main outcome and measures: The primary outcome was delirium as determined by bedside nurse assessment using the Nursing Delirium Screening Scale (Nu-DESC) recorded once per nursing shift. The Nu-DESC is a standardized delirium screening tool that defines delirium with a score ≥2. Secondary outcomes included sleep opportunity (ie, EHR-based sleep metrics that reflected the maximum time between iatrogenic interruptions, such as nighttime vital sign checks) and patient satisfaction (ie, patient satisfaction measured by standardized Hospital Consumer Assessment of Healthcare Providers and Systems [HCAHPS] survey). Potential balancing outcomes were assessed to ensure that reduced vital sign checks were not causing harms; these included ICU transfers, rapid response calls, and code blue alarms. All analyses were conducted on the basis of intention-to-treat.

Main results: A total of 3025 inpatient encounters were screened and 1930 encounters were randomized (966 SPV CDS intervention; 964 usual care). The randomized encounters consisted of 1699 patients; demographic factors between the 2 trial arms were similar. Specifically, the intervention arm included 566 men (59%) and mean (SD) age was 53 (15) years. The incidence of delirium was similar between the intervention and usual care arms: 108 (11%) vs 123 (13%) (P = .32). Compared to the usual care arm, the intervention arm had a higher mean (SD) number of sleep opportunity hours per night (4.95 [1.45] vs 4.57 [1.30], P < .001) and fewer nighttime vital sign checks (0.97 [0.95] vs 1.41 [0.86], P < .001). The post-discharge HCAHPS survey measuring patient satisfaction was completed by only 5% of patients (53 intervention, 49 usual care), and survey results were similar between the 2 arms (P = .86). In addition, safety outcomes including ICU transfers (49 [5%] vs 47 [5%], P = .92), rapid response calls (68 [7%] vs 55 [6%], P = .27), and code blue alarms (2 [0.2%] vs 9 [0.9%], P = .07) were similar between the study arms.

Conclusion: In this randomized clinical trial, a CDS tool utilizing a real-time prediction algorithm embedded in EHR did not reduce the incidence of delirium in hospitalized patients. However, this SPV CDS intervention helped physicians identify clinically stable patients who can forgo routine nighttime vital sign checks and facilitated greater opportunity for patients to sleep. These findings suggest that augmenting physician judgment using a real-time prediction algorithm can help to improve sleep opportunity without an accompanying increased risk of clinical decompensation during acute care.

Commentary

High-quality sleep is fundamental to health and well-being. Sleep deprivation and disorders are associated with many adverse health outcomes, including increased risks for obesity, diabetes, hypertension, myocardial infarction, and depression.¹ In hospitalized patients who are acutely ill, restorative sleep is critical to facilitating recovery. However, poor sleep is exceedingly common in hospitalized patients and is associated with deleterious outcomes, such as high blood pressure, hyperglycemia, and delirium.^2,3 Moreover, some of these adverse sleep-induced cardiometabolic outcomes, as well as sleep disruption itself, may persist after hospital discharge.⁴ Factors that precipitate interrupted sleep during hospitalization include iatrogenic causes such as frequent vital sign checks, nighttime procedures or early morning blood draws, and environmental factors such as loud ambient noise.³ Thus, a potential intervention to improve sleep quality in the hospital is to reduce nighttime interruptions such as frequent vital sign checks.

In the current study, Najafi and colleagues conducted a randomized trial to evaluate whether a CDS tool embedded in EHR, powered by a real-time prediction algorithm of patient data, can be utilized to identify patients in whom vital sign checks can be safely discontinued at nighttime. The authors found a modest but statistically significant reduction in the number of nighttime vital sign checks in patients who underwent the SPV CDS intervention, and a corresponding higher sleep opportunity per night in those who received the intervention. Importantly, this reduction in nighttime vital sign checks did not cause a higher risk of clinical decompensation as measured by ICU transfers, rapid response calls, or code blue alarms. Thus, the results demonstrated the feasibility of using a real-time, patient data-driven CDS tool to augment physician judgment in managing sleep disruption, an important hospital-associated stressor and a common hazard of hospitalization in older patients.

Delirium is a common clinical problem in hospitalized older patients that is associated with prolonged hospitalization, functional and cognitive decline, institutionalization, death, and increased health care costs.⁵ Despite a potential benefit of SPV CDS intervention in reducing vital sign checks and increasing sleep opportunity, this intervention did not reduce the incidence of delirium in hospitalized patients. This finding is not surprising given that delirium has a multifactorial etiology (eg, metabolic derangements, infections, medication side effects and drug toxicity, hospital environment). A small modification in nighttime vital sign checks and sleep opportunity may have limited impact on optimizing sleep quality and does not address other risk factors for delirium. As such, a multicomponent nonpharmacologic approach that includes sleep enhancement, early mobilization, feeding assistance, fluid repletion, infection prevention, and other interventions should guide delirium prevention in the hospital setting. The SPV CDS intervention may play a role in the delivery of a multifaceted, nonpharmacologic delirium prevention intervention in high-risk individuals.

Sleep disruption is one of the multiple hazards of hospitalization frequently experience by hospitalized older patients. Other hazards, or hospital-associated stressors, include mobility restriction (eg, physical restraints such as urinary catheters and intravenous lines, bed elevation and rails), malnourishment and dehydration (eg, frequent use of no-food-by-mouth order, lack of easy access to hydration), and pain (eg, poor pain control). Extended exposures to these stressors may lead to a maladaptive state called allostatic overload that transiently increases vulnerability to post-hospitalization adverse events, including emergency department use, hospital readmission, or death (ie, post-hospital syndrome).⁶ Thus, the optimization of sleep during hospitalization in vulnerable patients may have benefits that extend beyond delirium prevention. It is perceivable that a CDS tool embedded in EHR, powered by a real-time prediction algorithm of patient data, may be applied to reduce some of these hazards of hospitalization in addition to improving sleep opportunity.

Applications for Clinical Practice

Findings from the current study indicate that a CDS tool embedded in EHR that utilizes a real-time prediction algorithm of patient data may help to safely improve sleep opportunity in hospitalized patients. The participants in the current study were relatively young (53 [15] years). Given that age is a risk factor for delirium, the effects of this intervention on delirium prevention in the most susceptible population (ie, those over the age of 65) remain unknown and further investigation is warranted. Additional studies are needed to determine whether this approach yields similar results in geriatric patients and improves clinical outcomes.

—Fred Ko, MD

Study Overview

Objective: This study evaluated whether a clinical-decision-support (CDS) tool that utilizes a real-time algorithm incorporating patient vital sign data can identify hospitalized patients who can forgo overnight vital sign checks and thus reduce delirium incidence.

Design: This was a parallel randomized clinical trial of adult inpatients admitted to the general medical service of a tertiary care academic medical center in the United States. The trial intervention consisted of a CDS notification in the electronic health record (EHR) that informed the physician if a patient had a high likelihood of nighttime vital signs within the reference ranges based on a logistic regression model of real-time patient data input. This notification provided the physician an opportunity to discontinue nighttime vital sign checks, dismiss the notification for 1 hour, or dismiss the notification until the next day.

Setting and participants: This clinical trial was conducted at the University of California, San Francisco Medical Center from March 11 to November 24, 2019. Participants included physicians who served on the primary team (eg, attending, resident) of 1699 patients on the general medical service who were outside of the intensive care unit (ICU). The hospital encounters were randomized (allocation ratio of 1:1) to sleep promotion vitals CDS (SPV CDS) intervention or usual care.

Main outcome and measures: The primary outcome was delirium as determined by bedside nurse assessment using the Nursing Delirium Screening Scale (Nu-DESC) recorded once per nursing shift. The Nu-DESC is a standardized delirium screening tool that defines delirium with a score ≥2. Secondary outcomes included sleep opportunity (ie, EHR-based sleep metrics that reflected the maximum time between iatrogenic interruptions, such as nighttime vital sign checks) and patient satisfaction (ie, patient satisfaction measured by standardized Hospital Consumer Assessment of Healthcare Providers and Systems [HCAHPS] survey). Potential balancing outcomes were assessed to ensure that reduced vital sign checks were not causing harms; these included ICU transfers, rapid response calls, and code blue alarms. All analyses were conducted on the basis of intention-to-treat.

Main results: A total of 3025 inpatient encounters were screened and 1930 encounters were randomized (966 SPV CDS intervention; 964 usual care). The randomized encounters consisted of 1699 patients; demographic factors between the 2 trial arms were similar. Specifically, the intervention arm included 566 men (59%) and mean (SD) age was 53 (15) years. The incidence of delirium was similar between the intervention and usual care arms: 108 (11%) vs 123 (13%) (P = .32). Compared to the usual care arm, the intervention arm had a higher mean (SD) number of sleep opportunity hours per night (4.95 [1.45] vs 4.57 [1.30], P < .001) and fewer nighttime vital sign checks (0.97 [0.95] vs 1.41 [0.86], P < .001). The post-discharge HCAHPS survey measuring patient satisfaction was completed by only 5% of patients (53 intervention, 49 usual care), and survey results were similar between the 2 arms (P = .86). In addition, safety outcomes including ICU transfers (49 [5%] vs 47 [5%], P = .92), rapid response calls (68 [7%] vs 55 [6%], P = .27), and code blue alarms (2 [0.2%] vs 9 [0.9%], P = .07) were similar between the study arms.

Conclusion: In this randomized clinical trial, a CDS tool utilizing a real-time prediction algorithm embedded in EHR did not reduce the incidence of delirium in hospitalized patients. However, this SPV CDS intervention helped physicians identify clinically stable patients who can forgo routine nighttime vital sign checks and facilitated greater opportunity for patients to sleep. These findings suggest that augmenting physician judgment using a real-time prediction algorithm can help to improve sleep opportunity without an accompanying increased risk of clinical decompensation during acute care.

Commentary

High-quality sleep is fundamental to health and well-being. Sleep deprivation and disorders are associated with many adverse health outcomes, including increased risks for obesity, diabetes, hypertension, myocardial infarction, and depression.¹ In hospitalized patients who are acutely ill, restorative sleep is critical to facilitating recovery. However, poor sleep is exceedingly common in hospitalized patients and is associated with deleterious outcomes, such as high blood pressure, hyperglycemia, and delirium.^2,3 Moreover, some of these adverse sleep-induced cardiometabolic outcomes, as well as sleep disruption itself, may persist after hospital discharge.⁴ Factors that precipitate interrupted sleep during hospitalization include iatrogenic causes such as frequent vital sign checks, nighttime procedures or early morning blood draws, and environmental factors such as loud ambient noise.³ Thus, a potential intervention to improve sleep quality in the hospital is to reduce nighttime interruptions such as frequent vital sign checks.

In the current study, Najafi and colleagues conducted a randomized trial to evaluate whether a CDS tool embedded in EHR, powered by a real-time prediction algorithm of patient data, can be utilized to identify patients in whom vital sign checks can be safely discontinued at nighttime. The authors found a modest but statistically significant reduction in the number of nighttime vital sign checks in patients who underwent the SPV CDS intervention, and a corresponding higher sleep opportunity per night in those who received the intervention. Importantly, this reduction in nighttime vital sign checks did not cause a higher risk of clinical decompensation as measured by ICU transfers, rapid response calls, or code blue alarms. Thus, the results demonstrated the feasibility of using a real-time, patient data-driven CDS tool to augment physician judgment in managing sleep disruption, an important hospital-associated stressor and a common hazard of hospitalization in older patients.

Delirium is a common clinical problem in hospitalized older patients that is associated with prolonged hospitalization, functional and cognitive decline, institutionalization, death, and increased health care costs.⁵ Despite a potential benefit of SPV CDS intervention in reducing vital sign checks and increasing sleep opportunity, this intervention did not reduce the incidence of delirium in hospitalized patients. This finding is not surprising given that delirium has a multifactorial etiology (eg, metabolic derangements, infections, medication side effects and drug toxicity, hospital environment). A small modification in nighttime vital sign checks and sleep opportunity may have limited impact on optimizing sleep quality and does not address other risk factors for delirium. As such, a multicomponent nonpharmacologic approach that includes sleep enhancement, early mobilization, feeding assistance, fluid repletion, infection prevention, and other interventions should guide delirium prevention in the hospital setting. The SPV CDS intervention may play a role in the delivery of a multifaceted, nonpharmacologic delirium prevention intervention in high-risk individuals.

Sleep disruption is one of the multiple hazards of hospitalization frequently experience by hospitalized older patients. Other hazards, or hospital-associated stressors, include mobility restriction (eg, physical restraints such as urinary catheters and intravenous lines, bed elevation and rails), malnourishment and dehydration (eg, frequent use of no-food-by-mouth order, lack of easy access to hydration), and pain (eg, poor pain control). Extended exposures to these stressors may lead to a maladaptive state called allostatic overload that transiently increases vulnerability to post-hospitalization adverse events, including emergency department use, hospital readmission, or death (ie, post-hospital syndrome).⁶ Thus, the optimization of sleep during hospitalization in vulnerable patients may have benefits that extend beyond delirium prevention. It is perceivable that a CDS tool embedded in EHR, powered by a real-time prediction algorithm of patient data, may be applied to reduce some of these hazards of hospitalization in addition to improving sleep opportunity.

Applications for Clinical Practice

Findings from the current study indicate that a CDS tool embedded in EHR that utilizes a real-time prediction algorithm of patient data may help to safely improve sleep opportunity in hospitalized patients. The participants in the current study were relatively young (53 [15] years). Given that age is a risk factor for delirium, the effects of this intervention on delirium prevention in the most susceptible population (ie, those over the age of 65) remain unknown and further investigation is warranted. Additional studies are needed to determine whether this approach yields similar results in geriatric patients and improves clinical outcomes.

—Fred Ko, MD

Study Overview

Objective: This study evaluated whether a clinical-decision-support (CDS) tool that utilizes a real-time algorithm incorporating patient vital sign data can identify hospitalized patients who can forgo overnight vital sign checks and thus reduce delirium incidence.

Design: This was a parallel randomized clinical trial of adult inpatients admitted to the general medical service of a tertiary care academic medical center in the United States. The trial intervention consisted of a CDS notification in the electronic health record (EHR) that informed the physician if a patient had a high likelihood of nighttime vital signs within the reference ranges based on a logistic regression model of real-time patient data input. This notification provided the physician an opportunity to discontinue nighttime vital sign checks, dismiss the notification for 1 hour, or dismiss the notification until the next day.

Setting and participants: This clinical trial was conducted at the University of California, San Francisco Medical Center from March 11 to November 24, 2019. Participants included physicians who served on the primary team (eg, attending, resident) of 1699 patients on the general medical service who were outside of the intensive care unit (ICU). The hospital encounters were randomized (allocation ratio of 1:1) to sleep promotion vitals CDS (SPV CDS) intervention or usual care.

Main outcome and measures: The primary outcome was delirium as determined by bedside nurse assessment using the Nursing Delirium Screening Scale (Nu-DESC) recorded once per nursing shift. The Nu-DESC is a standardized delirium screening tool that defines delirium with a score ≥2. Secondary outcomes included sleep opportunity (ie, EHR-based sleep metrics that reflected the maximum time between iatrogenic interruptions, such as nighttime vital sign checks) and patient satisfaction (ie, patient satisfaction measured by standardized Hospital Consumer Assessment of Healthcare Providers and Systems [HCAHPS] survey). Potential balancing outcomes were assessed to ensure that reduced vital sign checks were not causing harms; these included ICU transfers, rapid response calls, and code blue alarms. All analyses were conducted on the basis of intention-to-treat.

Main results: A total of 3025 inpatient encounters were screened and 1930 encounters were randomized (966 SPV CDS intervention; 964 usual care). The randomized encounters consisted of 1699 patients; demographic factors between the 2 trial arms were similar. Specifically, the intervention arm included 566 men (59%) and mean (SD) age was 53 (15) years. The incidence of delirium was similar between the intervention and usual care arms: 108 (11%) vs 123 (13%) (P = .32). Compared to the usual care arm, the intervention arm had a higher mean (SD) number of sleep opportunity hours per night (4.95 [1.45] vs 4.57 [1.30], P < .001) and fewer nighttime vital sign checks (0.97 [0.95] vs 1.41 [0.86], P < .001). The post-discharge HCAHPS survey measuring patient satisfaction was completed by only 5% of patients (53 intervention, 49 usual care), and survey results were similar between the 2 arms (P = .86). In addition, safety outcomes including ICU transfers (49 [5%] vs 47 [5%], P = .92), rapid response calls (68 [7%] vs 55 [6%], P = .27), and code blue alarms (2 [0.2%] vs 9 [0.9%], P = .07) were similar between the study arms.

Conclusion: In this randomized clinical trial, a CDS tool utilizing a real-time prediction algorithm embedded in EHR did not reduce the incidence of delirium in hospitalized patients. However, this SPV CDS intervention helped physicians identify clinically stable patients who can forgo routine nighttime vital sign checks and facilitated greater opportunity for patients to sleep. These findings suggest that augmenting physician judgment using a real-time prediction algorithm can help to improve sleep opportunity without an accompanying increased risk of clinical decompensation during acute care.

Commentary

High-quality sleep is fundamental to health and well-being. Sleep deprivation and disorders are associated with many adverse health outcomes, including increased risks for obesity, diabetes, hypertension, myocardial infarction, and depression.¹ In hospitalized patients who are acutely ill, restorative sleep is critical to facilitating recovery. However, poor sleep is exceedingly common in hospitalized patients and is associated with deleterious outcomes, such as high blood pressure, hyperglycemia, and delirium.^2,3 Moreover, some of these adverse sleep-induced cardiometabolic outcomes, as well as sleep disruption itself, may persist after hospital discharge.⁴ Factors that precipitate interrupted sleep during hospitalization include iatrogenic causes such as frequent vital sign checks, nighttime procedures or early morning blood draws, and environmental factors such as loud ambient noise.³ Thus, a potential intervention to improve sleep quality in the hospital is to reduce nighttime interruptions such as frequent vital sign checks.

In the current study, Najafi and colleagues conducted a randomized trial to evaluate whether a CDS tool embedded in EHR, powered by a real-time prediction algorithm of patient data, can be utilized to identify patients in whom vital sign checks can be safely discontinued at nighttime. The authors found a modest but statistically significant reduction in the number of nighttime vital sign checks in patients who underwent the SPV CDS intervention, and a corresponding higher sleep opportunity per night in those who received the intervention. Importantly, this reduction in nighttime vital sign checks did not cause a higher risk of clinical decompensation as measured by ICU transfers, rapid response calls, or code blue alarms. Thus, the results demonstrated the feasibility of using a real-time, patient data-driven CDS tool to augment physician judgment in managing sleep disruption, an important hospital-associated stressor and a common hazard of hospitalization in older patients.

Delirium is a common clinical problem in hospitalized older patients that is associated with prolonged hospitalization, functional and cognitive decline, institutionalization, death, and increased health care costs.⁵ Despite a potential benefit of SPV CDS intervention in reducing vital sign checks and increasing sleep opportunity, this intervention did not reduce the incidence of delirium in hospitalized patients. This finding is not surprising given that delirium has a multifactorial etiology (eg, metabolic derangements, infections, medication side effects and drug toxicity, hospital environment). A small modification in nighttime vital sign checks and sleep opportunity may have limited impact on optimizing sleep quality and does not address other risk factors for delirium. As such, a multicomponent nonpharmacologic approach that includes sleep enhancement, early mobilization, feeding assistance, fluid repletion, infection prevention, and other interventions should guide delirium prevention in the hospital setting. The SPV CDS intervention may play a role in the delivery of a multifaceted, nonpharmacologic delirium prevention intervention in high-risk individuals.

Sleep disruption is one of the multiple hazards of hospitalization frequently experience by hospitalized older patients. Other hazards, or hospital-associated stressors, include mobility restriction (eg, physical restraints such as urinary catheters and intravenous lines, bed elevation and rails), malnourishment and dehydration (eg, frequent use of no-food-by-mouth order, lack of easy access to hydration), and pain (eg, poor pain control). Extended exposures to these stressors may lead to a maladaptive state called allostatic overload that transiently increases vulnerability to post-hospitalization adverse events, including emergency department use, hospital readmission, or death (ie, post-hospital syndrome).⁶ Thus, the optimization of sleep during hospitalization in vulnerable patients may have benefits that extend beyond delirium prevention. It is perceivable that a CDS tool embedded in EHR, powered by a real-time prediction algorithm of patient data, may be applied to reduce some of these hazards of hospitalization in addition to improving sleep opportunity.

Applications for Clinical Practice

Findings from the current study indicate that a CDS tool embedded in EHR that utilizes a real-time prediction algorithm of patient data may help to safely improve sleep opportunity in hospitalized patients. The participants in the current study were relatively young (53 [15] years). Given that age is a risk factor for delirium, the effects of this intervention on delirium prevention in the most susceptible population (ie, those over the age of 65) remain unknown and further investigation is warranted. Additional studies are needed to determine whether this approach yields similar results in geriatric patients and improves clinical outcomes.

—Fred Ko, MD

Study Overview

Objective: To assess the safety and efficacy of early hospital discharge (EHD) for selected low-risk patients with ST-segment elevation myocardial infarction (STEMI) after primary percutaneous coronary intervention (PCI).

Design: Single-center retrospective analysis of prospectively collected data.

Setting and participants: An EHD group comprised of 600 patients who were discharged at <48 hours between April 2020 and June 2021 was compared to a control group of 700 patients who met EHD criteria but were discharged at >48 hour between October 2018 and June 2021. Patients were selected into the EHD group based on the following criteria, in accordance with recommendations from the European Society of Cardiology, and all patients had close follow-up with a combination of structured telephone follow-up at 48 hours post discharge and virtual visits at 2, 6, and 8 weeks and at 3 months:

• Left ventricular ejection fraction ≥40%
• Successful primary PCI (that achieved thrombolysis in myocardial infarction flow grade 3)
• Absence of severe nonculprit disease requiring further inpatient revascularization
• Absence of ischemic symptoms post PCI
• Absence of heart failure or hemodynamic instability
• Absence of significant arrhythmia (ventricular fibrillation, ventricular tachycardia, or atrial fibrillation or atrial flutter requiring prolonged stay)
• Mobility with suitable social circumstances for discharge

Main outcome measures: The outcomes measured were length of hospitalization and a composite primary endpoint of cardiovascular mortality and major adverse cardiovascular event (MACE) rates, defined as a composite of all-cause mortality, recurrent MI, and target lesion revascularization.

Main results: The median length of stay of hospitalization in the EHD group was 24.6 hours compared to 56.1 hours in the >48-hour historical control group. On median follow-up of 271 days, the EHD group demonstrated 0% cardiovascular mortality and a MACE rate of 1.2%. This was shown to be noninferior compared to the >48-hour historical control group, which had mortality of 0.7% and a MACE rate of 1.9%.

Conclusion: Selected low-risk STEMI patients can be safely discharged early with appropriate follow-up after primary PCI.

Commentary

Patients with STEMI have a higher risk of postprocedural adverse events such as MI, arrhythmia, or acute heart failure compared to patients with stable ischemic heart disease, and thus are monitored after primary PCI. Although patients were traditionally monitored for 5 to 7 days a few decades ago,¹ with improvements in PCI techniques, devices, and pharmacotherapy as well as in door-to-balloon time, the in-hospital complication rates for patients with STEMI have been decreasing, leading to earlier discharge. Currently in the United States, patients are most commonly monitored for 48 to 72 hours post PCI.² The current guidelines support this practice, recommending early discharge within 48 to 72 hours in selected low-risk patients if adequate follow-up and rehabilitation are arranged.³

Given the COVID-19 pandemic and decreased hospital bed availability, Rathod et al took one step further on the question of whether low-risk STEMI patients with primary PCI can be discharged safely within 48 hours with adequate follow-up. They found that at a median follow-up of 271 days, EHD patients had 2 COVID-related deaths, with 0% cardiovascular mortality and a MACE rate of 1.2%, including deaths, MI, and ischemic revascularization. The median time to discharge was 25 hours. This was noninferior to the >48-hour historical control group, which had mortality of 0.7% (P = 0.349) and a MACE rate of 1.9% (P = .674). The results remained similar after propensity matching for mortality (0.34% vs 0.69%, P = .410) or MACE (1.2% vs 1.9%, P = .342).

This is the first prospective study to systematically assess the safety and feasibility of discharge of low-risk STEMI patients with primary PCI within 48 hours. This study is unique in that it involved the use of telemedicine, including a virtual platform to collect data such as heart rate, blood pressure, and blood glucose, and virtual visits to facilitate follow-up and reduce clinic travel, cost, and potential COVID-19 exposure. The investigators’ protocol included virtual follow-up by cardiology advanced practitioners at 2, 6, and 8 weeks and by an interventional cardiologist at 12 weeks. This protocol led to an increase in patient satisfaction. The study’s main limitation is that it is a single-center trial with a smaller sample size. Further studies are necessary to confirm the safety and feasibility of this approach. In addition, further refinement of the patient selection criteria for EHD should be considered.

Applications for Clinical Practice

In low-risk STEMI patients after primary PCI, discharge within 48 hours may be considered if close follow-up is arranged. However, further studies are necessary to confirm this finding.

—Thai Nguyen, MD, Albert Chan, MD, and Taishi Hirai MD

Study Overview

Objective: To assess the safety and efficacy of early hospital discharge (EHD) for selected low-risk patients with ST-segment elevation myocardial infarction (STEMI) after primary percutaneous coronary intervention (PCI).

Design: Single-center retrospective analysis of prospectively collected data.

Setting and participants: An EHD group comprised of 600 patients who were discharged at <48 hours between April 2020 and June 2021 was compared to a control group of 700 patients who met EHD criteria but were discharged at >48 hour between October 2018 and June 2021. Patients were selected into the EHD group based on the following criteria, in accordance with recommendations from the European Society of Cardiology, and all patients had close follow-up with a combination of structured telephone follow-up at 48 hours post discharge and virtual visits at 2, 6, and 8 weeks and at 3 months:

• Left ventricular ejection fraction ≥40%
• Successful primary PCI (that achieved thrombolysis in myocardial infarction flow grade 3)
• Absence of severe nonculprit disease requiring further inpatient revascularization
• Absence of ischemic symptoms post PCI
• Absence of heart failure or hemodynamic instability
• Absence of significant arrhythmia (ventricular fibrillation, ventricular tachycardia, or atrial fibrillation or atrial flutter requiring prolonged stay)
• Mobility with suitable social circumstances for discharge

Main outcome measures: The outcomes measured were length of hospitalization and a composite primary endpoint of cardiovascular mortality and major adverse cardiovascular event (MACE) rates, defined as a composite of all-cause mortality, recurrent MI, and target lesion revascularization.

Main results: The median length of stay of hospitalization in the EHD group was 24.6 hours compared to 56.1 hours in the >48-hour historical control group. On median follow-up of 271 days, the EHD group demonstrated 0% cardiovascular mortality and a MACE rate of 1.2%. This was shown to be noninferior compared to the >48-hour historical control group, which had mortality of 0.7% and a MACE rate of 1.9%.

Conclusion: Selected low-risk STEMI patients can be safely discharged early with appropriate follow-up after primary PCI.

Commentary

Patients with STEMI have a higher risk of postprocedural adverse events such as MI, arrhythmia, or acute heart failure compared to patients with stable ischemic heart disease, and thus are monitored after primary PCI. Although patients were traditionally monitored for 5 to 7 days a few decades ago,¹ with improvements in PCI techniques, devices, and pharmacotherapy as well as in door-to-balloon time, the in-hospital complication rates for patients with STEMI have been decreasing, leading to earlier discharge. Currently in the United States, patients are most commonly monitored for 48 to 72 hours post PCI.² The current guidelines support this practice, recommending early discharge within 48 to 72 hours in selected low-risk patients if adequate follow-up and rehabilitation are arranged.³

Given the COVID-19 pandemic and decreased hospital bed availability, Rathod et al took one step further on the question of whether low-risk STEMI patients with primary PCI can be discharged safely within 48 hours with adequate follow-up. They found that at a median follow-up of 271 days, EHD patients had 2 COVID-related deaths, with 0% cardiovascular mortality and a MACE rate of 1.2%, including deaths, MI, and ischemic revascularization. The median time to discharge was 25 hours. This was noninferior to the >48-hour historical control group, which had mortality of 0.7% (P = 0.349) and a MACE rate of 1.9% (P = .674). The results remained similar after propensity matching for mortality (0.34% vs 0.69%, P = .410) or MACE (1.2% vs 1.9%, P = .342).

This is the first prospective study to systematically assess the safety and feasibility of discharge of low-risk STEMI patients with primary PCI within 48 hours. This study is unique in that it involved the use of telemedicine, including a virtual platform to collect data such as heart rate, blood pressure, and blood glucose, and virtual visits to facilitate follow-up and reduce clinic travel, cost, and potential COVID-19 exposure. The investigators’ protocol included virtual follow-up by cardiology advanced practitioners at 2, 6, and 8 weeks and by an interventional cardiologist at 12 weeks. This protocol led to an increase in patient satisfaction. The study’s main limitation is that it is a single-center trial with a smaller sample size. Further studies are necessary to confirm the safety and feasibility of this approach. In addition, further refinement of the patient selection criteria for EHD should be considered.

Applications for Clinical Practice

In low-risk STEMI patients after primary PCI, discharge within 48 hours may be considered if close follow-up is arranged. However, further studies are necessary to confirm this finding.

—Thai Nguyen, MD, Albert Chan, MD, and Taishi Hirai MD

Study Overview

Objective: To assess the safety and efficacy of early hospital discharge (EHD) for selected low-risk patients with ST-segment elevation myocardial infarction (STEMI) after primary percutaneous coronary intervention (PCI).

Design: Single-center retrospective analysis of prospectively collected data.

Setting and participants: An EHD group comprised of 600 patients who were discharged at <48 hours between April 2020 and June 2021 was compared to a control group of 700 patients who met EHD criteria but were discharged at >48 hour between October 2018 and June 2021. Patients were selected into the EHD group based on the following criteria, in accordance with recommendations from the European Society of Cardiology, and all patients had close follow-up with a combination of structured telephone follow-up at 48 hours post discharge and virtual visits at 2, 6, and 8 weeks and at 3 months:

• Left ventricular ejection fraction ≥40%
• Successful primary PCI (that achieved thrombolysis in myocardial infarction flow grade 3)
• Absence of severe nonculprit disease requiring further inpatient revascularization
• Absence of ischemic symptoms post PCI
• Absence of heart failure or hemodynamic instability
• Absence of significant arrhythmia (ventricular fibrillation, ventricular tachycardia, or atrial fibrillation or atrial flutter requiring prolonged stay)
• Mobility with suitable social circumstances for discharge

Main outcome measures: The outcomes measured were length of hospitalization and a composite primary endpoint of cardiovascular mortality and major adverse cardiovascular event (MACE) rates, defined as a composite of all-cause mortality, recurrent MI, and target lesion revascularization.

Main results: The median length of stay of hospitalization in the EHD group was 24.6 hours compared to 56.1 hours in the >48-hour historical control group. On median follow-up of 271 days, the EHD group demonstrated 0% cardiovascular mortality and a MACE rate of 1.2%. This was shown to be noninferior compared to the >48-hour historical control group, which had mortality of 0.7% and a MACE rate of 1.9%.

Conclusion: Selected low-risk STEMI patients can be safely discharged early with appropriate follow-up after primary PCI.

Commentary

Patients with STEMI have a higher risk of postprocedural adverse events such as MI, arrhythmia, or acute heart failure compared to patients with stable ischemic heart disease, and thus are monitored after primary PCI. Although patients were traditionally monitored for 5 to 7 days a few decades ago,¹ with improvements in PCI techniques, devices, and pharmacotherapy as well as in door-to-balloon time, the in-hospital complication rates for patients with STEMI have been decreasing, leading to earlier discharge. Currently in the United States, patients are most commonly monitored for 48 to 72 hours post PCI.² The current guidelines support this practice, recommending early discharge within 48 to 72 hours in selected low-risk patients if adequate follow-up and rehabilitation are arranged.³

Given the COVID-19 pandemic and decreased hospital bed availability, Rathod et al took one step further on the question of whether low-risk STEMI patients with primary PCI can be discharged safely within 48 hours with adequate follow-up. They found that at a median follow-up of 271 days, EHD patients had 2 COVID-related deaths, with 0% cardiovascular mortality and a MACE rate of 1.2%, including deaths, MI, and ischemic revascularization. The median time to discharge was 25 hours. This was noninferior to the >48-hour historical control group, which had mortality of 0.7% (P = 0.349) and a MACE rate of 1.9% (P = .674). The results remained similar after propensity matching for mortality (0.34% vs 0.69%, P = .410) or MACE (1.2% vs 1.9%, P = .342).

This is the first prospective study to systematically assess the safety and feasibility of discharge of low-risk STEMI patients with primary PCI within 48 hours. This study is unique in that it involved the use of telemedicine, including a virtual platform to collect data such as heart rate, blood pressure, and blood glucose, and virtual visits to facilitate follow-up and reduce clinic travel, cost, and potential COVID-19 exposure. The investigators’ protocol included virtual follow-up by cardiology advanced practitioners at 2, 6, and 8 weeks and by an interventional cardiologist at 12 weeks. This protocol led to an increase in patient satisfaction. The study’s main limitation is that it is a single-center trial with a smaller sample size. Further studies are necessary to confirm the safety and feasibility of this approach. In addition, further refinement of the patient selection criteria for EHD should be considered.

Applications for Clinical Practice

In low-risk STEMI patients after primary PCI, discharge within 48 hours may be considered if close follow-up is arranged. However, further studies are necessary to confirm this finding.

—Thai Nguyen, MD, Albert Chan, MD, and Taishi Hirai MD

Older adults admitted to post-acute settings frequently have complex rehabilitation needs and multimorbidity, which predisposes them to pain management challenges.^1,2 The prevalence of pain in post-acute and long-term care is as high as 65%, and opioid use is common among this population with 1 in 7 residents receiving long-term opioids.^3,4

Opioids that do not adequately control pain represent a missed opportunity for deprescribing. There is limited evidence regarding efficacy of long-term opioid use (> 90 days) for improving pain and physical functioning.⁵ In addition, long-term opioid use carries significant risks, including overdose-related death, dependence, and increased emergency department visits.⁵ These risks are likely to be pronounced among veterans receiving post-acute care (PAC) who are older, have comorbid psychiatric disorders, are prescribed several centrally acting medications, and experience substance use disorder (SUD).⁶

Older adults are at increased risk for opioid toxicity because of reduced drug clearance and smaller therapeutic window.⁵ Centers for Disease Control and Prevention (CDC) guidelines recommend frequently assessing patients for benefit in terms of sustained improvement in pain as well as physical function.⁵ If pain and functional improvements are minimal, opioid use and nonopioid pain management strategies should be considered. Some patients will struggle with this approach. Directly asking patients about the effectiveness of opioids is challenging. Opioid users with chronic pain frequently report problems with opioids even as they describe them as indispensable for pain management.^7,8

Earlier studies have assessed patient perspectives regarding opioid difficulties as well as their helpfulness, which could introduce recall bias. Patient-level factors that contribute to a global sense of distress, in addition to the presence of painful physical conditions, also could contribute to patients requesting opioids without experiencing adequate pain relief. One study in veterans residing in PAC facilities found that individuals with depression, posttraumatic stress disorder (PTSD), and SUD were more likely to report pain and receive scheduled analgesics; this effect persisted in individuals with PTSD even after adjusting for demographic and functional status variables.⁹ The study looked only at analgesics as a class and did not examine opioids specifically. It is possible that distressed individuals, such as those with uncontrolled depression, PTSD, and SUD, might be more likely to report high pain levels and receive opioids with inadequate benefit and increased risk. Identifying the primary condition causing distress and targeting treatment to that condition (ie, depression) is preferable to escalating opioids in an attempt to treat pain in the context of nonresponse. Assessing an individual’s aggregate response to opioids rather than relying on a single self-report is a useful addition to current pain management strategies.

The goal of this study was to pilot a method of identifying opioid-nonresponsive pain using administrative data, measure its prevalence in a PAC population of veterans, and explore clinical and demographic correlates with particular attention to variates that could indicate high levels of psychological and physical distress. Identifying pain that is poorly responsive to opioids would give clinicians the opportunity to avoid or minimize opioid use and prioritize treatments that are likely to improve the resident’s pain, quality of life, and physical function while minimizing recall bias. We hypothesized that pain that responds poorly to opioids would be prevalent among veterans residing in a PAC unit. We considered that veterans with pain poorly responsive to opioids would be more likely to have factors that would place them at increased risk of adverse effects, such as comorbid psychiatric conditions, history of SUD, and multimorbidity, providing further rationale for clinical equipoise in that population.⁶

Methods

This was a small, retrospective cross-sectional study using administrative data and chart review. The study included veterans who were administered opioids while residing in a single US Department of Veterans Affairs (VA) community living center PAC (CLC-PAC) unit during at least 1 of 4 nonconsecutive, random days in 2016 and 2017. The study was approved by the institutional review board of the Ann Arbor VA Health System (#2017-1034) as part of a larger project involving models of care in vulnerable older veterans.

Inclusion criteria were the presence of at least moderate pain (≥ 4 on a 0 to 10 scale); receiving ≥ 2 opioids ordered as needed over the prespecified 24-hour observation period; and having ≥ 2 pre-and postopioid administration pain scores during the observation period. Veterans who did not meet these criteria were excluded. At the time of initial sample selection, we did not capture information related to coprescribed analgesics, including a standing order of opioids. To obtain the sample, we initially characterized all veterans on the 4 days residing in the CLC-PAC unit as those reporting at least moderate pain (≥ 4) and those who reported no or mild pain (< 4). The cut point of 4 of 10 is consistent with moderate pain based on earlier work showing higher likelihood of pain that interferes with physical function.¹⁰ We then restricted the sample to veterans who received ≥ 2 opioids ordered as needed for pain and had ≥ 2 pre- and postopioid administration numeric pain rating scores during the 24-hour observation period. This methodology was chosen to enrich our sample for those who received opioids regularly for ongoing pain. Opioids were defined as full µ-opioid receptor agonists and included hydrocodone, oxycodone, morphine, hydromorphone, fentanyl, tramadol, and methadone.

Outcome of Interest

The primary outcome of interest was an individual’s response to as-needed opioids, which we defined as change in the pain score after opioid administration. The pre-opioid pain score was the score that immediately preceded administration of an as-needed opioid. The postopioid administration pain score was the first score after opioid administration if obtained within 3 hours of administration. Scores collected > 3 hours after opioid administration were excluded because they no longer accurately reflected the impact of the opioid due to the short half-lives. Observations were excluded if an opioid was administered without a recorded pain score; this occurred once for 6 individuals. Observations also were excluded if an opioid was administered but the data were captured on the following day (outside of the 24-hour window); this occurred once for 3 individuals.

We calculated a ∆ score by subtracting the postopioid pain rating score from the pre-opioid score. Individual ∆ scores were then averaged over the 24-hour period (range, 2-5 opioid doses). For example, if an individual reported a pre-opioid pain score of 10, and a postopioid pain score of 2, the ∆ was recorded as 8. If the individual’s next pre-opioid score was 10, and post-opioid score was 6, the ∆ was recorded as 4. ∆ scores over the 24-hour period were averaged together to determine that individual’s response to as-needed opioids. In the previous example, the mean ∆ score is 6. Lower mean ∆ scores reflect decreased responsiveness to opioids’ analgesic effect.

Demographic and clinical data were obtained from electronic health record review using a standardized assessment form. These data included information about medical and psychiatric comorbidities, specialist consultations, and CLC-PAC unit admission indications and diagnoses. Medications of interest were categorized as antidepressants, antipsychotics, benzodiazepines, muscle relaxants, hypnotics, stimulants, antiepileptic drugs/mood stabilizers (including gabapentin and pregabalin), and all adjuvant analgesics. Adjuvant analgesics were defined as medications administered for pain as documented by chart notes or those ordered as needed for pain, and analyzed as a composite variable. Antidepressants with analgesic properties (serotonin-norepinephrine reuptake inhibitors and tricyclic antidepressants) were considered adjuvant analgesics. Psychiatric information collected included presence of mood, anxiety, and psychotic disorders, and PTSD. SUD information was collected separately from other psychiatric disorders.

Analyses

The study population was described using tabulations for categorical data and means and standard deviations for continuous data. Responsiveness to opioids was analyzed as a continuous variable. Those with higher mean ∆ scores were considered to have pain relatively more responsive to opioids, while lower mean ∆ scores indicated pain less responsive to opioids. We constructed linear regression models controlling for average pre-opioid pain rating scores to explore associations between opioid responsiveness and variables of interest. All analyses were completed using Stata version 15. This study was not adequately powered to detect differences across the spectrum of opioid responsiveness, although the authors have reported differences in this article.

Results

Over the 4-day observational period there were 146 resident-days. Of these, 88 (60.3%) reported at least 1 pain score of ≥ 4. Of those, 61 (41.8%) received ≥ 1 as-needed opioid for pain. We identified 46 resident-days meeting study criteria of ≥ 2 pre- and postanalgesic scores. We identified 41 unique individuals (Figure 1). Two individuals were admitted to the CLC-PAC unit on 2 of the 4 observation days, and 1 individual was admitted to the CLC-PAC unit on 3 of the 4 observation days. For individuals admitted several days, we included data only from the initial observation day.

Response to opioids varied greatly in this sample. The mean (SD) ∆ pain score was 3.4 (1.6) and ranged from 0.5 to 6.3. Using linear regression, we found no relationship between admission indication, medical comorbidities (including active cancer), and opioid responsiveness (Table).

A poorer response to opioids was associated with a significantly higher as-needed opioid dosage (β = −0.02; 95% CI, −0.04 to −0.01; P = .002) as well as a trend toward higher total opioid dosage (β = −0.005; 95% CI, −0.01 to 0.0003; P = .06) (Figure 4). Thirty-eight (92.7%) participants received nonopioid adjuvant analgesics for pain. More than half (56.1%) received antidepressants or gabapentinoids (51.2%), although we did not assess whether they were prescribed for pain or another indication. We did not identify a relationship between any specific psychoactive drug class and opioid responsiveness in this sample.

Discussion

This exploratory study used readily available administrative data in a CLC-PAC unit to assess responsiveness to opioids via a numeric mean ∆ score, with higher values indicating more pain relief in response to opioids. We then constructed linear regression models to characterize the relationship between the mean ∆ score and factors known to be associated with difficult-to-control pain and psychosocial distress. As expected, opioid responsiveness was highly variable among residents; some residents experienced essentially no reduction in pain, on average, despite receiving opioids. Psychiatric comorbidity, higher dosage in OMEs, and the presence of a pain service consult significantly correlated with poorer response to opioids. To our knowledge, this is the first study to quantify opioid responsiveness and describe the relationship with clinical correlates in the understudied PAC population.

Earlier research has demonstrated a relationship between the presence of psychiatric disorders and increased likelihood of receiving any analgesics among veterans residing in PAC.⁹ Our study adds to the literature by quantifying opioid response using readily available administrative data and examining associations with psychiatric diagnoses. These findings highlight the possibility that attempting to treat high levels of pain by escalating the opioid dosage in patients with a comorbid psychiatric diagnosis should be re-addressed, particularly if there is no meaningful pain reduction at lower opioid dosages. Our sample had a variety of admission diagnoses and medical comorbidities, however, we did not identify a relationship with opioid responsiveness, including an active cancer diagnosis. Although SUDs were highly prevalent in our sample, there was no relationship with opioid responsiveness. This suggests that lack of response to opioids is not merely a matter of drug tolerance or an indication of drug-seeking behavior.

Factors Impacting Response

Many factors could affect whether an individual obtains an adequate analgesic response to opioids or other pain medications, including variations in genes encoding opioid receptors and hepatic enzymes involved in drug metabolism and an individual’s opioid exposure history.¹³ The phenomenon of requiring more drug to produce the same relief after repeated exposures (ie, tolerance) is well known.¹⁴ Opioid-induced hyperalgesia is a phenomenon whereby a patient’s overall pain increases while receiving opioids, but each opioid dose might be perceived as beneficial.¹⁵ Increasingly, psychosocial distress is an important factor in opioid response. Adverse selection is the process culminating in those with psychosocial distress and/or SUDs being prescribed more opioids for longer durations.¹⁶ Our data suggests that this process could play a role in PAC settings. In addition, exaggerating pain to obtain additional opioids for nonmedical purposes, such as euphoria or relaxation, also is possible.¹⁷

When clinically assessing an individual whose pain is not well controlled despite escalating opioid dosages, prescribers must consider which of these factors likely is predominant. However, the first step of determining who has a poor opioid response is not straightforward. Directly asking patients is challenging; many individuals perceive opioids to be helpful while simultaneously reporting inadequately controlled pain.^7,8 The primary value of this study is the possibility of providing prescribers a quick, simple method of assessing a patient’s response to opioids. Using this method, individuals who are responding poorly to opioids, including those who might exaggerate pain for secondary gain, could be identified. Health care professionals could consider revisiting pain management strategies, assess for the presence of OUD, or evaluate other contributors to inadequately controlled pain. Although we only collected data regarding response to opioids in this study, any pain medication administered as needed (ie, nonsteroidal anti-inflammatory drugs, acetaminophen) could be analyzed using this methodology, allowing identification of other helpful pain management strategies. We began the validation process with extensive chart review, but further validation is required before this method can be applied to routine clinical practice.

Patients who report uncontrolled pain despite receiving opioids are a clinically challenging population. The traditional strategy has been to escalate opioids, which is recommended by the World Health Organization stepladder approach for patients with cancer pain and limited life expectancy.¹⁸ Applying this approach to a general population of patients with chronic pain is ineffective and dangerous.¹⁹ The CDC and the VA/US Department of Defense (VA/DoD) guidelines both recommend carefully reassessing risks and benefits at total daily dosages > 50 OME and avoid increasing dosages to > 90 OME daily in most circumstances.^5,20 Our finding that participants taking higher dosages of opioids were not more likely to have better control over their pain supports this recommendation.

Limitations

This study has several limitations, the most significant is its small sample size because of the exploratory nature of the project. Results are based on a small pilot sample enriched to include individuals with at least moderate pain who receive opioids frequently at 1 VA CLC-PAC unit; therefore, the results might not be representative of all veterans or a more general population. Our small sample size limits power to detect small differences. Data collected should be used to inform formal power calculations before subsequent larger studies to select adequate sample size. Validation studies, including samples from the same population using different dates, which reproduce findings are an important step. Moreover, we only had data on a single dimension of pain (intensity/severity), as measured by the pain scale, which nursing staff used to make a real-time clinical decision of whether to administer an as-needed opioid. Future studies should consider using pain measures that provide multidimensional assessment (ie, severity, functional interference) and/or were developed specifically for veterans, such as the Defense and Veterans Pain Rating Scale.²¹

Our study was cross-sectional in nature and addressed a single 24-hour period of data per participant. The years of data collection (2016 and 2017) followed a decline in overall opioid prescribing that has continued, likely influenced by CDC and VA/DoD guidelines.²² It is unclear whether our observations are an accurate reflection of individuals’ response over time or whether prescribing practices in PAC have shifted.

We did not consider the type of pain being treated or explore clinicians’ reasons for prescribing opioids, therefore limiting our ability to know whether opioids were indicated. Information regarding OUD and other SUDs was limited to what was documented in the chart during the CLC-PAC unit admission. We did not have information on length of exposure to opioids. It is possible that opioid tolerance could play a role in reducing opioid responsiveness. However, simple tolerance would not be expected to explain robust correlations with psychiatric comorbidities. Also, simple tolerance would be expected to be overcome with higher opioid dosages, whereas our study demonstrates less responsiveness. These data suggests that some individuals’ pain might be poorly opioid responsive, and psychiatric factors could increase this risk. We used a novel data source in combination with chart review; to our knowledge, barcode medication administration data have not been used in this manner previously. Future work needs to validate this method, using larger sample sizes and several clinical sites. Finally, we used regression models that controlled for average pre-opioid pain rating scores, which is only 1 covariate important for examining effects. Larger studies with adequate power should control for multiple covariates known to be associated with pain and opioid response.

Conclusions

Opioid responsiveness is important clinically yet challenging to assess. This pilot study identifies a way of classifying pain as relatively opioid nonresponsive using administrative data but requires further validation before considering scaling for more general use. The possibility that a substantial percentage of residents in a CLC-PAC unit could be receiving increasing dosages of opioids without adequate benefit justifies the need for more research and underscores the need for prescribers to assess individuals frequently for ongoing benefit of opioids regardless of diagnosis or mechanism of pain.

Acknowledgments

The authors thank Andrzej Galecki, Corey Powell, and the University of Michigan Consulting for Statistics, Computing and Analytics Research Center for assistance with statistical analysis.

Older adults admitted to post-acute settings frequently have complex rehabilitation needs and multimorbidity, which predisposes them to pain management challenges.^1,2 The prevalence of pain in post-acute and long-term care is as high as 65%, and opioid use is common among this population with 1 in 7 residents receiving long-term opioids.^3,4

Opioids that do not adequately control pain represent a missed opportunity for deprescribing. There is limited evidence regarding efficacy of long-term opioid use (> 90 days) for improving pain and physical functioning.⁵ In addition, long-term opioid use carries significant risks, including overdose-related death, dependence, and increased emergency department visits.⁵ These risks are likely to be pronounced among veterans receiving post-acute care (PAC) who are older, have comorbid psychiatric disorders, are prescribed several centrally acting medications, and experience substance use disorder (SUD).⁶

Older adults are at increased risk for opioid toxicity because of reduced drug clearance and smaller therapeutic window.⁵ Centers for Disease Control and Prevention (CDC) guidelines recommend frequently assessing patients for benefit in terms of sustained improvement in pain as well as physical function.⁵ If pain and functional improvements are minimal, opioid use and nonopioid pain management strategies should be considered. Some patients will struggle with this approach. Directly asking patients about the effectiveness of opioids is challenging. Opioid users with chronic pain frequently report problems with opioids even as they describe them as indispensable for pain management.^7,8

Earlier studies have assessed patient perspectives regarding opioid difficulties as well as their helpfulness, which could introduce recall bias. Patient-level factors that contribute to a global sense of distress, in addition to the presence of painful physical conditions, also could contribute to patients requesting opioids without experiencing adequate pain relief. One study in veterans residing in PAC facilities found that individuals with depression, posttraumatic stress disorder (PTSD), and SUD were more likely to report pain and receive scheduled analgesics; this effect persisted in individuals with PTSD even after adjusting for demographic and functional status variables.⁹ The study looked only at analgesics as a class and did not examine opioids specifically. It is possible that distressed individuals, such as those with uncontrolled depression, PTSD, and SUD, might be more likely to report high pain levels and receive opioids with inadequate benefit and increased risk. Identifying the primary condition causing distress and targeting treatment to that condition (ie, depression) is preferable to escalating opioids in an attempt to treat pain in the context of nonresponse. Assessing an individual’s aggregate response to opioids rather than relying on a single self-report is a useful addition to current pain management strategies.

The goal of this study was to pilot a method of identifying opioid-nonresponsive pain using administrative data, measure its prevalence in a PAC population of veterans, and explore clinical and demographic correlates with particular attention to variates that could indicate high levels of psychological and physical distress. Identifying pain that is poorly responsive to opioids would give clinicians the opportunity to avoid or minimize opioid use and prioritize treatments that are likely to improve the resident’s pain, quality of life, and physical function while minimizing recall bias. We hypothesized that pain that responds poorly to opioids would be prevalent among veterans residing in a PAC unit. We considered that veterans with pain poorly responsive to opioids would be more likely to have factors that would place them at increased risk of adverse effects, such as comorbid psychiatric conditions, history of SUD, and multimorbidity, providing further rationale for clinical equipoise in that population.⁶

Methods

This was a small, retrospective cross-sectional study using administrative data and chart review. The study included veterans who were administered opioids while residing in a single US Department of Veterans Affairs (VA) community living center PAC (CLC-PAC) unit during at least 1 of 4 nonconsecutive, random days in 2016 and 2017. The study was approved by the institutional review board of the Ann Arbor VA Health System (#2017-1034) as part of a larger project involving models of care in vulnerable older veterans.

Inclusion criteria were the presence of at least moderate pain (≥ 4 on a 0 to 10 scale); receiving ≥ 2 opioids ordered as needed over the prespecified 24-hour observation period; and having ≥ 2 pre-and postopioid administration pain scores during the observation period. Veterans who did not meet these criteria were excluded. At the time of initial sample selection, we did not capture information related to coprescribed analgesics, including a standing order of opioids. To obtain the sample, we initially characterized all veterans on the 4 days residing in the CLC-PAC unit as those reporting at least moderate pain (≥ 4) and those who reported no or mild pain (< 4). The cut point of 4 of 10 is consistent with moderate pain based on earlier work showing higher likelihood of pain that interferes with physical function.¹⁰ We then restricted the sample to veterans who received ≥ 2 opioids ordered as needed for pain and had ≥ 2 pre- and postopioid administration numeric pain rating scores during the 24-hour observation period. This methodology was chosen to enrich our sample for those who received opioids regularly for ongoing pain. Opioids were defined as full µ-opioid receptor agonists and included hydrocodone, oxycodone, morphine, hydromorphone, fentanyl, tramadol, and methadone.

Outcome of Interest

The primary outcome of interest was an individual’s response to as-needed opioids, which we defined as change in the pain score after opioid administration. The pre-opioid pain score was the score that immediately preceded administration of an as-needed opioid. The postopioid administration pain score was the first score after opioid administration if obtained within 3 hours of administration. Scores collected > 3 hours after opioid administration were excluded because they no longer accurately reflected the impact of the opioid due to the short half-lives. Observations were excluded if an opioid was administered without a recorded pain score; this occurred once for 6 individuals. Observations also were excluded if an opioid was administered but the data were captured on the following day (outside of the 24-hour window); this occurred once for 3 individuals.

We calculated a ∆ score by subtracting the postopioid pain rating score from the pre-opioid score. Individual ∆ scores were then averaged over the 24-hour period (range, 2-5 opioid doses). For example, if an individual reported a pre-opioid pain score of 10, and a postopioid pain score of 2, the ∆ was recorded as 8. If the individual’s next pre-opioid score was 10, and post-opioid score was 6, the ∆ was recorded as 4. ∆ scores over the 24-hour period were averaged together to determine that individual’s response to as-needed opioids. In the previous example, the mean ∆ score is 6. Lower mean ∆ scores reflect decreased responsiveness to opioids’ analgesic effect.

Demographic and clinical data were obtained from electronic health record review using a standardized assessment form. These data included information about medical and psychiatric comorbidities, specialist consultations, and CLC-PAC unit admission indications and diagnoses. Medications of interest were categorized as antidepressants, antipsychotics, benzodiazepines, muscle relaxants, hypnotics, stimulants, antiepileptic drugs/mood stabilizers (including gabapentin and pregabalin), and all adjuvant analgesics. Adjuvant analgesics were defined as medications administered for pain as documented by chart notes or those ordered as needed for pain, and analyzed as a composite variable. Antidepressants with analgesic properties (serotonin-norepinephrine reuptake inhibitors and tricyclic antidepressants) were considered adjuvant analgesics. Psychiatric information collected included presence of mood, anxiety, and psychotic disorders, and PTSD. SUD information was collected separately from other psychiatric disorders.

Analyses

The study population was described using tabulations for categorical data and means and standard deviations for continuous data. Responsiveness to opioids was analyzed as a continuous variable. Those with higher mean ∆ scores were considered to have pain relatively more responsive to opioids, while lower mean ∆ scores indicated pain less responsive to opioids. We constructed linear regression models controlling for average pre-opioid pain rating scores to explore associations between opioid responsiveness and variables of interest. All analyses were completed using Stata version 15. This study was not adequately powered to detect differences across the spectrum of opioid responsiveness, although the authors have reported differences in this article.

Results

Over the 4-day observational period there were 146 resident-days. Of these, 88 (60.3%) reported at least 1 pain score of ≥ 4. Of those, 61 (41.8%) received ≥ 1 as-needed opioid for pain. We identified 46 resident-days meeting study criteria of ≥ 2 pre- and postanalgesic scores. We identified 41 unique individuals (Figure 1). Two individuals were admitted to the CLC-PAC unit on 2 of the 4 observation days, and 1 individual was admitted to the CLC-PAC unit on 3 of the 4 observation days. For individuals admitted several days, we included data only from the initial observation day.

Response to opioids varied greatly in this sample. The mean (SD) ∆ pain score was 3.4 (1.6) and ranged from 0.5 to 6.3. Using linear regression, we found no relationship between admission indication, medical comorbidities (including active cancer), and opioid responsiveness (Table).

A poorer response to opioids was associated with a significantly higher as-needed opioid dosage (β = −0.02; 95% CI, −0.04 to −0.01; P = .002) as well as a trend toward higher total opioid dosage (β = −0.005; 95% CI, −0.01 to 0.0003; P = .06) (Figure 4). Thirty-eight (92.7%) participants received nonopioid adjuvant analgesics for pain. More than half (56.1%) received antidepressants or gabapentinoids (51.2%), although we did not assess whether they were prescribed for pain or another indication. We did not identify a relationship between any specific psychoactive drug class and opioid responsiveness in this sample.

Discussion

This exploratory study used readily available administrative data in a CLC-PAC unit to assess responsiveness to opioids via a numeric mean ∆ score, with higher values indicating more pain relief in response to opioids. We then constructed linear regression models to characterize the relationship between the mean ∆ score and factors known to be associated with difficult-to-control pain and psychosocial distress. As expected, opioid responsiveness was highly variable among residents; some residents experienced essentially no reduction in pain, on average, despite receiving opioids. Psychiatric comorbidity, higher dosage in OMEs, and the presence of a pain service consult significantly correlated with poorer response to opioids. To our knowledge, this is the first study to quantify opioid responsiveness and describe the relationship with clinical correlates in the understudied PAC population.

Earlier research has demonstrated a relationship between the presence of psychiatric disorders and increased likelihood of receiving any analgesics among veterans residing in PAC.⁹ Our study adds to the literature by quantifying opioid response using readily available administrative data and examining associations with psychiatric diagnoses. These findings highlight the possibility that attempting to treat high levels of pain by escalating the opioid dosage in patients with a comorbid psychiatric diagnosis should be re-addressed, particularly if there is no meaningful pain reduction at lower opioid dosages. Our sample had a variety of admission diagnoses and medical comorbidities, however, we did not identify a relationship with opioid responsiveness, including an active cancer diagnosis. Although SUDs were highly prevalent in our sample, there was no relationship with opioid responsiveness. This suggests that lack of response to opioids is not merely a matter of drug tolerance or an indication of drug-seeking behavior.

Factors Impacting Response

Many factors could affect whether an individual obtains an adequate analgesic response to opioids or other pain medications, including variations in genes encoding opioid receptors and hepatic enzymes involved in drug metabolism and an individual’s opioid exposure history.¹³ The phenomenon of requiring more drug to produce the same relief after repeated exposures (ie, tolerance) is well known.¹⁴ Opioid-induced hyperalgesia is a phenomenon whereby a patient’s overall pain increases while receiving opioids, but each opioid dose might be perceived as beneficial.¹⁵ Increasingly, psychosocial distress is an important factor in opioid response. Adverse selection is the process culminating in those with psychosocial distress and/or SUDs being prescribed more opioids for longer durations.¹⁶ Our data suggests that this process could play a role in PAC settings. In addition, exaggerating pain to obtain additional opioids for nonmedical purposes, such as euphoria or relaxation, also is possible.¹⁷

When clinically assessing an individual whose pain is not well controlled despite escalating opioid dosages, prescribers must consider which of these factors likely is predominant. However, the first step of determining who has a poor opioid response is not straightforward. Directly asking patients is challenging; many individuals perceive opioids to be helpful while simultaneously reporting inadequately controlled pain.^7,8 The primary value of this study is the possibility of providing prescribers a quick, simple method of assessing a patient’s response to opioids. Using this method, individuals who are responding poorly to opioids, including those who might exaggerate pain for secondary gain, could be identified. Health care professionals could consider revisiting pain management strategies, assess for the presence of OUD, or evaluate other contributors to inadequately controlled pain. Although we only collected data regarding response to opioids in this study, any pain medication administered as needed (ie, nonsteroidal anti-inflammatory drugs, acetaminophen) could be analyzed using this methodology, allowing identification of other helpful pain management strategies. We began the validation process with extensive chart review, but further validation is required before this method can be applied to routine clinical practice.

Patients who report uncontrolled pain despite receiving opioids are a clinically challenging population. The traditional strategy has been to escalate opioids, which is recommended by the World Health Organization stepladder approach for patients with cancer pain and limited life expectancy.¹⁸ Applying this approach to a general population of patients with chronic pain is ineffective and dangerous.¹⁹ The CDC and the VA/US Department of Defense (VA/DoD) guidelines both recommend carefully reassessing risks and benefits at total daily dosages > 50 OME and avoid increasing dosages to > 90 OME daily in most circumstances.^5,20 Our finding that participants taking higher dosages of opioids were not more likely to have better control over their pain supports this recommendation.

Limitations

This study has several limitations, the most significant is its small sample size because of the exploratory nature of the project. Results are based on a small pilot sample enriched to include individuals with at least moderate pain who receive opioids frequently at 1 VA CLC-PAC unit; therefore, the results might not be representative of all veterans or a more general population. Our small sample size limits power to detect small differences. Data collected should be used to inform formal power calculations before subsequent larger studies to select adequate sample size. Validation studies, including samples from the same population using different dates, which reproduce findings are an important step. Moreover, we only had data on a single dimension of pain (intensity/severity), as measured by the pain scale, which nursing staff used to make a real-time clinical decision of whether to administer an as-needed opioid. Future studies should consider using pain measures that provide multidimensional assessment (ie, severity, functional interference) and/or were developed specifically for veterans, such as the Defense and Veterans Pain Rating Scale.²¹

Our study was cross-sectional in nature and addressed a single 24-hour period of data per participant. The years of data collection (2016 and 2017) followed a decline in overall opioid prescribing that has continued, likely influenced by CDC and VA/DoD guidelines.²² It is unclear whether our observations are an accurate reflection of individuals’ response over time or whether prescribing practices in PAC have shifted.

We did not consider the type of pain being treated or explore clinicians’ reasons for prescribing opioids, therefore limiting our ability to know whether opioids were indicated. Information regarding OUD and other SUDs was limited to what was documented in the chart during the CLC-PAC unit admission. We did not have information on length of exposure to opioids. It is possible that opioid tolerance could play a role in reducing opioid responsiveness. However, simple tolerance would not be expected to explain robust correlations with psychiatric comorbidities. Also, simple tolerance would be expected to be overcome with higher opioid dosages, whereas our study demonstrates less responsiveness. These data suggests that some individuals’ pain might be poorly opioid responsive, and psychiatric factors could increase this risk. We used a novel data source in combination with chart review; to our knowledge, barcode medication administration data have not been used in this manner previously. Future work needs to validate this method, using larger sample sizes and several clinical sites. Finally, we used regression models that controlled for average pre-opioid pain rating scores, which is only 1 covariate important for examining effects. Larger studies with adequate power should control for multiple covariates known to be associated with pain and opioid response.

Conclusions

Opioid responsiveness is important clinically yet challenging to assess. This pilot study identifies a way of classifying pain as relatively opioid nonresponsive using administrative data but requires further validation before considering scaling for more general use. The possibility that a substantial percentage of residents in a CLC-PAC unit could be receiving increasing dosages of opioids without adequate benefit justifies the need for more research and underscores the need for prescribers to assess individuals frequently for ongoing benefit of opioids regardless of diagnosis or mechanism of pain.

Acknowledgments

The authors thank Andrzej Galecki, Corey Powell, and the University of Michigan Consulting for Statistics, Computing and Analytics Research Center for assistance with statistical analysis.

Older adults admitted to post-acute settings frequently have complex rehabilitation needs and multimorbidity, which predisposes them to pain management challenges.^1,2 The prevalence of pain in post-acute and long-term care is as high as 65%, and opioid use is common among this population with 1 in 7 residents receiving long-term opioids.^3,4

Opioids that do not adequately control pain represent a missed opportunity for deprescribing. There is limited evidence regarding efficacy of long-term opioid use (> 90 days) for improving pain and physical functioning.⁵ In addition, long-term opioid use carries significant risks, including overdose-related death, dependence, and increased emergency department visits.⁵ These risks are likely to be pronounced among veterans receiving post-acute care (PAC) who are older, have comorbid psychiatric disorders, are prescribed several centrally acting medications, and experience substance use disorder (SUD).⁶

Older adults are at increased risk for opioid toxicity because of reduced drug clearance and smaller therapeutic window.⁵ Centers for Disease Control and Prevention (CDC) guidelines recommend frequently assessing patients for benefit in terms of sustained improvement in pain as well as physical function.⁵ If pain and functional improvements are minimal, opioid use and nonopioid pain management strategies should be considered. Some patients will struggle with this approach. Directly asking patients about the effectiveness of opioids is challenging. Opioid users with chronic pain frequently report problems with opioids even as they describe them as indispensable for pain management.^7,8

Earlier studies have assessed patient perspectives regarding opioid difficulties as well as their helpfulness, which could introduce recall bias. Patient-level factors that contribute to a global sense of distress, in addition to the presence of painful physical conditions, also could contribute to patients requesting opioids without experiencing adequate pain relief. One study in veterans residing in PAC facilities found that individuals with depression, posttraumatic stress disorder (PTSD), and SUD were more likely to report pain and receive scheduled analgesics; this effect persisted in individuals with PTSD even after adjusting for demographic and functional status variables.⁹ The study looked only at analgesics as a class and did not examine opioids specifically. It is possible that distressed individuals, such as those with uncontrolled depression, PTSD, and SUD, might be more likely to report high pain levels and receive opioids with inadequate benefit and increased risk. Identifying the primary condition causing distress and targeting treatment to that condition (ie, depression) is preferable to escalating opioids in an attempt to treat pain in the context of nonresponse. Assessing an individual’s aggregate response to opioids rather than relying on a single self-report is a useful addition to current pain management strategies.

The goal of this study was to pilot a method of identifying opioid-nonresponsive pain using administrative data, measure its prevalence in a PAC population of veterans, and explore clinical and demographic correlates with particular attention to variates that could indicate high levels of psychological and physical distress. Identifying pain that is poorly responsive to opioids would give clinicians the opportunity to avoid or minimize opioid use and prioritize treatments that are likely to improve the resident’s pain, quality of life, and physical function while minimizing recall bias. We hypothesized that pain that responds poorly to opioids would be prevalent among veterans residing in a PAC unit. We considered that veterans with pain poorly responsive to opioids would be more likely to have factors that would place them at increased risk of adverse effects, such as comorbid psychiatric conditions, history of SUD, and multimorbidity, providing further rationale for clinical equipoise in that population.⁶

Methods

This was a small, retrospective cross-sectional study using administrative data and chart review. The study included veterans who were administered opioids while residing in a single US Department of Veterans Affairs (VA) community living center PAC (CLC-PAC) unit during at least 1 of 4 nonconsecutive, random days in 2016 and 2017. The study was approved by the institutional review board of the Ann Arbor VA Health System (#2017-1034) as part of a larger project involving models of care in vulnerable older veterans.

Inclusion criteria were the presence of at least moderate pain (≥ 4 on a 0 to 10 scale); receiving ≥ 2 opioids ordered as needed over the prespecified 24-hour observation period; and having ≥ 2 pre-and postopioid administration pain scores during the observation period. Veterans who did not meet these criteria were excluded. At the time of initial sample selection, we did not capture information related to coprescribed analgesics, including a standing order of opioids. To obtain the sample, we initially characterized all veterans on the 4 days residing in the CLC-PAC unit as those reporting at least moderate pain (≥ 4) and those who reported no or mild pain (< 4). The cut point of 4 of 10 is consistent with moderate pain based on earlier work showing higher likelihood of pain that interferes with physical function.¹⁰ We then restricted the sample to veterans who received ≥ 2 opioids ordered as needed for pain and had ≥ 2 pre- and postopioid administration numeric pain rating scores during the 24-hour observation period. This methodology was chosen to enrich our sample for those who received opioids regularly for ongoing pain. Opioids were defined as full µ-opioid receptor agonists and included hydrocodone, oxycodone, morphine, hydromorphone, fentanyl, tramadol, and methadone.

Outcome of Interest

The primary outcome of interest was an individual’s response to as-needed opioids, which we defined as change in the pain score after opioid administration. The pre-opioid pain score was the score that immediately preceded administration of an as-needed opioid. The postopioid administration pain score was the first score after opioid administration if obtained within 3 hours of administration. Scores collected > 3 hours after opioid administration were excluded because they no longer accurately reflected the impact of the opioid due to the short half-lives. Observations were excluded if an opioid was administered without a recorded pain score; this occurred once for 6 individuals. Observations also were excluded if an opioid was administered but the data were captured on the following day (outside of the 24-hour window); this occurred once for 3 individuals.

We calculated a ∆ score by subtracting the postopioid pain rating score from the pre-opioid score. Individual ∆ scores were then averaged over the 24-hour period (range, 2-5 opioid doses). For example, if an individual reported a pre-opioid pain score of 10, and a postopioid pain score of 2, the ∆ was recorded as 8. If the individual’s next pre-opioid score was 10, and post-opioid score was 6, the ∆ was recorded as 4. ∆ scores over the 24-hour period were averaged together to determine that individual’s response to as-needed opioids. In the previous example, the mean ∆ score is 6. Lower mean ∆ scores reflect decreased responsiveness to opioids’ analgesic effect.

Demographic and clinical data were obtained from electronic health record review using a standardized assessment form. These data included information about medical and psychiatric comorbidities, specialist consultations, and CLC-PAC unit admission indications and diagnoses. Medications of interest were categorized as antidepressants, antipsychotics, benzodiazepines, muscle relaxants, hypnotics, stimulants, antiepileptic drugs/mood stabilizers (including gabapentin and pregabalin), and all adjuvant analgesics. Adjuvant analgesics were defined as medications administered for pain as documented by chart notes or those ordered as needed for pain, and analyzed as a composite variable. Antidepressants with analgesic properties (serotonin-norepinephrine reuptake inhibitors and tricyclic antidepressants) were considered adjuvant analgesics. Psychiatric information collected included presence of mood, anxiety, and psychotic disorders, and PTSD. SUD information was collected separately from other psychiatric disorders.

Analyses

The study population was described using tabulations for categorical data and means and standard deviations for continuous data. Responsiveness to opioids was analyzed as a continuous variable. Those with higher mean ∆ scores were considered to have pain relatively more responsive to opioids, while lower mean ∆ scores indicated pain less responsive to opioids. We constructed linear regression models controlling for average pre-opioid pain rating scores to explore associations between opioid responsiveness and variables of interest. All analyses were completed using Stata version 15. This study was not adequately powered to detect differences across the spectrum of opioid responsiveness, although the authors have reported differences in this article.

Results

Over the 4-day observational period there were 146 resident-days. Of these, 88 (60.3%) reported at least 1 pain score of ≥ 4. Of those, 61 (41.8%) received ≥ 1 as-needed opioid for pain. We identified 46 resident-days meeting study criteria of ≥ 2 pre- and postanalgesic scores. We identified 41 unique individuals (Figure 1). Two individuals were admitted to the CLC-PAC unit on 2 of the 4 observation days, and 1 individual was admitted to the CLC-PAC unit on 3 of the 4 observation days. For individuals admitted several days, we included data only from the initial observation day.

Response to opioids varied greatly in this sample. The mean (SD) ∆ pain score was 3.4 (1.6) and ranged from 0.5 to 6.3. Using linear regression, we found no relationship between admission indication, medical comorbidities (including active cancer), and opioid responsiveness (Table).

A poorer response to opioids was associated with a significantly higher as-needed opioid dosage (β = −0.02; 95% CI, −0.04 to −0.01; P = .002) as well as a trend toward higher total opioid dosage (β = −0.005; 95% CI, −0.01 to 0.0003; P = .06) (Figure 4). Thirty-eight (92.7%) participants received nonopioid adjuvant analgesics for pain. More than half (56.1%) received antidepressants or gabapentinoids (51.2%), although we did not assess whether they were prescribed for pain or another indication. We did not identify a relationship between any specific psychoactive drug class and opioid responsiveness in this sample.

Discussion

This exploratory study used readily available administrative data in a CLC-PAC unit to assess responsiveness to opioids via a numeric mean ∆ score, with higher values indicating more pain relief in response to opioids. We then constructed linear regression models to characterize the relationship between the mean ∆ score and factors known to be associated with difficult-to-control pain and psychosocial distress. As expected, opioid responsiveness was highly variable among residents; some residents experienced essentially no reduction in pain, on average, despite receiving opioids. Psychiatric comorbidity, higher dosage in OMEs, and the presence of a pain service consult significantly correlated with poorer response to opioids. To our knowledge, this is the first study to quantify opioid responsiveness and describe the relationship with clinical correlates in the understudied PAC population.

Earlier research has demonstrated a relationship between the presence of psychiatric disorders and increased likelihood of receiving any analgesics among veterans residing in PAC.⁹ Our study adds to the literature by quantifying opioid response using readily available administrative data and examining associations with psychiatric diagnoses. These findings highlight the possibility that attempting to treat high levels of pain by escalating the opioid dosage in patients with a comorbid psychiatric diagnosis should be re-addressed, particularly if there is no meaningful pain reduction at lower opioid dosages. Our sample had a variety of admission diagnoses and medical comorbidities, however, we did not identify a relationship with opioid responsiveness, including an active cancer diagnosis. Although SUDs were highly prevalent in our sample, there was no relationship with opioid responsiveness. This suggests that lack of response to opioids is not merely a matter of drug tolerance or an indication of drug-seeking behavior.

Factors Impacting Response

Many factors could affect whether an individual obtains an adequate analgesic response to opioids or other pain medications, including variations in genes encoding opioid receptors and hepatic enzymes involved in drug metabolism and an individual’s opioid exposure history.¹³ The phenomenon of requiring more drug to produce the same relief after repeated exposures (ie, tolerance) is well known.¹⁴ Opioid-induced hyperalgesia is a phenomenon whereby a patient’s overall pain increases while receiving opioids, but each opioid dose might be perceived as beneficial.¹⁵ Increasingly, psychosocial distress is an important factor in opioid response. Adverse selection is the process culminating in those with psychosocial distress and/or SUDs being prescribed more opioids for longer durations.¹⁶ Our data suggests that this process could play a role in PAC settings. In addition, exaggerating pain to obtain additional opioids for nonmedical purposes, such as euphoria or relaxation, also is possible.¹⁷

When clinically assessing an individual whose pain is not well controlled despite escalating opioid dosages, prescribers must consider which of these factors likely is predominant. However, the first step of determining who has a poor opioid response is not straightforward. Directly asking patients is challenging; many individuals perceive opioids to be helpful while simultaneously reporting inadequately controlled pain.^7,8 The primary value of this study is the possibility of providing prescribers a quick, simple method of assessing a patient’s response to opioids. Using this method, individuals who are responding poorly to opioids, including those who might exaggerate pain for secondary gain, could be identified. Health care professionals could consider revisiting pain management strategies, assess for the presence of OUD, or evaluate other contributors to inadequately controlled pain. Although we only collected data regarding response to opioids in this study, any pain medication administered as needed (ie, nonsteroidal anti-inflammatory drugs, acetaminophen) could be analyzed using this methodology, allowing identification of other helpful pain management strategies. We began the validation process with extensive chart review, but further validation is required before this method can be applied to routine clinical practice.

Patients who report uncontrolled pain despite receiving opioids are a clinically challenging population. The traditional strategy has been to escalate opioids, which is recommended by the World Health Organization stepladder approach for patients with cancer pain and limited life expectancy.¹⁸ Applying this approach to a general population of patients with chronic pain is ineffective and dangerous.¹⁹ The CDC and the VA/US Department of Defense (VA/DoD) guidelines both recommend carefully reassessing risks and benefits at total daily dosages > 50 OME and avoid increasing dosages to > 90 OME daily in most circumstances.^5,20 Our finding that participants taking higher dosages of opioids were not more likely to have better control over their pain supports this recommendation.

Limitations

This study has several limitations, the most significant is its small sample size because of the exploratory nature of the project. Results are based on a small pilot sample enriched to include individuals with at least moderate pain who receive opioids frequently at 1 VA CLC-PAC unit; therefore, the results might not be representative of all veterans or a more general population. Our small sample size limits power to detect small differences. Data collected should be used to inform formal power calculations before subsequent larger studies to select adequate sample size. Validation studies, including samples from the same population using different dates, which reproduce findings are an important step. Moreover, we only had data on a single dimension of pain (intensity/severity), as measured by the pain scale, which nursing staff used to make a real-time clinical decision of whether to administer an as-needed opioid. Future studies should consider using pain measures that provide multidimensional assessment (ie, severity, functional interference) and/or were developed specifically for veterans, such as the Defense and Veterans Pain Rating Scale.²¹

Our study was cross-sectional in nature and addressed a single 24-hour period of data per participant. The years of data collection (2016 and 2017) followed a decline in overall opioid prescribing that has continued, likely influenced by CDC and VA/DoD guidelines.²² It is unclear whether our observations are an accurate reflection of individuals’ response over time or whether prescribing practices in PAC have shifted.

We did not consider the type of pain being treated or explore clinicians’ reasons for prescribing opioids, therefore limiting our ability to know whether opioids were indicated. Information regarding OUD and other SUDs was limited to what was documented in the chart during the CLC-PAC unit admission. We did not have information on length of exposure to opioids. It is possible that opioid tolerance could play a role in reducing opioid responsiveness. However, simple tolerance would not be expected to explain robust correlations with psychiatric comorbidities. Also, simple tolerance would be expected to be overcome with higher opioid dosages, whereas our study demonstrates less responsiveness. These data suggests that some individuals’ pain might be poorly opioid responsive, and psychiatric factors could increase this risk. We used a novel data source in combination with chart review; to our knowledge, barcode medication administration data have not been used in this manner previously. Future work needs to validate this method, using larger sample sizes and several clinical sites. Finally, we used regression models that controlled for average pre-opioid pain rating scores, which is only 1 covariate important for examining effects. Larger studies with adequate power should control for multiple covariates known to be associated with pain and opioid response.

Conclusions

Opioid responsiveness is important clinically yet challenging to assess. This pilot study identifies a way of classifying pain as relatively opioid nonresponsive using administrative data but requires further validation before considering scaling for more general use. The possibility that a substantial percentage of residents in a CLC-PAC unit could be receiving increasing dosages of opioids without adequate benefit justifies the need for more research and underscores the need for prescribers to assess individuals frequently for ongoing benefit of opioids regardless of diagnosis or mechanism of pain.

Acknowledgments

The authors thank Andrzej Galecki, Corey Powell, and the University of Michigan Consulting for Statistics, Computing and Analytics Research Center for assistance with statistical analysis.

Testing is ongoing after more than 4,600 veterans who had received care at the Carl Vinson Veterans Affairs Medical Center in Dublin, Georgia, were alerted that they may have been exposed to HIV, hepatitis B, and hepatitis C. The exposure was due to improperly sterilized equipment. At least some of the patients have tested positive, but the facility has not indicated the number, the diseases, or whether the infections were the result of the exposure.

A mid-January internal review at the hospital found that not all steps were being followed in the procedures for sterilizing equipment between patients. Patients who had dentistry, endoscopy, urology, podiatry, optometry, or surgical procedures in 2021 may have been exposed to blood-borne pathogens.

In response, the VA sent teams from other hospitals to help, including a team from the Augusta Veterans Affairs Medical Center to reprocess all equipment and staff from VA facilities in Atlanta, South Carolina, and Alabama to provide personnel training. All staff at Carl Vinson Veterans Affairs Medical Center have since been retrained on all current guidelines.

The hospital says it’s still testing exposed veterans. Hospital spokesperson James Huckfeldt told a Macon-based newspaper, The Telegraph, that veterans with positive test results will undergo additional testing to determine whether the transmission is new or preexisting. “The findings from the additional testing will be used to accurately diagnose any impacted veterans and ensure that they receive appropriate medical treatment,” he said.

Manuel M. Davila, director of the hospital, sent letters to the patients at risk, alerting them to the exposure. “We sincerely apologize and accept responsibility for this mistake and are taking steps to prevent it from happening in the future,” Davilla wrote. “This event is unacceptable to us as well, and we want to work with you to correct the situation and ensure your safety and well-being. Because your safety is important to us and because we want to honor your trust in us, we want you to know that when concerns are raised over our processes or procedures, we take immediate steps to stop everything and make sure things are.”

Davilla reassured the veterans that “we are confident that the risk of infectious disease is very low.”

The Carl Vinson Medical Center has set up a communication center to answer questions for veterans: (478) 274-5400.

Testing is ongoing after more than 4,600 veterans who had received care at the Carl Vinson Veterans Affairs Medical Center in Dublin, Georgia, were alerted that they may have been exposed to HIV, hepatitis B, and hepatitis C. The exposure was due to improperly sterilized equipment. At least some of the patients have tested positive, but the facility has not indicated the number, the diseases, or whether the infections were the result of the exposure.

A mid-January internal review at the hospital found that not all steps were being followed in the procedures for sterilizing equipment between patients. Patients who had dentistry, endoscopy, urology, podiatry, optometry, or surgical procedures in 2021 may have been exposed to blood-borne pathogens.

In response, the VA sent teams from other hospitals to help, including a team from the Augusta Veterans Affairs Medical Center to reprocess all equipment and staff from VA facilities in Atlanta, South Carolina, and Alabama to provide personnel training. All staff at Carl Vinson Veterans Affairs Medical Center have since been retrained on all current guidelines.

The hospital says it’s still testing exposed veterans. Hospital spokesperson James Huckfeldt told a Macon-based newspaper, The Telegraph, that veterans with positive test results will undergo additional testing to determine whether the transmission is new or preexisting. “The findings from the additional testing will be used to accurately diagnose any impacted veterans and ensure that they receive appropriate medical treatment,” he said.

Manuel M. Davila, director of the hospital, sent letters to the patients at risk, alerting them to the exposure. “We sincerely apologize and accept responsibility for this mistake and are taking steps to prevent it from happening in the future,” Davilla wrote. “This event is unacceptable to us as well, and we want to work with you to correct the situation and ensure your safety and well-being. Because your safety is important to us and because we want to honor your trust in us, we want you to know that when concerns are raised over our processes or procedures, we take immediate steps to stop everything and make sure things are.”

Davilla reassured the veterans that “we are confident that the risk of infectious disease is very low.”

The Carl Vinson Medical Center has set up a communication center to answer questions for veterans: (478) 274-5400.

Testing is ongoing after more than 4,600 veterans who had received care at the Carl Vinson Veterans Affairs Medical Center in Dublin, Georgia, were alerted that they may have been exposed to HIV, hepatitis B, and hepatitis C. The exposure was due to improperly sterilized equipment. At least some of the patients have tested positive, but the facility has not indicated the number, the diseases, or whether the infections were the result of the exposure.

A mid-January internal review at the hospital found that not all steps were being followed in the procedures for sterilizing equipment between patients. Patients who had dentistry, endoscopy, urology, podiatry, optometry, or surgical procedures in 2021 may have been exposed to blood-borne pathogens.

In response, the VA sent teams from other hospitals to help, including a team from the Augusta Veterans Affairs Medical Center to reprocess all equipment and staff from VA facilities in Atlanta, South Carolina, and Alabama to provide personnel training. All staff at Carl Vinson Veterans Affairs Medical Center have since been retrained on all current guidelines.

The hospital says it’s still testing exposed veterans. Hospital spokesperson James Huckfeldt told a Macon-based newspaper, The Telegraph, that veterans with positive test results will undergo additional testing to determine whether the transmission is new or preexisting. “The findings from the additional testing will be used to accurately diagnose any impacted veterans and ensure that they receive appropriate medical treatment,” he said.

Manuel M. Davila, director of the hospital, sent letters to the patients at risk, alerting them to the exposure. “We sincerely apologize and accept responsibility for this mistake and are taking steps to prevent it from happening in the future,” Davilla wrote. “This event is unacceptable to us as well, and we want to work with you to correct the situation and ensure your safety and well-being. Because your safety is important to us and because we want to honor your trust in us, we want you to know that when concerns are raised over our processes or procedures, we take immediate steps to stop everything and make sure things are.”

Davilla reassured the veterans that “we are confident that the risk of infectious disease is very low.”

The Carl Vinson Medical Center has set up a communication center to answer questions for veterans: (478) 274-5400.

Cirrhosis is the most common cause of ascites in the United States. In patients with compensated cirrhosis, the 10-year probability of developing ascites is 47%. Developing ascites portends a poor prognosis. Fifteen percent of patients who receive this diagnosis die within 1 year, and 44% within 5 years.¹ First-line treatment of cirrhotic ascites consists of dietary sodium restriction and diuretic therapy. Refractory ascites is defined as ascites that cannot be easily mobilized despite adhering to a dietary sodium intake of ≤ 2 g daily and daily doses of spironolactone 400 mg and furosemide 160 mg.

Patients who cannot tolerate diuretics because of complications are defined as having diuretic intractable ascites. Diuretic-induced complications include hepatic encephalopathy, renal impairment, hyponatremia, and hypo- or hyperkalemia. Because these patients are either unresponsive to or intolerant of diuretics, second-line treatments, such as regular large-volume paracentesis (LVP) or the insertion of a transjugular intrahepatic portosystemic shunt (TIPS) are needed to manage their ascites. These patients also should be considered for liver transplantation unless there is a contraindication.²

Serial LVP has been shown to be safe and effective in controlling refractory ascites.³ TIPS will decrease the need for repeated LVP in patients with refractory LVP. However, given the uncertainty as to the effect of TIPS creation on survival and the increased risk of encephalopathy, the American Association for the Study of Liver Diseases (AASLD) recommends that TIPS should be used only in those patients who cannot tolerate repeated LVP.⁴ Repeated LVP also has been shown to be safe and effective in controlling malignant ascites.^5,6

LVP can be done in different health care settings. These include the emergency department (ED), interventional radiology suite, inpatient bed, or an outpatient paracentesis clinic. There have been various descriptions of outpatient paracentesis clinics. Reports from the United Kingdom have revealed that paracenteses in these outpatient clinics can be performed safely by nurse practitioners or a liver specialist nurse, that these clinics are highly rated by the patients, and are cost effective.^7-10 Gashau and colleagues describe a clinic in Great Britain run by gastroenterology (GI) fellows using an endoscopy suite.¹¹ A nurse practitioner outpatient paracentesis clinic in the US has been described as well.¹² Grabau and colleagues present a clinic run by GI endoscopy assistants (licensed practical nurses) using a dedicated paracentesis room in the endoscopy suite.¹³ Cheng and colleagues describe an outpatient paracentesis clinic in a radiology department run by a single advanced practitioner with assistance from an ultrasound technologist.¹⁴ Wang and colleagues present outpatient paracenteses in an outpatient transitional care program by a physician or an advanced practitioner supervised by a physician.¹⁵ Sehgal and colleagues describe (in abstract) the creation of a hospitalist-run paracentesis clinic.¹⁶

Traditionally, at Veterans Affairs Pittsburgh Healthcare System (VAPHS) in Pennsylvania, if a patient needed LVP, they were admitted to a medicine bed. LVP is not done in the ED, and interventional radiology cannot accommodate the number of patients requiring LVP because of their caseload. The procedure was done by an attending hospitalist or medical residents under the supervision of an attending hospitalist. To improve patient flow and decrease the number of patients using inpatients beds, we created an outpatient paracentesis clinic in 2014. Here, we present the logistics of the clinic, patient demographics, the amount of ascites removed, and the time required to remove the ascites. As part of ongoing quality assurance, we keep track of any complications and report these as well.

Methods

The setting of the outpatient paracentesis clinic is a room in the VAPHS endoscopy suite. The clinic operates 1 half-day per week with up to 3 patients receiving a paracentesis. We use the existing logistics in the endoscopy suite. There are 1 or 2 registered nurses (RNs) who assist the physician performing the paracentesis. The proceduralist is an academic hospitalist who at the time is not on service with residents. The patients are referred to the clinic by the ED, hepatology clinic, palliative care, primary care physicians, or at hospital discharge. In the clinic consult, patients are required to have at least an estimated 3 L of ascites and systolic blood pressure (SBP) ≥ 90. The patients can eat and take medications the morning of the procedure except diuretics. Patients are checked in to the endoscopy suite and a peripheral IV is placed. Blood tests, such as a complete blood count and coagulation studies, are not checked routinely since the AASLD guidelines state that routine prophylactic use of fresh frozen plasma or platelets before paracentesis is not recommended because bleeding is uncommon.³ The proceduralist can order blood work at their discretion.

After the procedure, patients are brought to the recovery area of the endoscopy suite and discharged. The patients are discharged usually within 15 to 30 minutes from arriving in the recovery area after it is assured that the SBP is within 10% of their baseline. Patient follow-up in the outpatient paracentesis clinic is determined by the proceduralist. Most patients need regularly scheduled paracenteses depending on how quickly they reaccumulate ascites. If a patient does not need a regularly scheduled paracentesis, the proceduralist ensures that the appropriate outpatient clinic visit has been scheduled or requested.

Procedure

Informed consent is obtained, and a time-out is performed before each paracentesis. The patient is attached to a cardiac monitor and pulse oximetry as per the endoscopy suite protocol. The proceduralist does a point-of-care ultrasound to find the optimal site and marks the site of puncture. The skin around the marked site is prepared with 3 chlorhexidine gluconate 2%/isopropyl alcohol 70% applicators. A fenestrated drape is used to form a sterile field. The Avanos Paracentesis Kit is routinely used for LVP at VAPHS. Local anesthesia with 1% lidocaine is used with a 25-gauge × 1-inch needle. Deeper anesthesia is obtained with 1% lidocaine, using a 22-gauge × 1.5-inch needle, injecting and aspirating while advancing the needle until ascites is aspirated.

A 15-gauge 3.3-inch Caldwell cannula with an inner needle is inserted into the peritoneal cavity and ascites is aspirated into a syringe. The inner needle is then removed, and the Caldwell cannula is left in the peritoneal cavity and tubing with a roller clamp is attached to the cannula. The tubing is then attached to a 1-L vacuum suction bottle by the RN. We use the CareFusion PleurX drainage bottle. The proceduralist maintains sterility and assures the cannula remains in place. The RN changes the drainage bottles after being filled with 1 L of ascites.

We drain as much ascites as possible until drainage stops on its own. The cannula is then removed, and pressure is held with a gauze pad. An adhesive bandage is then placed over the site. Consistent with AASLD guideline, 25 g of IV albumin 25% is infused for every 3 L of albumin removed provided > 5 L of ascites is removed.³ The albumin is infused during the procedure and not after to limit the time of the procedure. A sample of ascites is sent for cell count with differential and culture.

Results

Between March 2014 and May 2020, 506 paracenteses were performed on 82 patients. The mean age was 66.4 years, and 80 of 82 patients were male. The etiology of the ascites is presented in the Table. Twelve percent of the patients had concomitant hepatocellular carcinoma. Data on the amount of ascites removed were available for all patients, but data on the amount of time it took to do the LVP were available for 392 of 506 paracenteses. The mean volume removed was 7.9 L (range, 0.2-22.9 L), and the mean time of the procedure was 33.3 minutes. The time of the procedure was the time difference between entering and leaving the procedure room. This does not include IV placement or the recovery area time.

There were 5 episodes of postprocedure hypotension that required IV fluid or admission. In all these events, the patients had received the appropriate amount of IV albumin. Three patients required admission, and 1 patient required IV fluid postparacentesis on 2 occasions and then was discharged home. One abdominal wall hematoma occurred. Two patients with umbilical hernias developed incarceration after the paracentesis; both required surgical repair. There were 3 episodes of leakage at the paracentesis site; a skin adhesive was used in 2 cases, and sutures were applied in the other. There were no deaths.

Possible Infections

Ascitic fluid infection is a risk for patients needing paracentesis. Spontaneous bacterial peritonitis (SBP) is a bacterial infection of ascites in the absence of a focal contiguous source. The polymorphonuclear leukocyte (PMN) count in the ascites is ≥ 250 cells/mm³ in the presence of a single organism on culture. Culture-negative neutrocytic ascites (CNNA) is an ascitic fluid PMN count ≥ 250 cells/mm³ in the absence of culture growth obtained before the administration of antibiotics. Monomicrobial nonneutrocytic bacterascites (MNB) is an ascitic fluid PMN count < 250 cells/mm³ with growth of a single organism on culture.¹⁷ There was one occasion where a patient developed symptomatic CNNA 3 days after having a therapeutic paracentesis in the clinic at which time his ascites had a normal neutrophil count and a negative culture. He presented with abdominal pain and fever 3 days later, and a diagnostic paracentesis was done in the ED. He was treated as though he had SBP and did well.

Ascites cell count and culture are routinely sent in the clinic, and 1 case of asymptomatic SBP and 3 cases of asymptomatic ascitic fluid infection variants were diagnosed. The patient with SBP grew vancomycin-resistant Enterococcus faecium in his ascites. Two cases were CNNA. These patients were admitted to the hospital and treated with IV antibiotics. One case of MNB occurred that grew Escherichia coli. The patient refused to return to the hospital for IV antibiotics and was treated with a 5-day course of oral ciprofloxacin.

Discussion

We describe an academic hospitalist–run outpatient LVP clinic where large volumes of ascites are removed efficiently and safely. The only other description of a hospitalist-run paracentesis clinic was in abstract form.¹⁶ Without the clinic, the patients would have been admitted to the hospital to get an LVP. Based on VAPHS data from fiscal year 2021, the average cost per day of a nontelemetry medicine admission was $3394. Over 74 months, 506 admissions were prevented, which averages to 82 admissions prevented per year, an approximate annual cost savings of $278,308 in the last fiscal year alone.

Possible Complications

The complications we report are congruent with those reported in the literature. Runyon reported that the rate of an abdominal wall hematoma requiring blood transfusion was 0.9%, and the rate of an abdominal wall hematoma not requiring blood transfusion was also 0.9%.¹⁸ We had 1 patient who developed an abdominal wall hematoma (0.2% of paracenteses). This patient required 4 units of packed red blood cells. The incidence of ascitic fluid leakage after paracentesis has been reported to be between 0.4% and 2.4%.¹² We had 3 episodes of leakage (0.6% of paracenteses). The Z-track technique has been purported to decrease postparacentesis leakage.² This involves creating a pathway that is nonlinear when anesthetizing the soft tissues and inserting the paracentesis needle. The Z-track technique was not used in any of the paracenteses in our clinic.

Postparacentesis hypotension has been reported to be 0.4% to 1.8%.^12,14 We report 5 episodes of hypotension (0.1% of paracenteses) of which 3 patients were admitted to the hospital. Interestingly, 4 of the 5 patients were on β-blockers. Serste and colleagues reported in a crossover trial that paracentesis-induced circulatory dysfunction (PICD) decreased from 80 to 10% when propranolol was discontinued.¹⁹ PICD is characterized by reduction of effective arterial blood volume with subsequent activation of vasoconstrictor and antinatriuretic factors that can cause rapid ascites recurrence rate, development of dilutional hyponatremia, hepatorenal syndrome, and increased mortality. IV albumin is given during LVP to prevent PICD. Discontinuing unnecessary antihypertensive medications, especially β-blockers, may mitigate postparacentesis hypotension. In a study of 515 paracenteses, De Gottardi and colleagues reported a 0.2% rate of iatrogenic percutaneous infection of ascites.²⁰ We had 1 patient return 3 days after LVP with fever, abdominal pain, and neutrocytic ascites. His blood and ascites cultures were negative. The etiology of his infected ascites could have been either a spontaneously developed CNNA infection or an iatrogenic percutaneous infection of ascites.

Two cases of incarceration and strangulation of umbilical hernias postparacentesis that required emergent surgical intervention were unanticipated complications. Incarceration of an existing umbilical hernia postparacentesis is an uncommon but serious complication of LVP described in the past in numerous case reports but whose incidence is otherwise unknown.^21-26 The fluid and pressure shifts before and after LVP are likely responsible for the hernia incarceration. When ascites is present, the umbilical hernia ring is kept patent by the pressure of the ascitic fluid, and the decrease in tension after removal of ascites may lead to decreased size of the hernia ring and trapping of contents in the hernia sac.^25-27 In most reported cases, symptoms and recognition of the incarcerated hernia have occurred within 2 days of the index paracentesis procedure. Most cases were in patients who required serial paracenteses for management of ascites and had relatively regular LVPs.

In both cases, the patients had regular visits for paracentesis, and incarceration occurred 0.5 hours postprocedure, in 1 case and 6 hours in the other. Umbilical hernias are common in patients with cirrhosis, with the prevalence approaching 20%.²⁸ The management of umbilical hernias in patients with ascites is complex and optimal guideline-based management involves elective repair when ascites is adequately controlled to prevent recurrence, with consideration of TIPS at the time of repair.³ However, patients enrolled in outpatient paracentesis clinics are unlikely to have adequate ascites control to be considered optimized for an elective repair. In addition, given the number of serial procedures that they require, it is not surprising that they may be at risk for complications that are otherwise thought to be rare. Although incarceration and strangulation of umbilical hernia is thought to be a rare complication of LVP, patients should be informed of this potential complication so that they are aware to seek medical attention should they develop signs or symptoms.

Guidelines

There are no guidelines on how much ascites can be removed and how quickly the ascites can be removed during LVP. The goal of a therapeutic paracentesis is to remove as much fluid as possible, and there are no limits on the amount that can be removed safely.¹ Concerning paracentesis flow rates, Elsabaawy and colleagues showed that ascites flow rate does not correlate with PICD. They looked at 3 groups with ascites flow rates of 80 mL/min, 180 mL/min and 270 mL/min.²⁹ We had data on the time in the procedure room in 77% of our procedures. Given our average amount of ascites removed (7.9 L) and average time in the procedure room (33.3 minutes), the average flow rate from our clinic was at least 237 mL/min (although the flow rate was likely higher because the average time from needle inserted to needle removed was < 33.3 minutes). Both the mean duration of LVP and the mean volume of ascites removed in an outpatient paracentesis clinic were reported in only 1 other study. In a study of 1100 patients, Grabau and colleagues reported the mean duration, defined as the time between when the patient entered and exited the procedure room (the same time period we reported) as 97 minutes and the mean volume of ascites removed as 8.7 L.¹³

The AASLD guidelines state that patients undergoing serial outpatient LVP should be tested only for cell count and differential without sending a bacterial culture. The reason given is that false positives may exceed true positives from ascites bacterial culture results in asymptomatic patients.³ Mohan and Venkataraman reported a 0.4% rate of SBP, 1.4% rate of CNNA, and 0.7% rate of MNB in asymptomatic patients undergoing LVP in an outpatient clinic.³⁰ We had a 0.2% rate of SBP, 0.4% rate of CNNA, and 0.2% rate of MNB. Given the low rates of SBP in outpatient paracenteses clinics, we will adopt the AASLD suggestions to only send an ascites cell count and not a culture in asymptomatic patients. Noteworthy, our patient with asymptomatic SBP grew vancomycin-resistant Enterococcus faecium, which was resistant to standard SBP antibiotic therapy. However, if ascites culture was not sent, he would have been treated with antibiotics for CNNA, and if he developed symptoms, he would have had a repeat paracentesis with cell count and culture sent.

Training

In 2015, faculty at VAPHS and the University of Pittsburgh School of Medicine designed a Mastering Paracentesis for Medical Residents course based on current guidelines on the management of ascites and published procedural guides. The course is mandatory for all postgraduate year-1 internal medicine residents and begins with 2 hours of didactic and simulation-based training with an ultrasound-compatible paracentesis mannequin. In the 3 weeks following simulation-based training, residents rotate through our outpatient paracentesis clinic and perform between 1 and 3 abdominal paracentesis procedures, receiving as-needed coaching and postprocedure feedback from faculty. Since the course’s inception, more than 150 internal medicine residents have been trained in paracentesis through our clinic.

Conclusions

We present a description of a successful outpatient paracentesis clinic at our hospital run by academic hospitalists. The clinic was created to decrease the number of admissions for LVP. We were fortunate to be able to use the GI endoscopy suite and their resources as the clinic setting. To create outpatient LVP clinics at other institutions, administrative support is essential. In conclusion, we have shown that an outpatient paracentesis clinic run by academic hospitalists can safely and quickly remove large volumes of ascites.

Cirrhosis is the most common cause of ascites in the United States. In patients with compensated cirrhosis, the 10-year probability of developing ascites is 47%. Developing ascites portends a poor prognosis. Fifteen percent of patients who receive this diagnosis die within 1 year, and 44% within 5 years.¹ First-line treatment of cirrhotic ascites consists of dietary sodium restriction and diuretic therapy. Refractory ascites is defined as ascites that cannot be easily mobilized despite adhering to a dietary sodium intake of ≤ 2 g daily and daily doses of spironolactone 400 mg and furosemide 160 mg.

Patients who cannot tolerate diuretics because of complications are defined as having diuretic intractable ascites. Diuretic-induced complications include hepatic encephalopathy, renal impairment, hyponatremia, and hypo- or hyperkalemia. Because these patients are either unresponsive to or intolerant of diuretics, second-line treatments, such as regular large-volume paracentesis (LVP) or the insertion of a transjugular intrahepatic portosystemic shunt (TIPS) are needed to manage their ascites. These patients also should be considered for liver transplantation unless there is a contraindication.²

Serial LVP has been shown to be safe and effective in controlling refractory ascites.³ TIPS will decrease the need for repeated LVP in patients with refractory LVP. However, given the uncertainty as to the effect of TIPS creation on survival and the increased risk of encephalopathy, the American Association for the Study of Liver Diseases (AASLD) recommends that TIPS should be used only in those patients who cannot tolerate repeated LVP.⁴ Repeated LVP also has been shown to be safe and effective in controlling malignant ascites.^5,6

LVP can be done in different health care settings. These include the emergency department (ED), interventional radiology suite, inpatient bed, or an outpatient paracentesis clinic. There have been various descriptions of outpatient paracentesis clinics. Reports from the United Kingdom have revealed that paracenteses in these outpatient clinics can be performed safely by nurse practitioners or a liver specialist nurse, that these clinics are highly rated by the patients, and are cost effective.^7-10 Gashau and colleagues describe a clinic in Great Britain run by gastroenterology (GI) fellows using an endoscopy suite.¹¹ A nurse practitioner outpatient paracentesis clinic in the US has been described as well.¹² Grabau and colleagues present a clinic run by GI endoscopy assistants (licensed practical nurses) using a dedicated paracentesis room in the endoscopy suite.¹³ Cheng and colleagues describe an outpatient paracentesis clinic in a radiology department run by a single advanced practitioner with assistance from an ultrasound technologist.¹⁴ Wang and colleagues present outpatient paracenteses in an outpatient transitional care program by a physician or an advanced practitioner supervised by a physician.¹⁵ Sehgal and colleagues describe (in abstract) the creation of a hospitalist-run paracentesis clinic.¹⁶

Traditionally, at Veterans Affairs Pittsburgh Healthcare System (VAPHS) in Pennsylvania, if a patient needed LVP, they were admitted to a medicine bed. LVP is not done in the ED, and interventional radiology cannot accommodate the number of patients requiring LVP because of their caseload. The procedure was done by an attending hospitalist or medical residents under the supervision of an attending hospitalist. To improve patient flow and decrease the number of patients using inpatients beds, we created an outpatient paracentesis clinic in 2014. Here, we present the logistics of the clinic, patient demographics, the amount of ascites removed, and the time required to remove the ascites. As part of ongoing quality assurance, we keep track of any complications and report these as well.

Methods

The setting of the outpatient paracentesis clinic is a room in the VAPHS endoscopy suite. The clinic operates 1 half-day per week with up to 3 patients receiving a paracentesis. We use the existing logistics in the endoscopy suite. There are 1 or 2 registered nurses (RNs) who assist the physician performing the paracentesis. The proceduralist is an academic hospitalist who at the time is not on service with residents. The patients are referred to the clinic by the ED, hepatology clinic, palliative care, primary care physicians, or at hospital discharge. In the clinic consult, patients are required to have at least an estimated 3 L of ascites and systolic blood pressure (SBP) ≥ 90. The patients can eat and take medications the morning of the procedure except diuretics. Patients are checked in to the endoscopy suite and a peripheral IV is placed. Blood tests, such as a complete blood count and coagulation studies, are not checked routinely since the AASLD guidelines state that routine prophylactic use of fresh frozen plasma or platelets before paracentesis is not recommended because bleeding is uncommon.³ The proceduralist can order blood work at their discretion.

After the procedure, patients are brought to the recovery area of the endoscopy suite and discharged. The patients are discharged usually within 15 to 30 minutes from arriving in the recovery area after it is assured that the SBP is within 10% of their baseline. Patient follow-up in the outpatient paracentesis clinic is determined by the proceduralist. Most patients need regularly scheduled paracenteses depending on how quickly they reaccumulate ascites. If a patient does not need a regularly scheduled paracentesis, the proceduralist ensures that the appropriate outpatient clinic visit has been scheduled or requested.

Procedure

Informed consent is obtained, and a time-out is performed before each paracentesis. The patient is attached to a cardiac monitor and pulse oximetry as per the endoscopy suite protocol. The proceduralist does a point-of-care ultrasound to find the optimal site and marks the site of puncture. The skin around the marked site is prepared with 3 chlorhexidine gluconate 2%/isopropyl alcohol 70% applicators. A fenestrated drape is used to form a sterile field. The Avanos Paracentesis Kit is routinely used for LVP at VAPHS. Local anesthesia with 1% lidocaine is used with a 25-gauge × 1-inch needle. Deeper anesthesia is obtained with 1% lidocaine, using a 22-gauge × 1.5-inch needle, injecting and aspirating while advancing the needle until ascites is aspirated.

A 15-gauge 3.3-inch Caldwell cannula with an inner needle is inserted into the peritoneal cavity and ascites is aspirated into a syringe. The inner needle is then removed, and the Caldwell cannula is left in the peritoneal cavity and tubing with a roller clamp is attached to the cannula. The tubing is then attached to a 1-L vacuum suction bottle by the RN. We use the CareFusion PleurX drainage bottle. The proceduralist maintains sterility and assures the cannula remains in place. The RN changes the drainage bottles after being filled with 1 L of ascites.

We drain as much ascites as possible until drainage stops on its own. The cannula is then removed, and pressure is held with a gauze pad. An adhesive bandage is then placed over the site. Consistent with AASLD guideline, 25 g of IV albumin 25% is infused for every 3 L of albumin removed provided > 5 L of ascites is removed.³ The albumin is infused during the procedure and not after to limit the time of the procedure. A sample of ascites is sent for cell count with differential and culture.

Results

Between March 2014 and May 2020, 506 paracenteses were performed on 82 patients. The mean age was 66.4 years, and 80 of 82 patients were male. The etiology of the ascites is presented in the Table. Twelve percent of the patients had concomitant hepatocellular carcinoma. Data on the amount of ascites removed were available for all patients, but data on the amount of time it took to do the LVP were available for 392 of 506 paracenteses. The mean volume removed was 7.9 L (range, 0.2-22.9 L), and the mean time of the procedure was 33.3 minutes. The time of the procedure was the time difference between entering and leaving the procedure room. This does not include IV placement or the recovery area time.

There were 5 episodes of postprocedure hypotension that required IV fluid or admission. In all these events, the patients had received the appropriate amount of IV albumin. Three patients required admission, and 1 patient required IV fluid postparacentesis on 2 occasions and then was discharged home. One abdominal wall hematoma occurred. Two patients with umbilical hernias developed incarceration after the paracentesis; both required surgical repair. There were 3 episodes of leakage at the paracentesis site; a skin adhesive was used in 2 cases, and sutures were applied in the other. There were no deaths.

Possible Infections

Ascitic fluid infection is a risk for patients needing paracentesis. Spontaneous bacterial peritonitis (SBP) is a bacterial infection of ascites in the absence of a focal contiguous source. The polymorphonuclear leukocyte (PMN) count in the ascites is ≥ 250 cells/mm³ in the presence of a single organism on culture. Culture-negative neutrocytic ascites (CNNA) is an ascitic fluid PMN count ≥ 250 cells/mm³ in the absence of culture growth obtained before the administration of antibiotics. Monomicrobial nonneutrocytic bacterascites (MNB) is an ascitic fluid PMN count < 250 cells/mm³ with growth of a single organism on culture.¹⁷ There was one occasion where a patient developed symptomatic CNNA 3 days after having a therapeutic paracentesis in the clinic at which time his ascites had a normal neutrophil count and a negative culture. He presented with abdominal pain and fever 3 days later, and a diagnostic paracentesis was done in the ED. He was treated as though he had SBP and did well.

Ascites cell count and culture are routinely sent in the clinic, and 1 case of asymptomatic SBP and 3 cases of asymptomatic ascitic fluid infection variants were diagnosed. The patient with SBP grew vancomycin-resistant Enterococcus faecium in his ascites. Two cases were CNNA. These patients were admitted to the hospital and treated with IV antibiotics. One case of MNB occurred that grew Escherichia coli. The patient refused to return to the hospital for IV antibiotics and was treated with a 5-day course of oral ciprofloxacin.

Discussion

We describe an academic hospitalist–run outpatient LVP clinic where large volumes of ascites are removed efficiently and safely. The only other description of a hospitalist-run paracentesis clinic was in abstract form.¹⁶ Without the clinic, the patients would have been admitted to the hospital to get an LVP. Based on VAPHS data from fiscal year 2021, the average cost per day of a nontelemetry medicine admission was $3394. Over 74 months, 506 admissions were prevented, which averages to 82 admissions prevented per year, an approximate annual cost savings of $278,308 in the last fiscal year alone.

Possible Complications

The complications we report are congruent with those reported in the literature. Runyon reported that the rate of an abdominal wall hematoma requiring blood transfusion was 0.9%, and the rate of an abdominal wall hematoma not requiring blood transfusion was also 0.9%.¹⁸ We had 1 patient who developed an abdominal wall hematoma (0.2% of paracenteses). This patient required 4 units of packed red blood cells. The incidence of ascitic fluid leakage after paracentesis has been reported to be between 0.4% and 2.4%.¹² We had 3 episodes of leakage (0.6% of paracenteses). The Z-track technique has been purported to decrease postparacentesis leakage.² This involves creating a pathway that is nonlinear when anesthetizing the soft tissues and inserting the paracentesis needle. The Z-track technique was not used in any of the paracenteses in our clinic.

Postparacentesis hypotension has been reported to be 0.4% to 1.8%.^12,14 We report 5 episodes of hypotension (0.1% of paracenteses) of which 3 patients were admitted to the hospital. Interestingly, 4 of the 5 patients were on β-blockers. Serste and colleagues reported in a crossover trial that paracentesis-induced circulatory dysfunction (PICD) decreased from 80 to 10% when propranolol was discontinued.¹⁹ PICD is characterized by reduction of effective arterial blood volume with subsequent activation of vasoconstrictor and antinatriuretic factors that can cause rapid ascites recurrence rate, development of dilutional hyponatremia, hepatorenal syndrome, and increased mortality. IV albumin is given during LVP to prevent PICD. Discontinuing unnecessary antihypertensive medications, especially β-blockers, may mitigate postparacentesis hypotension. In a study of 515 paracenteses, De Gottardi and colleagues reported a 0.2% rate of iatrogenic percutaneous infection of ascites.²⁰ We had 1 patient return 3 days after LVP with fever, abdominal pain, and neutrocytic ascites. His blood and ascites cultures were negative. The etiology of his infected ascites could have been either a spontaneously developed CNNA infection or an iatrogenic percutaneous infection of ascites.

Two cases of incarceration and strangulation of umbilical hernias postparacentesis that required emergent surgical intervention were unanticipated complications. Incarceration of an existing umbilical hernia postparacentesis is an uncommon but serious complication of LVP described in the past in numerous case reports but whose incidence is otherwise unknown.^21-26 The fluid and pressure shifts before and after LVP are likely responsible for the hernia incarceration. When ascites is present, the umbilical hernia ring is kept patent by the pressure of the ascitic fluid, and the decrease in tension after removal of ascites may lead to decreased size of the hernia ring and trapping of contents in the hernia sac.^25-27 In most reported cases, symptoms and recognition of the incarcerated hernia have occurred within 2 days of the index paracentesis procedure. Most cases were in patients who required serial paracenteses for management of ascites and had relatively regular LVPs.

In both cases, the patients had regular visits for paracentesis, and incarceration occurred 0.5 hours postprocedure, in 1 case and 6 hours in the other. Umbilical hernias are common in patients with cirrhosis, with the prevalence approaching 20%.²⁸ The management of umbilical hernias in patients with ascites is complex and optimal guideline-based management involves elective repair when ascites is adequately controlled to prevent recurrence, with consideration of TIPS at the time of repair.³ However, patients enrolled in outpatient paracentesis clinics are unlikely to have adequate ascites control to be considered optimized for an elective repair. In addition, given the number of serial procedures that they require, it is not surprising that they may be at risk for complications that are otherwise thought to be rare. Although incarceration and strangulation of umbilical hernia is thought to be a rare complication of LVP, patients should be informed of this potential complication so that they are aware to seek medical attention should they develop signs or symptoms.

Guidelines

There are no guidelines on how much ascites can be removed and how quickly the ascites can be removed during LVP. The goal of a therapeutic paracentesis is to remove as much fluid as possible, and there are no limits on the amount that can be removed safely.¹ Concerning paracentesis flow rates, Elsabaawy and colleagues showed that ascites flow rate does not correlate with PICD. They looked at 3 groups with ascites flow rates of 80 mL/min, 180 mL/min and 270 mL/min.²⁹ We had data on the time in the procedure room in 77% of our procedures. Given our average amount of ascites removed (7.9 L) and average time in the procedure room (33.3 minutes), the average flow rate from our clinic was at least 237 mL/min (although the flow rate was likely higher because the average time from needle inserted to needle removed was < 33.3 minutes). Both the mean duration of LVP and the mean volume of ascites removed in an outpatient paracentesis clinic were reported in only 1 other study. In a study of 1100 patients, Grabau and colleagues reported the mean duration, defined as the time between when the patient entered and exited the procedure room (the same time period we reported) as 97 minutes and the mean volume of ascites removed as 8.7 L.¹³

The AASLD guidelines state that patients undergoing serial outpatient LVP should be tested only for cell count and differential without sending a bacterial culture. The reason given is that false positives may exceed true positives from ascites bacterial culture results in asymptomatic patients.³ Mohan and Venkataraman reported a 0.4% rate of SBP, 1.4% rate of CNNA, and 0.7% rate of MNB in asymptomatic patients undergoing LVP in an outpatient clinic.³⁰ We had a 0.2% rate of SBP, 0.4% rate of CNNA, and 0.2% rate of MNB. Given the low rates of SBP in outpatient paracenteses clinics, we will adopt the AASLD suggestions to only send an ascites cell count and not a culture in asymptomatic patients. Noteworthy, our patient with asymptomatic SBP grew vancomycin-resistant Enterococcus faecium, which was resistant to standard SBP antibiotic therapy. However, if ascites culture was not sent, he would have been treated with antibiotics for CNNA, and if he developed symptoms, he would have had a repeat paracentesis with cell count and culture sent.

Training

In 2015, faculty at VAPHS and the University of Pittsburgh School of Medicine designed a Mastering Paracentesis for Medical Residents course based on current guidelines on the management of ascites and published procedural guides. The course is mandatory for all postgraduate year-1 internal medicine residents and begins with 2 hours of didactic and simulation-based training with an ultrasound-compatible paracentesis mannequin. In the 3 weeks following simulation-based training, residents rotate through our outpatient paracentesis clinic and perform between 1 and 3 abdominal paracentesis procedures, receiving as-needed coaching and postprocedure feedback from faculty. Since the course’s inception, more than 150 internal medicine residents have been trained in paracentesis through our clinic.

Conclusions

We present a description of a successful outpatient paracentesis clinic at our hospital run by academic hospitalists. The clinic was created to decrease the number of admissions for LVP. We were fortunate to be able to use the GI endoscopy suite and their resources as the clinic setting. To create outpatient LVP clinics at other institutions, administrative support is essential. In conclusion, we have shown that an outpatient paracentesis clinic run by academic hospitalists can safely and quickly remove large volumes of ascites.

Cirrhosis is the most common cause of ascites in the United States. In patients with compensated cirrhosis, the 10-year probability of developing ascites is 47%. Developing ascites portends a poor prognosis. Fifteen percent of patients who receive this diagnosis die within 1 year, and 44% within 5 years.¹ First-line treatment of cirrhotic ascites consists of dietary sodium restriction and diuretic therapy. Refractory ascites is defined as ascites that cannot be easily mobilized despite adhering to a dietary sodium intake of ≤ 2 g daily and daily doses of spironolactone 400 mg and furosemide 160 mg.

Patients who cannot tolerate diuretics because of complications are defined as having diuretic intractable ascites. Diuretic-induced complications include hepatic encephalopathy, renal impairment, hyponatremia, and hypo- or hyperkalemia. Because these patients are either unresponsive to or intolerant of diuretics, second-line treatments, such as regular large-volume paracentesis (LVP) or the insertion of a transjugular intrahepatic portosystemic shunt (TIPS) are needed to manage their ascites. These patients also should be considered for liver transplantation unless there is a contraindication.²

Serial LVP has been shown to be safe and effective in controlling refractory ascites.³ TIPS will decrease the need for repeated LVP in patients with refractory LVP. However, given the uncertainty as to the effect of TIPS creation on survival and the increased risk of encephalopathy, the American Association for the Study of Liver Diseases (AASLD) recommends that TIPS should be used only in those patients who cannot tolerate repeated LVP.⁴ Repeated LVP also has been shown to be safe and effective in controlling malignant ascites.^5,6

LVP can be done in different health care settings. These include the emergency department (ED), interventional radiology suite, inpatient bed, or an outpatient paracentesis clinic. There have been various descriptions of outpatient paracentesis clinics. Reports from the United Kingdom have revealed that paracenteses in these outpatient clinics can be performed safely by nurse practitioners or a liver specialist nurse, that these clinics are highly rated by the patients, and are cost effective.^7-10 Gashau and colleagues describe a clinic in Great Britain run by gastroenterology (GI) fellows using an endoscopy suite.¹¹ A nurse practitioner outpatient paracentesis clinic in the US has been described as well.¹² Grabau and colleagues present a clinic run by GI endoscopy assistants (licensed practical nurses) using a dedicated paracentesis room in the endoscopy suite.¹³ Cheng and colleagues describe an outpatient paracentesis clinic in a radiology department run by a single advanced practitioner with assistance from an ultrasound technologist.¹⁴ Wang and colleagues present outpatient paracenteses in an outpatient transitional care program by a physician or an advanced practitioner supervised by a physician.¹⁵ Sehgal and colleagues describe (in abstract) the creation of a hospitalist-run paracentesis clinic.¹⁶

Traditionally, at Veterans Affairs Pittsburgh Healthcare System (VAPHS) in Pennsylvania, if a patient needed LVP, they were admitted to a medicine bed. LVP is not done in the ED, and interventional radiology cannot accommodate the number of patients requiring LVP because of their caseload. The procedure was done by an attending hospitalist or medical residents under the supervision of an attending hospitalist. To improve patient flow and decrease the number of patients using inpatients beds, we created an outpatient paracentesis clinic in 2014. Here, we present the logistics of the clinic, patient demographics, the amount of ascites removed, and the time required to remove the ascites. As part of ongoing quality assurance, we keep track of any complications and report these as well.

Methods

The setting of the outpatient paracentesis clinic is a room in the VAPHS endoscopy suite. The clinic operates 1 half-day per week with up to 3 patients receiving a paracentesis. We use the existing logistics in the endoscopy suite. There are 1 or 2 registered nurses (RNs) who assist the physician performing the paracentesis. The proceduralist is an academic hospitalist who at the time is not on service with residents. The patients are referred to the clinic by the ED, hepatology clinic, palliative care, primary care physicians, or at hospital discharge. In the clinic consult, patients are required to have at least an estimated 3 L of ascites and systolic blood pressure (SBP) ≥ 90. The patients can eat and take medications the morning of the procedure except diuretics. Patients are checked in to the endoscopy suite and a peripheral IV is placed. Blood tests, such as a complete blood count and coagulation studies, are not checked routinely since the AASLD guidelines state that routine prophylactic use of fresh frozen plasma or platelets before paracentesis is not recommended because bleeding is uncommon.³ The proceduralist can order blood work at their discretion.

After the procedure, patients are brought to the recovery area of the endoscopy suite and discharged. The patients are discharged usually within 15 to 30 minutes from arriving in the recovery area after it is assured that the SBP is within 10% of their baseline. Patient follow-up in the outpatient paracentesis clinic is determined by the proceduralist. Most patients need regularly scheduled paracenteses depending on how quickly they reaccumulate ascites. If a patient does not need a regularly scheduled paracentesis, the proceduralist ensures that the appropriate outpatient clinic visit has been scheduled or requested.

Procedure

Informed consent is obtained, and a time-out is performed before each paracentesis. The patient is attached to a cardiac monitor and pulse oximetry as per the endoscopy suite protocol. The proceduralist does a point-of-care ultrasound to find the optimal site and marks the site of puncture. The skin around the marked site is prepared with 3 chlorhexidine gluconate 2%/isopropyl alcohol 70% applicators. A fenestrated drape is used to form a sterile field. The Avanos Paracentesis Kit is routinely used for LVP at VAPHS. Local anesthesia with 1% lidocaine is used with a 25-gauge × 1-inch needle. Deeper anesthesia is obtained with 1% lidocaine, using a 22-gauge × 1.5-inch needle, injecting and aspirating while advancing the needle until ascites is aspirated.

A 15-gauge 3.3-inch Caldwell cannula with an inner needle is inserted into the peritoneal cavity and ascites is aspirated into a syringe. The inner needle is then removed, and the Caldwell cannula is left in the peritoneal cavity and tubing with a roller clamp is attached to the cannula. The tubing is then attached to a 1-L vacuum suction bottle by the RN. We use the CareFusion PleurX drainage bottle. The proceduralist maintains sterility and assures the cannula remains in place. The RN changes the drainage bottles after being filled with 1 L of ascites.

We drain as much ascites as possible until drainage stops on its own. The cannula is then removed, and pressure is held with a gauze pad. An adhesive bandage is then placed over the site. Consistent with AASLD guideline, 25 g of IV albumin 25% is infused for every 3 L of albumin removed provided > 5 L of ascites is removed.³ The albumin is infused during the procedure and not after to limit the time of the procedure. A sample of ascites is sent for cell count with differential and culture.

Results

Between March 2014 and May 2020, 506 paracenteses were performed on 82 patients. The mean age was 66.4 years, and 80 of 82 patients were male. The etiology of the ascites is presented in the Table. Twelve percent of the patients had concomitant hepatocellular carcinoma. Data on the amount of ascites removed were available for all patients, but data on the amount of time it took to do the LVP were available for 392 of 506 paracenteses. The mean volume removed was 7.9 L (range, 0.2-22.9 L), and the mean time of the procedure was 33.3 minutes. The time of the procedure was the time difference between entering and leaving the procedure room. This does not include IV placement or the recovery area time.

There were 5 episodes of postprocedure hypotension that required IV fluid or admission. In all these events, the patients had received the appropriate amount of IV albumin. Three patients required admission, and 1 patient required IV fluid postparacentesis on 2 occasions and then was discharged home. One abdominal wall hematoma occurred. Two patients with umbilical hernias developed incarceration after the paracentesis; both required surgical repair. There were 3 episodes of leakage at the paracentesis site; a skin adhesive was used in 2 cases, and sutures were applied in the other. There were no deaths.

Possible Infections

Ascitic fluid infection is a risk for patients needing paracentesis. Spontaneous bacterial peritonitis (SBP) is a bacterial infection of ascites in the absence of a focal contiguous source. The polymorphonuclear leukocyte (PMN) count in the ascites is ≥ 250 cells/mm³ in the presence of a single organism on culture. Culture-negative neutrocytic ascites (CNNA) is an ascitic fluid PMN count ≥ 250 cells/mm³ in the absence of culture growth obtained before the administration of antibiotics. Monomicrobial nonneutrocytic bacterascites (MNB) is an ascitic fluid PMN count < 250 cells/mm³ with growth of a single organism on culture.¹⁷ There was one occasion where a patient developed symptomatic CNNA 3 days after having a therapeutic paracentesis in the clinic at which time his ascites had a normal neutrophil count and a negative culture. He presented with abdominal pain and fever 3 days later, and a diagnostic paracentesis was done in the ED. He was treated as though he had SBP and did well.

Ascites cell count and culture are routinely sent in the clinic, and 1 case of asymptomatic SBP and 3 cases of asymptomatic ascitic fluid infection variants were diagnosed. The patient with SBP grew vancomycin-resistant Enterococcus faecium in his ascites. Two cases were CNNA. These patients were admitted to the hospital and treated with IV antibiotics. One case of MNB occurred that grew Escherichia coli. The patient refused to return to the hospital for IV antibiotics and was treated with a 5-day course of oral ciprofloxacin.

Discussion

We describe an academic hospitalist–run outpatient LVP clinic where large volumes of ascites are removed efficiently and safely. The only other description of a hospitalist-run paracentesis clinic was in abstract form.¹⁶ Without the clinic, the patients would have been admitted to the hospital to get an LVP. Based on VAPHS data from fiscal year 2021, the average cost per day of a nontelemetry medicine admission was $3394. Over 74 months, 506 admissions were prevented, which averages to 82 admissions prevented per year, an approximate annual cost savings of $278,308 in the last fiscal year alone.

Possible Complications

The complications we report are congruent with those reported in the literature. Runyon reported that the rate of an abdominal wall hematoma requiring blood transfusion was 0.9%, and the rate of an abdominal wall hematoma not requiring blood transfusion was also 0.9%.¹⁸ We had 1 patient who developed an abdominal wall hematoma (0.2% of paracenteses). This patient required 4 units of packed red blood cells. The incidence of ascitic fluid leakage after paracentesis has been reported to be between 0.4% and 2.4%.¹² We had 3 episodes of leakage (0.6% of paracenteses). The Z-track technique has been purported to decrease postparacentesis leakage.² This involves creating a pathway that is nonlinear when anesthetizing the soft tissues and inserting the paracentesis needle. The Z-track technique was not used in any of the paracenteses in our clinic.

Postparacentesis hypotension has been reported to be 0.4% to 1.8%.^12,14 We report 5 episodes of hypotension (0.1% of paracenteses) of which 3 patients were admitted to the hospital. Interestingly, 4 of the 5 patients were on β-blockers. Serste and colleagues reported in a crossover trial that paracentesis-induced circulatory dysfunction (PICD) decreased from 80 to 10% when propranolol was discontinued.¹⁹ PICD is characterized by reduction of effective arterial blood volume with subsequent activation of vasoconstrictor and antinatriuretic factors that can cause rapid ascites recurrence rate, development of dilutional hyponatremia, hepatorenal syndrome, and increased mortality. IV albumin is given during LVP to prevent PICD. Discontinuing unnecessary antihypertensive medications, especially β-blockers, may mitigate postparacentesis hypotension. In a study of 515 paracenteses, De Gottardi and colleagues reported a 0.2% rate of iatrogenic percutaneous infection of ascites.²⁰ We had 1 patient return 3 days after LVP with fever, abdominal pain, and neutrocytic ascites. His blood and ascites cultures were negative. The etiology of his infected ascites could have been either a spontaneously developed CNNA infection or an iatrogenic percutaneous infection of ascites.

Two cases of incarceration and strangulation of umbilical hernias postparacentesis that required emergent surgical intervention were unanticipated complications. Incarceration of an existing umbilical hernia postparacentesis is an uncommon but serious complication of LVP described in the past in numerous case reports but whose incidence is otherwise unknown.^21-26 The fluid and pressure shifts before and after LVP are likely responsible for the hernia incarceration. When ascites is present, the umbilical hernia ring is kept patent by the pressure of the ascitic fluid, and the decrease in tension after removal of ascites may lead to decreased size of the hernia ring and trapping of contents in the hernia sac.^25-27 In most reported cases, symptoms and recognition of the incarcerated hernia have occurred within 2 days of the index paracentesis procedure. Most cases were in patients who required serial paracenteses for management of ascites and had relatively regular LVPs.

In both cases, the patients had regular visits for paracentesis, and incarceration occurred 0.5 hours postprocedure, in 1 case and 6 hours in the other. Umbilical hernias are common in patients with cirrhosis, with the prevalence approaching 20%.²⁸ The management of umbilical hernias in patients with ascites is complex and optimal guideline-based management involves elective repair when ascites is adequately controlled to prevent recurrence, with consideration of TIPS at the time of repair.³ However, patients enrolled in outpatient paracentesis clinics are unlikely to have adequate ascites control to be considered optimized for an elective repair. In addition, given the number of serial procedures that they require, it is not surprising that they may be at risk for complications that are otherwise thought to be rare. Although incarceration and strangulation of umbilical hernia is thought to be a rare complication of LVP, patients should be informed of this potential complication so that they are aware to seek medical attention should they develop signs or symptoms.

Guidelines

There are no guidelines on how much ascites can be removed and how quickly the ascites can be removed during LVP. The goal of a therapeutic paracentesis is to remove as much fluid as possible, and there are no limits on the amount that can be removed safely.¹ Concerning paracentesis flow rates, Elsabaawy and colleagues showed that ascites flow rate does not correlate with PICD. They looked at 3 groups with ascites flow rates of 80 mL/min, 180 mL/min and 270 mL/min.²⁹ We had data on the time in the procedure room in 77% of our procedures. Given our average amount of ascites removed (7.9 L) and average time in the procedure room (33.3 minutes), the average flow rate from our clinic was at least 237 mL/min (although the flow rate was likely higher because the average time from needle inserted to needle removed was < 33.3 minutes). Both the mean duration of LVP and the mean volume of ascites removed in an outpatient paracentesis clinic were reported in only 1 other study. In a study of 1100 patients, Grabau and colleagues reported the mean duration, defined as the time between when the patient entered and exited the procedure room (the same time period we reported) as 97 minutes and the mean volume of ascites removed as 8.7 L.¹³

The AASLD guidelines state that patients undergoing serial outpatient LVP should be tested only for cell count and differential without sending a bacterial culture. The reason given is that false positives may exceed true positives from ascites bacterial culture results in asymptomatic patients.³ Mohan and Venkataraman reported a 0.4% rate of SBP, 1.4% rate of CNNA, and 0.7% rate of MNB in asymptomatic patients undergoing LVP in an outpatient clinic.³⁰ We had a 0.2% rate of SBP, 0.4% rate of CNNA, and 0.2% rate of MNB. Given the low rates of SBP in outpatient paracenteses clinics, we will adopt the AASLD suggestions to only send an ascites cell count and not a culture in asymptomatic patients. Noteworthy, our patient with asymptomatic SBP grew vancomycin-resistant Enterococcus faecium, which was resistant to standard SBP antibiotic therapy. However, if ascites culture was not sent, he would have been treated with antibiotics for CNNA, and if he developed symptoms, he would have had a repeat paracentesis with cell count and culture sent.

Training

In 2015, faculty at VAPHS and the University of Pittsburgh School of Medicine designed a Mastering Paracentesis for Medical Residents course based on current guidelines on the management of ascites and published procedural guides. The course is mandatory for all postgraduate year-1 internal medicine residents and begins with 2 hours of didactic and simulation-based training with an ultrasound-compatible paracentesis mannequin. In the 3 weeks following simulation-based training, residents rotate through our outpatient paracentesis clinic and perform between 1 and 3 abdominal paracentesis procedures, receiving as-needed coaching and postprocedure feedback from faculty. Since the course’s inception, more than 150 internal medicine residents have been trained in paracentesis through our clinic.

Conclusions

We present a description of a successful outpatient paracentesis clinic at our hospital run by academic hospitalists. The clinic was created to decrease the number of admissions for LVP. We were fortunate to be able to use the GI endoscopy suite and their resources as the clinic setting. To create outpatient LVP clinics at other institutions, administrative support is essential. In conclusion, we have shown that an outpatient paracentesis clinic run by academic hospitalists can safely and quickly remove large volumes of ascites.

Automated urine cultures (UCs) following urinalysis (UA) are often used in emergency departments (EDs) to identify urinary tract infections (UTIs). The fast-paced environment of the ED makes this method of proactive collection and facilitation of UC favorable. However, results are often reported as no organism growth or the growth of clinically insignificant organisms, leading to the overdetection and overtreatment of asymptomatic bacteriuria (ASB).^1-3 An estimated 30 to 60% of patients with ASB receive unwarranted antibiotic treatment, which is associated with an increased risk of developing Clostridioides difficile infection and contributes to the development of antimicrobial resistance.^4-10 The costs associated with UC are an important consideration given the use of resources, the time and effort required to collect and process large numbers of negative cultures, and further efforts devoted to the follow-up of ED culture results.

Changes in traditional testing involving testing of both a UA and UC to reflex testing where urine specimens undergo culture only if they meet certain criteria have been described.^11-14 This change in traditional testing aims to reduce the number of potentially unnecessary cultures performed without compromising clinical care. Leukocyte quantity in the UA has been shown to be a reliable predictor of true infection.^11,15 Fok and colleagues demonstrated that reflex urine testing in ambulatory male urology patients in which cultures were done on only urine specimens with > 5 white blood cells per high-power field (WBC/HPF) would have missed only 7% of positive UCs, while avoiding 69% of cultures.¹¹

At the Edward Hines, Jr Veterans Affairs Hospital (Hines VA), inappropriate UC ordering and treatment for ASB has been identified as an area needing improvement. An evaluation was conducted at the facility to determine the population of inpatient veterans with a positive UC who were appropriately managed. Of the 113 study patients with a positive UC included in this review, 77 (68%) had a diagnosis of ASB, with > 80% of patients with ASB (and no other suspected infections) receiving antimicrobial therapy.⁸ A subsequent evaluation was conducted at the Hines VA ED to evaluate UTI treatment and follow-up. Of the 173 ED patients included, 23% received antibiotic therapy for an ASB and 60% had a UA and UC collected but did not report symptoms.⁹ Finally, a review by the Hines VA laboratory showed that in May 2017, of 359 UCs sent from various locations of the hospital, 38% were obtained in the setting of a negative UA.

A multidisciplinary group with representation from primary care, infectious diseases, pharmacy, nursing, laboratory, and informatics was created with a goal to improve the workup and management of UTIs. In addition to periodic education for the clinicians regarding appropriate use and interpretation of UA and UC along with judicious use of antimicrobials especially in the setting of ASB, a UA to reflex culture process change was implemented. This allowed for automatic cancellation of a UC in the setting of a negative UA, which was designed to help facilitate appropriate UC ordering.

Methods

The primary objective of this study was to compare the frequency of inappropriate UC use and inappropriate antibiotic prescribing pre- and postimplementation of this UA to reflex culture process change. An inappropriate UC was defined as a UC ordered despite a negative UA in asymptomatic patients. Inappropriate antibiotic prescribing was defined as treatment of patients with ASB. The secondary objective evaluated postintervention data to assess the frequency of outpatient, ED, and hospital visits for UTI-related symptoms in the group of patients that had a UC cancelled as a result of the new process change (within a 7-day period of the initial UA) to determine whether patients with true infections were missed due to the process change.

Study Design and Setting

This pre-post quality improvement (QI) study analyzed the UC-ordering practices for UTIs sent from the ED at the Hines VA. This VA is a 483-bed tertiary care hospital in Chicago, Illinois, and serves > 57,000 veterans and about 23,000 ED visits annually. This study was approved by the Edward Hines, Jr VA Institutional Review Board as a quality assurance/QI proposal prior to data collection.

Patient Selection

All patients who received a UA with or without a UC sent from the ED between October 17, 2017 and January 17, 2018 were identified by the microbiology laboratory and a list was generated. Postintervention data were compared with data from a previous analysis performed at the Hines VA in 2015 (baseline data), which found that UCs were collected frequently despite negative UA, and often resulted in the prescribing of unnecessary antibiotics.⁹

When comparing postintervention data with preintervention data for the primary study objective, the same exclusion criteria from the 2015 study were applied to the present study, which excluded ED patients who were admitted for inpatient care, concurrent antibiotic therapy for a non-UTI indication, duplicate cultures, and use of chronic bladder management devices. All patients identified as receiving a UA during the specified postintervention study period were included for evaluation of the secondary study objective.

Interventions

After physician education, an ED process change was implemented on October 3, 2017. This process change involved the creation of new order sets in the EHR that allowed clinicians to order a UA only, a UA with culture that would be cancelled by laboratory personnel if the UA did not result in > 5 WBC/HPF, and a UA with culture designated as do not cancel, where the UC was processed regardless of the UA results. The scenarios in which the latter option was considered appropriate were listed on the ordering screen and included pregnancy, a genitourinary procedure with necessary preoperative culture, and neutropenia.

Measurements

Postimplementation, all UAs were reviewed and grouped as follows: (1) positive UA with subsequent UC; (2) negative UA, culture cancelled; (3) only UA ordered (no culture); or (4) do not cancel UC ordered. Of the UAs that were analyzed, the following data were collected: demographics, comorbidities, concurrent medications for benign prostatic hyperplasia (BPH) and/or overactive bladder (OAB), documented allergies/adverse drug reactions to antibiotics, date of ED visit, documented UTI signs/symptoms (defined as frequency, urgency, dysuria, fever, suprapubic pain, or altered mental status in patients unable to verbalize urinary symptoms), UC results and susceptibilities, number of UCs repeated within 7 days after initial UA, requirement of antibiotic for UTI within 7 days of initial UA, antibiotic prescribed, duration of antibiotic therapy, and outpatient visits, ED visits, or need for hospital admission within 7 days of the initial UA for UTI-related symptoms. Other relevant UA and UC data that could not be obtained from the EHR were collected by generating a report using the Veterans Information Systems and Technology Architecture (VistA).

Analysis

Statistical analysis was performed using SAS v9.4. Independent t tests and Fisher exact tests were used to describe difference pre- and postintervention. Statistical significance was considered for P < .05. Based on results from the previous study conducted at this facility in addition to a literature review, it was determined that 92 patients in each group (pre- and postintervention) would be necessary to detect a 15% increase in percentage of patients appropriately treated for a UTI.

Results

There were 684 UAs evaluated from ED visits, 429 preintervention and 255 postintervention. The 255 patients were evaluated for the secondary objective of the study. Of the 255 patients with UAs identified postintervention, 150 were excluded based on the predefined exclusion criteria, and the remaining 105 were compared with the 173 patients from the preintervention group and were included in the analysis for the primary objective (Figure 1).

Patients in the postintervention group were younger than those in the preintervention group (P < .02): otherwise the groups were similar (Table 1). Inappropriate antibiotics for ASB decreased from 10.2% preintervention to 1.9% postintervention (odds ratio, 0.17; P = .01) (Table 2). UC processing despite a negative UA significantly decreased from 100% preintervention to 38.6% postintervention (P < .001) (Table 3). In patients with a negative UA, antibiotic prescribing decreased by 25.3% postintervention, but this difference was not statistically significant.

Discussion

A simple process change at the Hines VA resulted in benefits related to antimicrobial stewardship without conferring adverse outcomes on patient safety. Both UC processing despite a negative UA and inappropriate antibiotic prescribing for ASB were reduced significantly postintervention. This process change was piloted in the ED where UCs are often included as part of the initial diagnostic testing in patients who may not report UTI-related symptoms but for whom a UC is often bundled with other infectious workup, depending on the patient presentation.

Reflex testing of urine specimens has been described in the literature, both in an exploratory nature where impact of a reflex UC cancellation protocol based on certain UA criteria is measured by percent reduction of UCs processed as well as results of such interventions implemented into clinical practice.^11-13 A retrospective study performed at the University of North Carolina Medical Center evaluated patients who presented to the ED during a 6-month period and had both an automated UA and UC collected. UC processing was restricted to UA that was positive for nitrites, leukocyte esterase, bacteria, or > 10 WBC/HPF. Use of this reflex culture cancellation protocol could have eliminated 604 of the 1546 (39.1%) cultures processed. However, 11 of the 314 (3.5%) positive cultures could have been missed.¹³ This same protocol was externally validated at another large academic ED setting, where similar results were found.¹⁴

Limitations

This single-center, pre-post QI study was not without limitations. Manual chart reviews were required, and accuracy of information was dependent on clinician documentation and assessment of UTI-related symptoms. The population studied was predominately older males; thus, results may not be applicable to females or young adults. Additionally, recognition of a negative UA and subsequent cancellation of the UC was dependent on laboratory personnel. As noted in the patient group with a positive UA, some of these UAs were negative and may have been overlooked; therefore, subsequent UCs were inappropriately processed. However, this occurred infrequently and confirmed the low probability of true UTI in the setting of a negative UA. Follow-up for UTI-related symptoms may not have been captured if a patient had presented to an outside facility. Last, definitions of a positive UA differed slightly between the pre- and postintervention groups. The preintervention study defined a positive UA as a WBC count > 5 WBC/HPF and positive leukocyte esterase, whereas the present study defined a positive UA with a WBC count > 5. This may have resulted in an overestimation of positive UA in the postintervention group.

Conclusions

Better selective use of UC testing may improve stewardship resources and reduce costs impacting both ED and clinical laboratories. Furthermore, benefits can include a reduction in the use of time and resources required to collect samples for culture, use of test supplies, the time and effort required to process the large number of negative cultures, and resources devoted to the follow-up of these ED culture results. The described UA to reflex culture process change demonstrated a significant reduction in the processing of inappropriate UC and unnecessary antibiotics for ASB. There were no missed UTIs or other adverse patient outcomes noted. This process change has been implemented in all departments at the Hines VA and additional data will be collected to ensure consistent outcomes.

Automated urine cultures (UCs) following urinalysis (UA) are often used in emergency departments (EDs) to identify urinary tract infections (UTIs). The fast-paced environment of the ED makes this method of proactive collection and facilitation of UC favorable. However, results are often reported as no organism growth or the growth of clinically insignificant organisms, leading to the overdetection and overtreatment of asymptomatic bacteriuria (ASB).^1-3 An estimated 30 to 60% of patients with ASB receive unwarranted antibiotic treatment, which is associated with an increased risk of developing Clostridioides difficile infection and contributes to the development of antimicrobial resistance.^4-10 The costs associated with UC are an important consideration given the use of resources, the time and effort required to collect and process large numbers of negative cultures, and further efforts devoted to the follow-up of ED culture results.

Changes in traditional testing involving testing of both a UA and UC to reflex testing where urine specimens undergo culture only if they meet certain criteria have been described.^11-14 This change in traditional testing aims to reduce the number of potentially unnecessary cultures performed without compromising clinical care. Leukocyte quantity in the UA has been shown to be a reliable predictor of true infection.^11,15 Fok and colleagues demonstrated that reflex urine testing in ambulatory male urology patients in which cultures were done on only urine specimens with > 5 white blood cells per high-power field (WBC/HPF) would have missed only 7% of positive UCs, while avoiding 69% of cultures.¹¹

At the Edward Hines, Jr Veterans Affairs Hospital (Hines VA), inappropriate UC ordering and treatment for ASB has been identified as an area needing improvement. An evaluation was conducted at the facility to determine the population of inpatient veterans with a positive UC who were appropriately managed. Of the 113 study patients with a positive UC included in this review, 77 (68%) had a diagnosis of ASB, with > 80% of patients with ASB (and no other suspected infections) receiving antimicrobial therapy.⁸ A subsequent evaluation was conducted at the Hines VA ED to evaluate UTI treatment and follow-up. Of the 173 ED patients included, 23% received antibiotic therapy for an ASB and 60% had a UA and UC collected but did not report symptoms.⁹ Finally, a review by the Hines VA laboratory showed that in May 2017, of 359 UCs sent from various locations of the hospital, 38% were obtained in the setting of a negative UA.

A multidisciplinary group with representation from primary care, infectious diseases, pharmacy, nursing, laboratory, and informatics was created with a goal to improve the workup and management of UTIs. In addition to periodic education for the clinicians regarding appropriate use and interpretation of UA and UC along with judicious use of antimicrobials especially in the setting of ASB, a UA to reflex culture process change was implemented. This allowed for automatic cancellation of a UC in the setting of a negative UA, which was designed to help facilitate appropriate UC ordering.

Methods

The primary objective of this study was to compare the frequency of inappropriate UC use and inappropriate antibiotic prescribing pre- and postimplementation of this UA to reflex culture process change. An inappropriate UC was defined as a UC ordered despite a negative UA in asymptomatic patients. Inappropriate antibiotic prescribing was defined as treatment of patients with ASB. The secondary objective evaluated postintervention data to assess the frequency of outpatient, ED, and hospital visits for UTI-related symptoms in the group of patients that had a UC cancelled as a result of the new process change (within a 7-day period of the initial UA) to determine whether patients with true infections were missed due to the process change.

Study Design and Setting

This pre-post quality improvement (QI) study analyzed the UC-ordering practices for UTIs sent from the ED at the Hines VA. This VA is a 483-bed tertiary care hospital in Chicago, Illinois, and serves > 57,000 veterans and about 23,000 ED visits annually. This study was approved by the Edward Hines, Jr VA Institutional Review Board as a quality assurance/QI proposal prior to data collection.

Patient Selection

All patients who received a UA with or without a UC sent from the ED between October 17, 2017 and January 17, 2018 were identified by the microbiology laboratory and a list was generated. Postintervention data were compared with data from a previous analysis performed at the Hines VA in 2015 (baseline data), which found that UCs were collected frequently despite negative UA, and often resulted in the prescribing of unnecessary antibiotics.⁹

When comparing postintervention data with preintervention data for the primary study objective, the same exclusion criteria from the 2015 study were applied to the present study, which excluded ED patients who were admitted for inpatient care, concurrent antibiotic therapy for a non-UTI indication, duplicate cultures, and use of chronic bladder management devices. All patients identified as receiving a UA during the specified postintervention study period were included for evaluation of the secondary study objective.

Interventions

After physician education, an ED process change was implemented on October 3, 2017. This process change involved the creation of new order sets in the EHR that allowed clinicians to order a UA only, a UA with culture that would be cancelled by laboratory personnel if the UA did not result in > 5 WBC/HPF, and a UA with culture designated as do not cancel, where the UC was processed regardless of the UA results. The scenarios in which the latter option was considered appropriate were listed on the ordering screen and included pregnancy, a genitourinary procedure with necessary preoperative culture, and neutropenia.

Measurements

Postimplementation, all UAs were reviewed and grouped as follows: (1) positive UA with subsequent UC; (2) negative UA, culture cancelled; (3) only UA ordered (no culture); or (4) do not cancel UC ordered. Of the UAs that were analyzed, the following data were collected: demographics, comorbidities, concurrent medications for benign prostatic hyperplasia (BPH) and/or overactive bladder (OAB), documented allergies/adverse drug reactions to antibiotics, date of ED visit, documented UTI signs/symptoms (defined as frequency, urgency, dysuria, fever, suprapubic pain, or altered mental status in patients unable to verbalize urinary symptoms), UC results and susceptibilities, number of UCs repeated within 7 days after initial UA, requirement of antibiotic for UTI within 7 days of initial UA, antibiotic prescribed, duration of antibiotic therapy, and outpatient visits, ED visits, or need for hospital admission within 7 days of the initial UA for UTI-related symptoms. Other relevant UA and UC data that could not be obtained from the EHR were collected by generating a report using the Veterans Information Systems and Technology Architecture (VistA).

Analysis

Statistical analysis was performed using SAS v9.4. Independent t tests and Fisher exact tests were used to describe difference pre- and postintervention. Statistical significance was considered for P < .05. Based on results from the previous study conducted at this facility in addition to a literature review, it was determined that 92 patients in each group (pre- and postintervention) would be necessary to detect a 15% increase in percentage of patients appropriately treated for a UTI.

Results

There were 684 UAs evaluated from ED visits, 429 preintervention and 255 postintervention. The 255 patients were evaluated for the secondary objective of the study. Of the 255 patients with UAs identified postintervention, 150 were excluded based on the predefined exclusion criteria, and the remaining 105 were compared with the 173 patients from the preintervention group and were included in the analysis for the primary objective (Figure 1).

Patients in the postintervention group were younger than those in the preintervention group (P < .02): otherwise the groups were similar (Table 1). Inappropriate antibiotics for ASB decreased from 10.2% preintervention to 1.9% postintervention (odds ratio, 0.17; P = .01) (Table 2). UC processing despite a negative UA significantly decreased from 100% preintervention to 38.6% postintervention (P < .001) (Table 3). In patients with a negative UA, antibiotic prescribing decreased by 25.3% postintervention, but this difference was not statistically significant.

Discussion

A simple process change at the Hines VA resulted in benefits related to antimicrobial stewardship without conferring adverse outcomes on patient safety. Both UC processing despite a negative UA and inappropriate antibiotic prescribing for ASB were reduced significantly postintervention. This process change was piloted in the ED where UCs are often included as part of the initial diagnostic testing in patients who may not report UTI-related symptoms but for whom a UC is often bundled with other infectious workup, depending on the patient presentation.

Reflex testing of urine specimens has been described in the literature, both in an exploratory nature where impact of a reflex UC cancellation protocol based on certain UA criteria is measured by percent reduction of UCs processed as well as results of such interventions implemented into clinical practice.^11-13 A retrospective study performed at the University of North Carolina Medical Center evaluated patients who presented to the ED during a 6-month period and had both an automated UA and UC collected. UC processing was restricted to UA that was positive for nitrites, leukocyte esterase, bacteria, or > 10 WBC/HPF. Use of this reflex culture cancellation protocol could have eliminated 604 of the 1546 (39.1%) cultures processed. However, 11 of the 314 (3.5%) positive cultures could have been missed.¹³ This same protocol was externally validated at another large academic ED setting, where similar results were found.¹⁴

Limitations

This single-center, pre-post QI study was not without limitations. Manual chart reviews were required, and accuracy of information was dependent on clinician documentation and assessment of UTI-related symptoms. The population studied was predominately older males; thus, results may not be applicable to females or young adults. Additionally, recognition of a negative UA and subsequent cancellation of the UC was dependent on laboratory personnel. As noted in the patient group with a positive UA, some of these UAs were negative and may have been overlooked; therefore, subsequent UCs were inappropriately processed. However, this occurred infrequently and confirmed the low probability of true UTI in the setting of a negative UA. Follow-up for UTI-related symptoms may not have been captured if a patient had presented to an outside facility. Last, definitions of a positive UA differed slightly between the pre- and postintervention groups. The preintervention study defined a positive UA as a WBC count > 5 WBC/HPF and positive leukocyte esterase, whereas the present study defined a positive UA with a WBC count > 5. This may have resulted in an overestimation of positive UA in the postintervention group.

Conclusions

Better selective use of UC testing may improve stewardship resources and reduce costs impacting both ED and clinical laboratories. Furthermore, benefits can include a reduction in the use of time and resources required to collect samples for culture, use of test supplies, the time and effort required to process the large number of negative cultures, and resources devoted to the follow-up of these ED culture results. The described UA to reflex culture process change demonstrated a significant reduction in the processing of inappropriate UC and unnecessary antibiotics for ASB. There were no missed UTIs or other adverse patient outcomes noted. This process change has been implemented in all departments at the Hines VA and additional data will be collected to ensure consistent outcomes.

Automated urine cultures (UCs) following urinalysis (UA) are often used in emergency departments (EDs) to identify urinary tract infections (UTIs). The fast-paced environment of the ED makes this method of proactive collection and facilitation of UC favorable. However, results are often reported as no organism growth or the growth of clinically insignificant organisms, leading to the overdetection and overtreatment of asymptomatic bacteriuria (ASB).^1-3 An estimated 30 to 60% of patients with ASB receive unwarranted antibiotic treatment, which is associated with an increased risk of developing Clostridioides difficile infection and contributes to the development of antimicrobial resistance.^4-10 The costs associated with UC are an important consideration given the use of resources, the time and effort required to collect and process large numbers of negative cultures, and further efforts devoted to the follow-up of ED culture results.

Changes in traditional testing involving testing of both a UA and UC to reflex testing where urine specimens undergo culture only if they meet certain criteria have been described.^11-14 This change in traditional testing aims to reduce the number of potentially unnecessary cultures performed without compromising clinical care. Leukocyte quantity in the UA has been shown to be a reliable predictor of true infection.^11,15 Fok and colleagues demonstrated that reflex urine testing in ambulatory male urology patients in which cultures were done on only urine specimens with > 5 white blood cells per high-power field (WBC/HPF) would have missed only 7% of positive UCs, while avoiding 69% of cultures.¹¹

At the Edward Hines, Jr Veterans Affairs Hospital (Hines VA), inappropriate UC ordering and treatment for ASB has been identified as an area needing improvement. An evaluation was conducted at the facility to determine the population of inpatient veterans with a positive UC who were appropriately managed. Of the 113 study patients with a positive UC included in this review, 77 (68%) had a diagnosis of ASB, with > 80% of patients with ASB (and no other suspected infections) receiving antimicrobial therapy.⁸ A subsequent evaluation was conducted at the Hines VA ED to evaluate UTI treatment and follow-up. Of the 173 ED patients included, 23% received antibiotic therapy for an ASB and 60% had a UA and UC collected but did not report symptoms.⁹ Finally, a review by the Hines VA laboratory showed that in May 2017, of 359 UCs sent from various locations of the hospital, 38% were obtained in the setting of a negative UA.

A multidisciplinary group with representation from primary care, infectious diseases, pharmacy, nursing, laboratory, and informatics was created with a goal to improve the workup and management of UTIs. In addition to periodic education for the clinicians regarding appropriate use and interpretation of UA and UC along with judicious use of antimicrobials especially in the setting of ASB, a UA to reflex culture process change was implemented. This allowed for automatic cancellation of a UC in the setting of a negative UA, which was designed to help facilitate appropriate UC ordering.

Methods

The primary objective of this study was to compare the frequency of inappropriate UC use and inappropriate antibiotic prescribing pre- and postimplementation of this UA to reflex culture process change. An inappropriate UC was defined as a UC ordered despite a negative UA in asymptomatic patients. Inappropriate antibiotic prescribing was defined as treatment of patients with ASB. The secondary objective evaluated postintervention data to assess the frequency of outpatient, ED, and hospital visits for UTI-related symptoms in the group of patients that had a UC cancelled as a result of the new process change (within a 7-day period of the initial UA) to determine whether patients with true infections were missed due to the process change.

Study Design and Setting

This pre-post quality improvement (QI) study analyzed the UC-ordering practices for UTIs sent from the ED at the Hines VA. This VA is a 483-bed tertiary care hospital in Chicago, Illinois, and serves > 57,000 veterans and about 23,000 ED visits annually. This study was approved by the Edward Hines, Jr VA Institutional Review Board as a quality assurance/QI proposal prior to data collection.

Patient Selection

All patients who received a UA with or without a UC sent from the ED between October 17, 2017 and January 17, 2018 were identified by the microbiology laboratory and a list was generated. Postintervention data were compared with data from a previous analysis performed at the Hines VA in 2015 (baseline data), which found that UCs were collected frequently despite negative UA, and often resulted in the prescribing of unnecessary antibiotics.⁹

When comparing postintervention data with preintervention data for the primary study objective, the same exclusion criteria from the 2015 study were applied to the present study, which excluded ED patients who were admitted for inpatient care, concurrent antibiotic therapy for a non-UTI indication, duplicate cultures, and use of chronic bladder management devices. All patients identified as receiving a UA during the specified postintervention study period were included for evaluation of the secondary study objective.

Interventions

After physician education, an ED process change was implemented on October 3, 2017. This process change involved the creation of new order sets in the EHR that allowed clinicians to order a UA only, a UA with culture that would be cancelled by laboratory personnel if the UA did not result in > 5 WBC/HPF, and a UA with culture designated as do not cancel, where the UC was processed regardless of the UA results. The scenarios in which the latter option was considered appropriate were listed on the ordering screen and included pregnancy, a genitourinary procedure with necessary preoperative culture, and neutropenia.

Measurements

Postimplementation, all UAs were reviewed and grouped as follows: (1) positive UA with subsequent UC; (2) negative UA, culture cancelled; (3) only UA ordered (no culture); or (4) do not cancel UC ordered. Of the UAs that were analyzed, the following data were collected: demographics, comorbidities, concurrent medications for benign prostatic hyperplasia (BPH) and/or overactive bladder (OAB), documented allergies/adverse drug reactions to antibiotics, date of ED visit, documented UTI signs/symptoms (defined as frequency, urgency, dysuria, fever, suprapubic pain, or altered mental status in patients unable to verbalize urinary symptoms), UC results and susceptibilities, number of UCs repeated within 7 days after initial UA, requirement of antibiotic for UTI within 7 days of initial UA, antibiotic prescribed, duration of antibiotic therapy, and outpatient visits, ED visits, or need for hospital admission within 7 days of the initial UA for UTI-related symptoms. Other relevant UA and UC data that could not be obtained from the EHR were collected by generating a report using the Veterans Information Systems and Technology Architecture (VistA).

Analysis

Statistical analysis was performed using SAS v9.4. Independent t tests and Fisher exact tests were used to describe difference pre- and postintervention. Statistical significance was considered for P < .05. Based on results from the previous study conducted at this facility in addition to a literature review, it was determined that 92 patients in each group (pre- and postintervention) would be necessary to detect a 15% increase in percentage of patients appropriately treated for a UTI.

Results

There were 684 UAs evaluated from ED visits, 429 preintervention and 255 postintervention. The 255 patients were evaluated for the secondary objective of the study. Of the 255 patients with UAs identified postintervention, 150 were excluded based on the predefined exclusion criteria, and the remaining 105 were compared with the 173 patients from the preintervention group and were included in the analysis for the primary objective (Figure 1).

Patients in the postintervention group were younger than those in the preintervention group (P < .02): otherwise the groups were similar (Table 1). Inappropriate antibiotics for ASB decreased from 10.2% preintervention to 1.9% postintervention (odds ratio, 0.17; P = .01) (Table 2). UC processing despite a negative UA significantly decreased from 100% preintervention to 38.6% postintervention (P < .001) (Table 3). In patients with a negative UA, antibiotic prescribing decreased by 25.3% postintervention, but this difference was not statistically significant.

Discussion

A simple process change at the Hines VA resulted in benefits related to antimicrobial stewardship without conferring adverse outcomes on patient safety. Both UC processing despite a negative UA and inappropriate antibiotic prescribing for ASB were reduced significantly postintervention. This process change was piloted in the ED where UCs are often included as part of the initial diagnostic testing in patients who may not report UTI-related symptoms but for whom a UC is often bundled with other infectious workup, depending on the patient presentation.

Reflex testing of urine specimens has been described in the literature, both in an exploratory nature where impact of a reflex UC cancellation protocol based on certain UA criteria is measured by percent reduction of UCs processed as well as results of such interventions implemented into clinical practice.^11-13 A retrospective study performed at the University of North Carolina Medical Center evaluated patients who presented to the ED during a 6-month period and had both an automated UA and UC collected. UC processing was restricted to UA that was positive for nitrites, leukocyte esterase, bacteria, or > 10 WBC/HPF. Use of this reflex culture cancellation protocol could have eliminated 604 of the 1546 (39.1%) cultures processed. However, 11 of the 314 (3.5%) positive cultures could have been missed.¹³ This same protocol was externally validated at another large academic ED setting, where similar results were found.¹⁴

Limitations

This single-center, pre-post QI study was not without limitations. Manual chart reviews were required, and accuracy of information was dependent on clinician documentation and assessment of UTI-related symptoms. The population studied was predominately older males; thus, results may not be applicable to females or young adults. Additionally, recognition of a negative UA and subsequent cancellation of the UC was dependent on laboratory personnel. As noted in the patient group with a positive UA, some of these UAs were negative and may have been overlooked; therefore, subsequent UCs were inappropriately processed. However, this occurred infrequently and confirmed the low probability of true UTI in the setting of a negative UA. Follow-up for UTI-related symptoms may not have been captured if a patient had presented to an outside facility. Last, definitions of a positive UA differed slightly between the pre- and postintervention groups. The preintervention study defined a positive UA as a WBC count > 5 WBC/HPF and positive leukocyte esterase, whereas the present study defined a positive UA with a WBC count > 5. This may have resulted in an overestimation of positive UA in the postintervention group.

Conclusions

Better selective use of UC testing may improve stewardship resources and reduce costs impacting both ED and clinical laboratories. Furthermore, benefits can include a reduction in the use of time and resources required to collect samples for culture, use of test supplies, the time and effort required to process the large number of negative cultures, and resources devoted to the follow-up of these ED culture results. The described UA to reflex culture process change demonstrated a significant reduction in the processing of inappropriate UC and unnecessary antibiotics for ASB. There were no missed UTIs or other adverse patient outcomes noted. This process change has been implemented in all departments at the Hines VA and additional data will be collected to ensure consistent outcomes.

From Hassenfeld Children’s Hospital at NYU Langone Health, New York, NY (Dr Gallagher), T1D Exchange, Boston, MA (Saketh Rompicherla; Drs Ebekozien, Noor, Odugbesan, and Mungmode; Nicole Rioles, Emma Ospelt), University of Mississippi School of Population Health, Jackson, MS (Dr. Ebekozien), Icahn School of Medicine at Mount Sinai, New York, NY (Drs. Wilkes, O’Malley, and Rapaport), Weill Cornell Medicine, New York, NY (Drs. Antal and Feuer), NYU Long Island School of Medicine, Mineola, NY (Dr. Gabriel), NYU Langone Health, New York, NY (Dr. Golden), Barbara Davis Center, Aurora, CO (Dr. Alonso), Texas Children’s Hospital/Baylor College of Medicine, Houston, TX (Dr. Lyons), Stanford University, Stanford, CA (Dr. Prahalad), Children Mercy Kansas City, MO (Dr. Clements), Indiana University School of Medicine, IN (Dr. Neyman), Rady Children’s Hospital, University of California, San Diego, CA (Dr. Demeterco-Berggren).

Background: Patient outcomes of COVID-19 have improved throughout the pandemic. However, because it is not known whether outcomes of COVID-19 in the type 1 diabetes (T1D) population improved over time, we investigated differences in COVID-19 outcomes for patients with T1D in the United States.

Methods: We analyzed data collected via a registry of patients with T1D and COVID-19 from 56 sites between April 2020 and January 2021. We grouped cases into first surge (April 9, 2020, to July 31, 2020, n = 188) and late surge (August 1, 2020, to January 31, 2021, n = 410), and then compared outcomes between both groups using descriptive statistics and logistic regression models.

Results: Adverse outcomes were more frequent during the first surge, including diabetic ketoacidosis (32% vs 15%, P < .001), severe hypoglycemia (4% vs 1%, P = .04), and hospitalization (52% vs 22%, P < .001). Patients in the first surge were older (28 [SD,18.8] years vs 18.0 [SD, 11.1] years, P < .001), had higher median hemoglobin A1c levels (9.3 [interquartile range {IQR}, 4.0] vs 8.4 (IQR, 2.8), P < .001), and were more likely to use public insurance (107 [57%] vs 154 [38%], P < .001). The odds of hospitalization for adults in the first surge were 5 times higher compared to the late surge (odds ratio, 5.01; 95% CI, 2.11-12.63).

Conclusion: Patients with T1D who presented with COVID-19 during the first surge had a higher proportion of adverse outcomes than those who presented in a later surge.

Keywords: TD1, diabetic ketoacidosis, hypoglycemia.

After the World Health Organization declared the disease caused by the novel coronavirus SARS-CoV-2, COVID-19, a pandemic on March 11, 2020, the Centers for Disease Control and Prevention identified patients with diabetes as high risk for severe illness.^1-7 The case-fatality rate for COVID-19 has significantly improved over the past 2 years. Public health measures, less severe COVID-19 variants, increased access to testing, and new treatments for COVID-19 have contributed to improved outcomes.

The T1D Exchange has previously published findings on COVID-19 outcomes for patients with type 1 diabetes (T1D) using data from the T1D COVID-19 Surveillance Registry.^8-12 Given improved outcomes in COVID-19 in the general population, we sought to determine if outcomes for cases of COVID-19 reported to this registry changed over time.

Methods

This study was coordinated by the T1D Exchange and approved as nonhuman subject research by the Western Institutional Review Board. All participating centers also obtained local institutional review board approval. No identifiable patient information was collected as part of this noninterventional, cross-sectional study.

The T1D Exchange Multi-center COVID-19 Surveillance Study collected data from endocrinology clinics that completed a retrospective chart review and submitted information to T1D Exchange via an online questionnaire for all patients with T1D at their sites who tested positive for COVID-19.^13,14 The questionnaire was administered using the Qualtrics survey platform (www.qualtrics.com version XM) and contained 33 pre-coded and free-text response fields to collect patient and clinical attributes.

Each participating center identified 1 team member for reporting to avoid duplicate case submission. Each submitted case was reviewed for potential errors and incomplete information. The coordinating center verified the number of cases per site for data quality assurance.

Quantitative data were represented as mean (standard deviation) or median (interquartile range). Categorical data were described as the number (percentage) of patients. Summary statistics, including frequency and percentage for categorical variables, were calculated for all patient-related and clinical characteristics. The date August 1, 2021, was selected as the end of the first surge based on a review of national COVID-19 surges.

We used the Fisher’s exact test to assess associations between hospitalization and demographics, HbA1c, diabetes duration, symptoms, and adverse outcomes. In addition, multivariate logistic regression was used to calculate odds ratios (OR). Logistic regression models were used to determine the association between time of surge and hospitalization separately for both the pediatric and adult populations. Each model was adjusted for potential sociodemographic confounders, specifically age, sex, race, insurance, and HbA1c.

All tests were 2-sided, with type 1 error set at 5%. Fisher’s exact test and logistic regression were performed using statistical software R, version 3.6.2 (R Foundation for Statistical Computing).

Results

The characteristics of COVID-19 cases in patients with T1D that were reported early in the pandemic, before August 1, 2020 (first surge), compared with those of cases reported on and after August 1, 2020 (later surges) are shown in Table 1.

Patients with T1D who presented with COVID-19 during the first surge as compared to the later surges were older (mean age 28 [SD, 18.0] years vs 18.8 [SD, 11.1] years; P < .001) and had a longer duration of diabetes (P < .001). The first-surge group also had more patients with >20 years’ diabetes duration (20% vs 9%, P < .001). Obesity, hypertension, and chronic kidney disease were also more commonly reported in first-surge cases (all P < .001).

There was a significant difference in race and ethnicity reported in the first surge vs the later surge cases, with fewer patients identifying as non-Hispanic White (39% vs, 63%, P < .001) and more patients identifying as non-Hispanic Black (29% vs 12%, P < .001). The groups also differed significantly in terms of insurance type, with more people on public insurance in the first-surge group (57% vs 38%, P < .001). In addition, median HbA1c was higher (9.3% vs 8.4%, P < .001) and continuous glucose monitor and insulin pump use were less common (P = .02 and <.001, respectively) in the early surge.

All symptoms and adverse outcomes were reported more often in the first surge, including diabetic ketoacidosis (DKA; 32% vs 15%; P < .001) and severe hypoglycemia (4% vs 1%, P = .04). Hospitalization (52% vs 13%, P < .001) and ICU admission (24% vs 9%, P < .001) were reported more often in the first-surge group.

Regression Analyses

Table 2 shows the results of logistic regression analyses for hospitalization in the pediatric (≤19 years of age) and adult (>19 years of age) groups, along with the odds of hospitalization during the first vs late surge among COVID-positive people with T1D. Adult patients who tested positive in the first surge were about 5 times more likely to be hospitalized than adults who tested positive for infection in the late surge after adjusting for age, insurance type, sex, race, and HbA1c levels. Pediatric patients also had an increased odds for hospitalization during the first surge, but this increase was not statistically significant.

Discussion

Our analysis of COVID-19 cases in patients with T1D reported by diabetes providers across the United States found that adverse outcomes were more prevalent early in the pandemic. There may be a number of reasons for this difference in outcomes between patients who presented in the first surge vs a later surge. First, because testing for COVID-19 was extremely limited and reserved for hospitalized patients early in the pandemic, the first-surge patients with confirmed COVID-19 likely represent a skewed population of higher-acuity patients. This may also explain the relative paucity of cases in younger patients reported early in the pandemic. Second, worse outcomes in the early surge may also have been associated with overwhelmed hospitals in New York City at the start of the outbreak. According to Cummings et al, the abrupt surge of critically ill patients hospitalized with severe acute respiratory distress syndrome initially outpaced their capacity to provide prone-positioning ventilation, which has been expanded since then.¹⁵ While there was very little hypertension, cardiovascular disease, or kidney disease reported in the pediatric groups, there was a higher prevalence of obesity in the pediatric group from the mid-Atlantic region. Obesity has been associated with a worse prognosis for COVID-19 illness in children.¹⁶ Finally, there were 5 deaths reported in this study, all of which were reported during the first surge. Older age and increased rates of cardiovascular disease and chronic kidney disease in the first surge cases likely contributed to worse outcomes for adults in mid-Atlantic region relative to the other regions. Minority race and the use of public insurance, risk factors for more severe outcomes in all regions, were also more common in cases reported from the mid-Atlantic region.

This study has several limitations. First, it is a cross-sectional study that relies upon voluntary provider reports. Second, availability of COVID-19 testing was limited in all regions in spring 2020. Third, different regions of the country experienced subsequent surges at different times within the reported timeframes in this analysis. Fourth, this report time period does not include the impact of the newer COVID-19 variants. Finally, trends in COVID-19 outcomes were affected by the evolution of care that developed throughout 2020.

Conclusion

Adult patients with T1D and COVID-19 who reported during the first surge had about 5 times higher hospitalization odds than those who presented in a later surge.

Corresponding author: Osagie Ebekozien, MD, MPH, 11 Avenue de Lafayette, Boston, MA 02111; oebekozien@t1dexchange.org

Disclosures: Dr Ebekozien reports receiving research grants from Medtronic Diabetes, Eli Lilly, and Dexcom, and receiving honoraria from Medtronic Diabetes.

From Hassenfeld Children’s Hospital at NYU Langone Health, New York, NY (Dr Gallagher), T1D Exchange, Boston, MA (Saketh Rompicherla; Drs Ebekozien, Noor, Odugbesan, and Mungmode; Nicole Rioles, Emma Ospelt), University of Mississippi School of Population Health, Jackson, MS (Dr. Ebekozien), Icahn School of Medicine at Mount Sinai, New York, NY (Drs. Wilkes, O’Malley, and Rapaport), Weill Cornell Medicine, New York, NY (Drs. Antal and Feuer), NYU Long Island School of Medicine, Mineola, NY (Dr. Gabriel), NYU Langone Health, New York, NY (Dr. Golden), Barbara Davis Center, Aurora, CO (Dr. Alonso), Texas Children’s Hospital/Baylor College of Medicine, Houston, TX (Dr. Lyons), Stanford University, Stanford, CA (Dr. Prahalad), Children Mercy Kansas City, MO (Dr. Clements), Indiana University School of Medicine, IN (Dr. Neyman), Rady Children’s Hospital, University of California, San Diego, CA (Dr. Demeterco-Berggren).

Background: Patient outcomes of COVID-19 have improved throughout the pandemic. However, because it is not known whether outcomes of COVID-19 in the type 1 diabetes (T1D) population improved over time, we investigated differences in COVID-19 outcomes for patients with T1D in the United States.

Methods: We analyzed data collected via a registry of patients with T1D and COVID-19 from 56 sites between April 2020 and January 2021. We grouped cases into first surge (April 9, 2020, to July 31, 2020, n = 188) and late surge (August 1, 2020, to January 31, 2021, n = 410), and then compared outcomes between both groups using descriptive statistics and logistic regression models.

Results: Adverse outcomes were more frequent during the first surge, including diabetic ketoacidosis (32% vs 15%, P < .001), severe hypoglycemia (4% vs 1%, P = .04), and hospitalization (52% vs 22%, P < .001). Patients in the first surge were older (28 [SD,18.8] years vs 18.0 [SD, 11.1] years, P < .001), had higher median hemoglobin A1c levels (9.3 [interquartile range {IQR}, 4.0] vs 8.4 (IQR, 2.8), P < .001), and were more likely to use public insurance (107 [57%] vs 154 [38%], P < .001). The odds of hospitalization for adults in the first surge were 5 times higher compared to the late surge (odds ratio, 5.01; 95% CI, 2.11-12.63).

Conclusion: Patients with T1D who presented with COVID-19 during the first surge had a higher proportion of adverse outcomes than those who presented in a later surge.

Keywords: TD1, diabetic ketoacidosis, hypoglycemia.

After the World Health Organization declared the disease caused by the novel coronavirus SARS-CoV-2, COVID-19, a pandemic on March 11, 2020, the Centers for Disease Control and Prevention identified patients with diabetes as high risk for severe illness.^1-7 The case-fatality rate for COVID-19 has significantly improved over the past 2 years. Public health measures, less severe COVID-19 variants, increased access to testing, and new treatments for COVID-19 have contributed to improved outcomes.

The T1D Exchange has previously published findings on COVID-19 outcomes for patients with type 1 diabetes (T1D) using data from the T1D COVID-19 Surveillance Registry.^8-12 Given improved outcomes in COVID-19 in the general population, we sought to determine if outcomes for cases of COVID-19 reported to this registry changed over time.

Methods

This study was coordinated by the T1D Exchange and approved as nonhuman subject research by the Western Institutional Review Board. All participating centers also obtained local institutional review board approval. No identifiable patient information was collected as part of this noninterventional, cross-sectional study.

The T1D Exchange Multi-center COVID-19 Surveillance Study collected data from endocrinology clinics that completed a retrospective chart review and submitted information to T1D Exchange via an online questionnaire for all patients with T1D at their sites who tested positive for COVID-19.^13,14 The questionnaire was administered using the Qualtrics survey platform (www.qualtrics.com version XM) and contained 33 pre-coded and free-text response fields to collect patient and clinical attributes.

Each participating center identified 1 team member for reporting to avoid duplicate case submission. Each submitted case was reviewed for potential errors and incomplete information. The coordinating center verified the number of cases per site for data quality assurance.

Quantitative data were represented as mean (standard deviation) or median (interquartile range). Categorical data were described as the number (percentage) of patients. Summary statistics, including frequency and percentage for categorical variables, were calculated for all patient-related and clinical characteristics. The date August 1, 2021, was selected as the end of the first surge based on a review of national COVID-19 surges.

We used the Fisher’s exact test to assess associations between hospitalization and demographics, HbA1c, diabetes duration, symptoms, and adverse outcomes. In addition, multivariate logistic regression was used to calculate odds ratios (OR). Logistic regression models were used to determine the association between time of surge and hospitalization separately for both the pediatric and adult populations. Each model was adjusted for potential sociodemographic confounders, specifically age, sex, race, insurance, and HbA1c.

All tests were 2-sided, with type 1 error set at 5%. Fisher’s exact test and logistic regression were performed using statistical software R, version 3.6.2 (R Foundation for Statistical Computing).

Results

The characteristics of COVID-19 cases in patients with T1D that were reported early in the pandemic, before August 1, 2020 (first surge), compared with those of cases reported on and after August 1, 2020 (later surges) are shown in Table 1.

Patients with T1D who presented with COVID-19 during the first surge as compared to the later surges were older (mean age 28 [SD, 18.0] years vs 18.8 [SD, 11.1] years; P < .001) and had a longer duration of diabetes (P < .001). The first-surge group also had more patients with >20 years’ diabetes duration (20% vs 9%, P < .001). Obesity, hypertension, and chronic kidney disease were also more commonly reported in first-surge cases (all P < .001).

There was a significant difference in race and ethnicity reported in the first surge vs the later surge cases, with fewer patients identifying as non-Hispanic White (39% vs, 63%, P < .001) and more patients identifying as non-Hispanic Black (29% vs 12%, P < .001). The groups also differed significantly in terms of insurance type, with more people on public insurance in the first-surge group (57% vs 38%, P < .001). In addition, median HbA1c was higher (9.3% vs 8.4%, P < .001) and continuous glucose monitor and insulin pump use were less common (P = .02 and <.001, respectively) in the early surge.

All symptoms and adverse outcomes were reported more often in the first surge, including diabetic ketoacidosis (DKA; 32% vs 15%; P < .001) and severe hypoglycemia (4% vs 1%, P = .04). Hospitalization (52% vs 13%, P < .001) and ICU admission (24% vs 9%, P < .001) were reported more often in the first-surge group.

Regression Analyses

Table 2 shows the results of logistic regression analyses for hospitalization in the pediatric (≤19 years of age) and adult (>19 years of age) groups, along with the odds of hospitalization during the first vs late surge among COVID-positive people with T1D. Adult patients who tested positive in the first surge were about 5 times more likely to be hospitalized than adults who tested positive for infection in the late surge after adjusting for age, insurance type, sex, race, and HbA1c levels. Pediatric patients also had an increased odds for hospitalization during the first surge, but this increase was not statistically significant.

Discussion

Our analysis of COVID-19 cases in patients with T1D reported by diabetes providers across the United States found that adverse outcomes were more prevalent early in the pandemic. There may be a number of reasons for this difference in outcomes between patients who presented in the first surge vs a later surge. First, because testing for COVID-19 was extremely limited and reserved for hospitalized patients early in the pandemic, the first-surge patients with confirmed COVID-19 likely represent a skewed population of higher-acuity patients. This may also explain the relative paucity of cases in younger patients reported early in the pandemic. Second, worse outcomes in the early surge may also have been associated with overwhelmed hospitals in New York City at the start of the outbreak. According to Cummings et al, the abrupt surge of critically ill patients hospitalized with severe acute respiratory distress syndrome initially outpaced their capacity to provide prone-positioning ventilation, which has been expanded since then.¹⁵ While there was very little hypertension, cardiovascular disease, or kidney disease reported in the pediatric groups, there was a higher prevalence of obesity in the pediatric group from the mid-Atlantic region. Obesity has been associated with a worse prognosis for COVID-19 illness in children.¹⁶ Finally, there were 5 deaths reported in this study, all of which were reported during the first surge. Older age and increased rates of cardiovascular disease and chronic kidney disease in the first surge cases likely contributed to worse outcomes for adults in mid-Atlantic region relative to the other regions. Minority race and the use of public insurance, risk factors for more severe outcomes in all regions, were also more common in cases reported from the mid-Atlantic region.

This study has several limitations. First, it is a cross-sectional study that relies upon voluntary provider reports. Second, availability of COVID-19 testing was limited in all regions in spring 2020. Third, different regions of the country experienced subsequent surges at different times within the reported timeframes in this analysis. Fourth, this report time period does not include the impact of the newer COVID-19 variants. Finally, trends in COVID-19 outcomes were affected by the evolution of care that developed throughout 2020.

Conclusion

Adult patients with T1D and COVID-19 who reported during the first surge had about 5 times higher hospitalization odds than those who presented in a later surge.

Corresponding author: Osagie Ebekozien, MD, MPH, 11 Avenue de Lafayette, Boston, MA 02111; oebekozien@t1dexchange.org

Disclosures: Dr Ebekozien reports receiving research grants from Medtronic Diabetes, Eli Lilly, and Dexcom, and receiving honoraria from Medtronic Diabetes.

From Hassenfeld Children’s Hospital at NYU Langone Health, New York, NY (Dr Gallagher), T1D Exchange, Boston, MA (Saketh Rompicherla; Drs Ebekozien, Noor, Odugbesan, and Mungmode; Nicole Rioles, Emma Ospelt), University of Mississippi School of Population Health, Jackson, MS (Dr. Ebekozien), Icahn School of Medicine at Mount Sinai, New York, NY (Drs. Wilkes, O’Malley, and Rapaport), Weill Cornell Medicine, New York, NY (Drs. Antal and Feuer), NYU Long Island School of Medicine, Mineola, NY (Dr. Gabriel), NYU Langone Health, New York, NY (Dr. Golden), Barbara Davis Center, Aurora, CO (Dr. Alonso), Texas Children’s Hospital/Baylor College of Medicine, Houston, TX (Dr. Lyons), Stanford University, Stanford, CA (Dr. Prahalad), Children Mercy Kansas City, MO (Dr. Clements), Indiana University School of Medicine, IN (Dr. Neyman), Rady Children’s Hospital, University of California, San Diego, CA (Dr. Demeterco-Berggren).

Background: Patient outcomes of COVID-19 have improved throughout the pandemic. However, because it is not known whether outcomes of COVID-19 in the type 1 diabetes (T1D) population improved over time, we investigated differences in COVID-19 outcomes for patients with T1D in the United States.

Methods: We analyzed data collected via a registry of patients with T1D and COVID-19 from 56 sites between April 2020 and January 2021. We grouped cases into first surge (April 9, 2020, to July 31, 2020, n = 188) and late surge (August 1, 2020, to January 31, 2021, n = 410), and then compared outcomes between both groups using descriptive statistics and logistic regression models.

Results: Adverse outcomes were more frequent during the first surge, including diabetic ketoacidosis (32% vs 15%, P < .001), severe hypoglycemia (4% vs 1%, P = .04), and hospitalization (52% vs 22%, P < .001). Patients in the first surge were older (28 [SD,18.8] years vs 18.0 [SD, 11.1] years, P < .001), had higher median hemoglobin A1c levels (9.3 [interquartile range {IQR}, 4.0] vs 8.4 (IQR, 2.8), P < .001), and were more likely to use public insurance (107 [57%] vs 154 [38%], P < .001). The odds of hospitalization for adults in the first surge were 5 times higher compared to the late surge (odds ratio, 5.01; 95% CI, 2.11-12.63).

Conclusion: Patients with T1D who presented with COVID-19 during the first surge had a higher proportion of adverse outcomes than those who presented in a later surge.

Keywords: TD1, diabetic ketoacidosis, hypoglycemia.

After the World Health Organization declared the disease caused by the novel coronavirus SARS-CoV-2, COVID-19, a pandemic on March 11, 2020, the Centers for Disease Control and Prevention identified patients with diabetes as high risk for severe illness.^1-7 The case-fatality rate for COVID-19 has significantly improved over the past 2 years. Public health measures, less severe COVID-19 variants, increased access to testing, and new treatments for COVID-19 have contributed to improved outcomes.

The T1D Exchange has previously published findings on COVID-19 outcomes for patients with type 1 diabetes (T1D) using data from the T1D COVID-19 Surveillance Registry.^8-12 Given improved outcomes in COVID-19 in the general population, we sought to determine if outcomes for cases of COVID-19 reported to this registry changed over time.

Methods

This study was coordinated by the T1D Exchange and approved as nonhuman subject research by the Western Institutional Review Board. All participating centers also obtained local institutional review board approval. No identifiable patient information was collected as part of this noninterventional, cross-sectional study.

The T1D Exchange Multi-center COVID-19 Surveillance Study collected data from endocrinology clinics that completed a retrospective chart review and submitted information to T1D Exchange via an online questionnaire for all patients with T1D at their sites who tested positive for COVID-19.^13,14 The questionnaire was administered using the Qualtrics survey platform (www.qualtrics.com version XM) and contained 33 pre-coded and free-text response fields to collect patient and clinical attributes.

Each participating center identified 1 team member for reporting to avoid duplicate case submission. Each submitted case was reviewed for potential errors and incomplete information. The coordinating center verified the number of cases per site for data quality assurance.

Quantitative data were represented as mean (standard deviation) or median (interquartile range). Categorical data were described as the number (percentage) of patients. Summary statistics, including frequency and percentage for categorical variables, were calculated for all patient-related and clinical characteristics. The date August 1, 2021, was selected as the end of the first surge based on a review of national COVID-19 surges.

We used the Fisher’s exact test to assess associations between hospitalization and demographics, HbA1c, diabetes duration, symptoms, and adverse outcomes. In addition, multivariate logistic regression was used to calculate odds ratios (OR). Logistic regression models were used to determine the association between time of surge and hospitalization separately for both the pediatric and adult populations. Each model was adjusted for potential sociodemographic confounders, specifically age, sex, race, insurance, and HbA1c.

All tests were 2-sided, with type 1 error set at 5%. Fisher’s exact test and logistic regression were performed using statistical software R, version 3.6.2 (R Foundation for Statistical Computing).

Results

The characteristics of COVID-19 cases in patients with T1D that were reported early in the pandemic, before August 1, 2020 (first surge), compared with those of cases reported on and after August 1, 2020 (later surges) are shown in Table 1.

Patients with T1D who presented with COVID-19 during the first surge as compared to the later surges were older (mean age 28 [SD, 18.0] years vs 18.8 [SD, 11.1] years; P < .001) and had a longer duration of diabetes (P < .001). The first-surge group also had more patients with >20 years’ diabetes duration (20% vs 9%, P < .001). Obesity, hypertension, and chronic kidney disease were also more commonly reported in first-surge cases (all P < .001).

There was a significant difference in race and ethnicity reported in the first surge vs the later surge cases, with fewer patients identifying as non-Hispanic White (39% vs, 63%, P < .001) and more patients identifying as non-Hispanic Black (29% vs 12%, P < .001). The groups also differed significantly in terms of insurance type, with more people on public insurance in the first-surge group (57% vs 38%, P < .001). In addition, median HbA1c was higher (9.3% vs 8.4%, P < .001) and continuous glucose monitor and insulin pump use were less common (P = .02 and <.001, respectively) in the early surge.

All symptoms and adverse outcomes were reported more often in the first surge, including diabetic ketoacidosis (DKA; 32% vs 15%; P < .001) and severe hypoglycemia (4% vs 1%, P = .04). Hospitalization (52% vs 13%, P < .001) and ICU admission (24% vs 9%, P < .001) were reported more often in the first-surge group.

Regression Analyses

Table 2 shows the results of logistic regression analyses for hospitalization in the pediatric (≤19 years of age) and adult (>19 years of age) groups, along with the odds of hospitalization during the first vs late surge among COVID-positive people with T1D. Adult patients who tested positive in the first surge were about 5 times more likely to be hospitalized than adults who tested positive for infection in the late surge after adjusting for age, insurance type, sex, race, and HbA1c levels. Pediatric patients also had an increased odds for hospitalization during the first surge, but this increase was not statistically significant.

Discussion

Our analysis of COVID-19 cases in patients with T1D reported by diabetes providers across the United States found that adverse outcomes were more prevalent early in the pandemic. There may be a number of reasons for this difference in outcomes between patients who presented in the first surge vs a later surge. First, because testing for COVID-19 was extremely limited and reserved for hospitalized patients early in the pandemic, the first-surge patients with confirmed COVID-19 likely represent a skewed population of higher-acuity patients. This may also explain the relative paucity of cases in younger patients reported early in the pandemic. Second, worse outcomes in the early surge may also have been associated with overwhelmed hospitals in New York City at the start of the outbreak. According to Cummings et al, the abrupt surge of critically ill patients hospitalized with severe acute respiratory distress syndrome initially outpaced their capacity to provide prone-positioning ventilation, which has been expanded since then.¹⁵ While there was very little hypertension, cardiovascular disease, or kidney disease reported in the pediatric groups, there was a higher prevalence of obesity in the pediatric group from the mid-Atlantic region. Obesity has been associated with a worse prognosis for COVID-19 illness in children.¹⁶ Finally, there were 5 deaths reported in this study, all of which were reported during the first surge. Older age and increased rates of cardiovascular disease and chronic kidney disease in the first surge cases likely contributed to worse outcomes for adults in mid-Atlantic region relative to the other regions. Minority race and the use of public insurance, risk factors for more severe outcomes in all regions, were also more common in cases reported from the mid-Atlantic region.

This study has several limitations. First, it is a cross-sectional study that relies upon voluntary provider reports. Second, availability of COVID-19 testing was limited in all regions in spring 2020. Third, different regions of the country experienced subsequent surges at different times within the reported timeframes in this analysis. Fourth, this report time period does not include the impact of the newer COVID-19 variants. Finally, trends in COVID-19 outcomes were affected by the evolution of care that developed throughout 2020.

Conclusion

Adult patients with T1D and COVID-19 who reported during the first surge had about 5 times higher hospitalization odds than those who presented in a later surge.

Corresponding author: Osagie Ebekozien, MD, MPH, 11 Avenue de Lafayette, Boston, MA 02111; oebekozien@t1dexchange.org

Disclosures: Dr Ebekozien reports receiving research grants from Medtronic Diabetes, Eli Lilly, and Dexcom, and receiving honoraria from Medtronic Diabetes.

From the Department of Emergency Medicine, Santa Croce e Carle Hospital, Cuneo, Italy (Drs. Abram, Tosello, Emanuele Bernardi, Allione, Cavalot, Dutto, Corsini, Martini, Sciolla, Sara Bernardi, and Lauria). From the School of Emergency Medicine, University of Turin, Turin, Italy (Drs. Paglietta and Giamello).

Objective: This retrospective and prospective cohort study was designed to describe the characteristics, treatments, and outcomes of patients with SARS-CoV-2 infection (COVID-19) admitted to subintensive care units (SICU) and to identify the variables associated with outcomes. SICUs have been extremely stressed during the pandemic, but most data regarding critically ill COVID-19 patients come from intensive care units (ICUs). Studies about COVID-19 patients in SICUs are lacking.

Setting and participants: The study included 88 COVID-19 patients admitted to our SICU in Cuneo, Italy, between March and May 2020.

Measurements: Clinical and ventilatory data were collected, and patients were divided by outcome. Multivariable logistic regression analysis examined the variables associated with negative outcomes (transfer to the ICU, palliation, or death in a SICU).

Results: A total of 60 patients (68%) had a positive outcome, and 28 patients (32%) had a negative outcome; 69 patients (78%) underwent continuous positive airway pressure (CPAP). Pronation (n = 37 [42%]) had been more frequently adopted in patients who had a positive outcome vs a negative outcome (n = 30 [50%] vs n = 7 [25%]; P = .048), and the median (interquartile range) Pao₂/Fio₂ ratio after 6 hours of prone positioning was lower in patients who had a negative outcome vs a positive outcome (144 [140-168] vs 249 [195-268], P = .006). Independent predictors of a negative outcome were diabetes (odds ratio [OR], 8.22; 95% CI, 1.50-44.70; P = .015), higher D-dimer (OR, 1.28; 95% CI, 1.04-1.57; P = .019), higher lactate dehydrogenase level (OR, 1.003; 95% CI, 1.000-1.006; P = .039), and lower lymphocytes count (OR, 0.996; 95% CI, 0.993-0.999; P = .004).

Conclusion: SICUs have a fundamental role in the treatment of critically ill patients with COVID-19, who require long-term CPAP and pronation cycles. Diabetes, lymphopenia, and high D-dimer and LDH levels are associated with negative outcomes.

Keywords: emergency medicine, noninvasive ventilation, prone position, continuous positive airway pressure.

The COVID-19 pandemic has led to large increases in hospital admissions. Subintensive care units (SICUs) are among the wards most under pressure worldwide,¹ dealing with the increased number of critically ill patients who need noninvasive ventilation, as well as serving as the best alternative to overfilled intensive care units (ICUs). In Italy, SICUs are playing a fundamental role in the management of COVID-19 patients, providing early treatment of respiratory failure by continuous noninvasive ventilation in order to reduce the need for intubation.^2-5 Nevertheless, the great majority of available data about critically ill COVID-19 patients comes from ICUs. Full studies about outcomes of patients in SICUs are lacking and need to be conducted.

We sought to evaluate the characteristics and outcomes of patients admitted to our SICU for COVID-19 to describe the treatments they needed and their impact on prognosis, and to identify the variables associated with patient outcomes.

Methods

Study Design

This cohort study used data from patients who were admitted in the very first weeks of the pandemic. Data were collected retrospectively as well as prospectively, since the ethical committee approved our project. The quality and quantity of data in the 2 groups were comparable.

Data were collected from electronic and written medical records gathered during the patient’s entire stay in our SICU. Data were entered in a database with limited and controlled access. This study complied with the Declaration of Helsinki and was approved by the local ethics committees (ID: MEDURG10).

Study Population

We studied 88 consecutive patients admitted to the SICU of the Santa Croce e Carle Teaching Hospital, Cuneo, Italy, for COVID-19, from March 8 to May 1, 2020. The diagnosis was based on acute respiratory failure associated with SARS-CoV-2 RNA detection on nasopharyngeal swab or tracheal aspirate and/or typical COVID-19 features on a pulmonary computed tomography (CT) scan.⁶ Exclusion criteria were age younger than 18 years and patient denial of permission to use their data for research purposes (the great majority of patients could actively give consent; for patients who were too sick to do so, family members were asked whether they were aware of any reason why the patient would deny consent).

The past medical history and recent symptoms description were obtained by manually reviewing medical records. Epidemiological exposure was defined as contact with SARS-CoV-2–positive people or staying in an epidemic outbreak area. Initial vital parameters, venous blood tests, arterial blood gas analysis, chest x-ray, as well as the result of the nasopharyngeal swab were gathered from the emergency department (ED) examination. (Additional swabs could be requested when the first one was negative but clinical suspicion for COVID-19 was high.) Upon admission to the SICU, a standardized panel of blood tests was performed, which was repeated the next day and then every 48 hours. Arterial blood gas analysis was performed when clinically indicated, at least twice a day, or following a scheduled time in patients undergoing pronation. Charlson Comorbidity Index⁷ and MuLBSTA score⁸ were calculated based on the collected data.

Imaging

Chest ultrasonography was performed in the ED at the time of hospitalization and once a day in the SICU. Pulmonary high-resolution computed tomography (HRCT) was performed when clinically indicated or when the results of nasopharyngeal swabs and/or x-ray results were discordant with COVID-19 clinical suspicion. Contrast CT was performed when pulmonary embolism was suspected.

Medical Therapy

Hydroxychloroquine, antiviral agents, tocilizumab, and ruxolitinib were used in the early phase of the pandemic, then were dismissed after evidence of no efficacy.^9-11 Steroids and low-molecular-weight heparin were used afterward. Enoxaparin was used at the standard prophylactic dosage, and 70% of the anticoagulant dosage was also adopted in patients with moderate-to-severe COVID-19 and D-dimer values >3 times the normal value.^12-14 Antibiotics were given when a bacterial superinfection was suspected.

Oxygen and Ventilatory Therapy

Oxygen support or noninvasive ventilation were started based on patients’ respiratory efficacy, estimated by respiratory rate and the ratio of partial pressure of arterial oxygen and fraction of inspired oxygen (P/F ratio).^15,16 Oxygen support was delivered through nasal cannula, Venturi mask, or reservoir mask. Noninvasive ventilation was performed by continuous positive airway pressure (CPAP) when the P/F ratio was <250 or the respiratory rate was >25 breaths per minute, using the helmet interface.^5,17 Prone positioning during CPAP^18-20 was adopted in patients meeting the acute respiratory distress syndrome (ARDS) criteria²¹ and having persistence of respiratory distress and P/F <300 after a 1-hour trial of CPAP.

The prone position was maintained based on patient tolerance. P/F ratio was measured before pronation (T0), after 1 hour of prone position (T1), before resupination (T2), and 6 hours after resupination (T3). With the same timing, the patient was asked to rate their comfort in each position, from 0 (lack of comfort) to 10 (optimal comfort). Delta P/F was defined as the difference between P/F at T3 and basal P/F at T0.

Outcomes

Positive outcomes were defined as patient discharge from the SICU or transfer to a lower-intensity care ward for treatment continuation. Negative outcomes were defined as need for transfer to the ICU, transfer to another ward for palliation, or death in the SICU.

Statistical Analysis

Continuous data are reported as median and interquartile range (IQR); normal distribution of variables was tested using the Shapiro-Wilk test. Categorical variables were reported as absolute number and percentage. The Mann-Whitney test was used to compare continuous variables between groups, and chi-square test with continuity correction was used for categorical variables. The variables that were most significantly associated with a negative outcome on the univariate analysis were included in a stepwise logistic regression analysis, in order to identify independent predictors of patient outcome. Statistical analysis was performed using JASP (JASP Team) software.

Study Population

Of the 88 patients included in the study, 70% were male; the median age was 66 years (IQR, 60-77). In most patients, the diagnosis of COVID-19 was derived from a positive SARS-CoV-2 nasopharyngeal swab. Six patients, however, maintained a negative swab at all determinations but had clinical and imaging features strongly suggesting COVID-19. No patients met the exclusion criteria. Most patients came from the ED (n = 58 [66%]) or general wards (n = 22 [25%]), while few were transferred from the ICU (n = 8 [9%]). The median length of stay in the SICU was 4 days (IQR, 2-7). An epidemiological link to affected persons or a known virus exposure was identifiable in 37 patients (42%).

Clinical, Laboratory, and Imaging Data

The clinical and anthropometric characteristics of patients are shown in Table 1. Hypertension and smoking habits were prevalent in our population, and the median Charlson Comorbidity Index was 3. Most patients experienced fever, dyspnea, and cough during the days before hospitalization.

Laboratory data showed a marked inflammatory milieu in all studied patients, both at baseline and after 24 and 72 hours. Lymphopenia was observed, along with a significant increase of lactate dehydrogenase (LDH), C-reactive protein (CPR), and D-dimer, and a mild increase of procalcitonin. N-terminal pro-brain natriuretic peptide (NT-proBNP) values were also increased, with normal troponin I values (Table 2).

Medical Therapy

Most patients (89%) received hydroxychloroquine, whereas steroids were used in one-third of the population (36%). Immunomodulators (tocilizumab and ruxolitinib) were restricted to 12 patients (14%). Empirical antiviral therapy was introduced in the first 41 patients (47%). Enoxaparin was the default agent for thromboembolism prophylaxis, and 6 patients (7%) received 70% of the anticoagulating dose.

Oxygen and Ventilatory Therapy

Basal median P/F ratio was 253 (IQR, 218-291), and respiratory rate at triage was 20 breaths/min (IQR, 16-28), underlining a moderate-to-severe respiratory insufficiency at presentation. A total of 69 patients (78%) underwent CPAP, with a median positive end-expiratory pressure (PEEP) of 10.0 cm H₂O (IQR, 7.5-10.0) and fraction of inspired oxygen (Fio₂) of 0.40 (IQR, 0.40-0.50). In 37 patients (42%) who received ongoing CPAP, prone positioning was adopted. In this subgroup, respiratory rate was not significantly different from baseline to resupination (24 vs 25 breaths/min). The median P/F improved from 197 (IQR, 154-236) at baseline to 217 (IQR, 180-262) after pronation (the duration of the prone position was variable, depending on patients’ tolerance: 1 to 6 hours or every pronation cycle). The median delta P/F ratio was 39.4 (IQR, –17.0 to 78.0).

Outcomes

A total of 28 patients (32%) had a negative outcome in the SICU: 8 patients (9%) died, having no clinical indication for higher-intensity care; 6 patients (7%) were transferred to general wards for palliation; and 14 patients (16%) needed an upgrade of cure intensity and were transferred to the ICU. Of these 14 patients, 9 died in the ICU. The total in-hospital mortality of COVID-19 patients, including patients transferred from the SICU to general wards in fair condition, was 27% (n = 24). Clinical, laboratory, and therapeutic characteristics between the 2 groups are shown in Table 4.

Patients who had a negative outcome were significantly older and had more comorbidities, as suggested by a significantly higher prevalence of diabetes and higher Charlson Comorbidity scores (reflecting the mortality risk based on age and comorbidities). The median MuLBSTA score, which estimates the 90-day mortality risk from viral pneumonia, was also higher in patients who had a negative outcome (9.33%). Symptom occurrence was not different in patients with a negative outcome (apart from cough, which was less frequent), but these patients underwent hospitalization earlier—since the appearance of their first COVID-19 symptoms—compared to patients who had a positive outcome. No difference was found in antihypertensive therapy with angiotensin-converting enzyme inhibitors or angiotensin-receptor blockers among outcome groups.

More pronounced laboratory abnormalities were found in patients who had a negative outcome, compared to patients who had a positive outcome: lower lymphocytes and higher C-reactive protein (CRP), procalcitonin, D-dimer, LDH, and NT-proBNP. We found no differences in the radiological distribution of pulmonary involvement in patients who had negative or positive outcomes, nor in the adopted medical treatment.

Data showed no difference in CPAP implementation in the 2 groups. However, prone positioning had been more frequently adopted in the group of patients who had a positive outcome, compared with patients who had a negative outcome. No differences of basal P/F were found in patients who had a negative or positive outcome, but the median P/F after 6 hours of prone position was significantly lower in patients who had a negative outcome. The delta P/F ratio did not differ in the 2 groups of patients.

Multivariate Analysis

A logistic regression model was created, including the variables significantly associated with outcomes in the univariate analysis (age, sex, history of diabetes, lymphocytes, CRP, procalcitonin, LDH, NT-proBNP, and D-dimer). In the multivariate analysis, independent predictors of a negative outcome were history of diabetes (odds ratio [OR], 8.22; 95% CI, 1.50-44.70; P =.015), high D-dimer values (OR, 1.28; CI, 1.04-1.57; P = .019), high LDH values (OR, 1.003; CI, 1.000-1.006; P = .039), and low lymphocytes count (OR, 0.996; CI, 0.993-0.999; P = .004).

Discussion

Role of Subintensive Units and Mortality

The novelty of our report is its attempt to investigate the specific group of COVID-19 patients admitted to a SICU. In Italy, SICUs receive acutely ill, spontaneously breathing patients who need (invasive) hemodynamic monitoring, vasoactive medication, renal replacement therapy, chest- tube placement, thrombolysis, and respiratory noninvasive support. The nurse-to-patient ratio is higher than for general wards (usually 1 nurse to every 4 or 5 patients), though lower than for ICUs. In northern Italy, a great number of COVID-19 patients have required this kind of high-intensity care during the pandemic: Noninvasive ventilation support had to be maintained for several days, pronation maneuvers required a high number of people 2 or 3 times a day, and strict monitoring had to be assured. The SICU setting allows patients to buy time as a bridge to progressive reduction of pulmonary involvement, sometimes preventing the need for intubation.

The high prevalence of negative outcomes in the SICU underlines the complexity of COVID-19 patients in this setting. In fact, published data about mortality for patients with severe COVID-19 pneumonia are similar to ours.^22,23

Clinical, Laboratory, and Imaging Data

Our analysis confirmed a high rate of comorbidities in COVID-19 patients²⁴ and their prognostic role with age.^25,26 A marked inflammatory milieu was a negative prognostic indicator, and associated concomitant bacterial superinfection could have led to a worse prognosis (procalcitonin was associated with negative outcomes).²⁷ The cardiovascular system was nevertheless stressed, as suggested by higher values of NT-proBNP in patients with negative outcomes, which could reflect sepsis-related systemic involvement.²⁸

It is known that the pulmonary damage caused by SARS-CoV-2 has a dynamic radiological and clinical course, with early areas of subsegmental consolidation, and bilateral ground-glass opacities predominating later in the course of the disease.²⁹ This could explain why in our population we found no specific radiological pattern leading to a worse outcome.

Medical Therapy

No specific pharmacological therapy was found to be associated with a positive outcome in our study, just like antiviral and immunomodulator therapies failed to demonstrate effectiveness in subsequent pandemic surges. The low statistical power of our study did not allow us to give insight into the effectiveness of steroids and heparin at any dosage.

PEEP Support and Prone Positioning

Continuous positive airway pressure was initiated in the majority of patients and maintained for several days. This was an absolute novelty, because we rarely had to keep patients in helmets for long. This was feasible thanks to the SICU’s high nurse-to-patient ratio and the possibility of providing monitored sedation. Patients who could no longer tolerate CPAP helmets or did not improve with CPAP support were evaluated with anesthetists for programming further management. No initial data on respiratory rate, level of hypoxemia, or oxygen support need (level of PEEP and Fio₂) could discriminate between outcomes.

Prone positioning during CPAP was implemented in 42% of our study population: P/F ratio amelioration after prone positioning was highly variable, ranging from very good P/F ratio improvements to few responses or no response. No significantly greater delta P/F ratio was seen after the first prone positioning cycle in patients who had a positive outcome, probably due to the small size of our population, but we observed a clear positive trend. Interestingly, patients showing a negative outcome had a lower percentage of long-term responses to prone positioning: 6 hours after resupination, they lost the benefit of prone positioning in terms of P/F ratio amelioration. Similarly, a greater number of patients tolerating prone positioning had a positive outcome. These data give insight on the possible benefits of prone positioning in a noninvasively supported cohort of patients, which has been mentioned in previous studies.^30,31

Outcomes and Variables Associated With Negative Outcomes

After correction for age and sex, we found in multiple regression analysis that higher D-dimer and LDH values, lymphopenia, and history of diabetes were independently associated with a worse outcome. Although our results had low statistical significance, we consider the trend of the obtained odds ratios important from a clinical point of view. These results could lead to greater attention being placed on COVID-19 patients who present with these characteristics upon their arrival to the ED because they have increased risk of death or intensive care need. Clinicians should consider SICU admission for these patients in order to guarantee closer monitoring and possibly more aggressive ventilatory treatments, earlier pronation, or earlier transfer to the ICU.

Limitations

The major limitation to our study is undoubtedly its statistical power, due to its relatively low patient population. Particularly, the small number of patients who underwent pronation did not allow speculation about the efficacy of this technique, although preliminary data seem promising. However, ours is among the first studies regarding patients with COVID-19 admitted to a SICU, and these preliminary data truthfully describe the Italian, and perhaps international, experience with the first surge of the pandemic.

Conclusions

Our data highlight the primary role of the SICU in COVID-19 in adequately treating critically ill patients who have high care needs different from intubation, and who require noninvasive ventilation for prolonged times as well as frequent pronation cycles. This setting of care may represent a valid, reliable, and effective option for critically ill respiratory patients. History of diabetes, lymphopenia, and high D-dimer and LDH values are independently associated with negative outcomes, and patients presenting with these characteristics should be strictly monitored.

Acknowledgments: The authors thank the Informatica System S.R.L., as well as Allessando Mendolia for the pro bono creation of the ISCovidCollect data collecting app.

Corresponding author: Sara Abram, MD, via Coppino, 12100 Cuneo, Italy; sara.abram84@gmail.com.

Disclosures: None.

From the Department of Emergency Medicine, Santa Croce e Carle Hospital, Cuneo, Italy (Drs. Abram, Tosello, Emanuele Bernardi, Allione, Cavalot, Dutto, Corsini, Martini, Sciolla, Sara Bernardi, and Lauria). From the School of Emergency Medicine, University of Turin, Turin, Italy (Drs. Paglietta and Giamello).

Objective: This retrospective and prospective cohort study was designed to describe the characteristics, treatments, and outcomes of patients with SARS-CoV-2 infection (COVID-19) admitted to subintensive care units (SICU) and to identify the variables associated with outcomes. SICUs have been extremely stressed during the pandemic, but most data regarding critically ill COVID-19 patients come from intensive care units (ICUs). Studies about COVID-19 patients in SICUs are lacking.

Setting and participants: The study included 88 COVID-19 patients admitted to our SICU in Cuneo, Italy, between March and May 2020.

Measurements: Clinical and ventilatory data were collected, and patients were divided by outcome. Multivariable logistic regression analysis examined the variables associated with negative outcomes (transfer to the ICU, palliation, or death in a SICU).

Results: A total of 60 patients (68%) had a positive outcome, and 28 patients (32%) had a negative outcome; 69 patients (78%) underwent continuous positive airway pressure (CPAP). Pronation (n = 37 [42%]) had been more frequently adopted in patients who had a positive outcome vs a negative outcome (n = 30 [50%] vs n = 7 [25%]; P = .048), and the median (interquartile range) Pao₂/Fio₂ ratio after 6 hours of prone positioning was lower in patients who had a negative outcome vs a positive outcome (144 [140-168] vs 249 [195-268], P = .006). Independent predictors of a negative outcome were diabetes (odds ratio [OR], 8.22; 95% CI, 1.50-44.70; P = .015), higher D-dimer (OR, 1.28; 95% CI, 1.04-1.57; P = .019), higher lactate dehydrogenase level (OR, 1.003; 95% CI, 1.000-1.006; P = .039), and lower lymphocytes count (OR, 0.996; 95% CI, 0.993-0.999; P = .004).

Conclusion: SICUs have a fundamental role in the treatment of critically ill patients with COVID-19, who require long-term CPAP and pronation cycles. Diabetes, lymphopenia, and high D-dimer and LDH levels are associated with negative outcomes.

Keywords: emergency medicine, noninvasive ventilation, prone position, continuous positive airway pressure.

The COVID-19 pandemic has led to large increases in hospital admissions. Subintensive care units (SICUs) are among the wards most under pressure worldwide,¹ dealing with the increased number of critically ill patients who need noninvasive ventilation, as well as serving as the best alternative to overfilled intensive care units (ICUs). In Italy, SICUs are playing a fundamental role in the management of COVID-19 patients, providing early treatment of respiratory failure by continuous noninvasive ventilation in order to reduce the need for intubation.^2-5 Nevertheless, the great majority of available data about critically ill COVID-19 patients comes from ICUs. Full studies about outcomes of patients in SICUs are lacking and need to be conducted.

We sought to evaluate the characteristics and outcomes of patients admitted to our SICU for COVID-19 to describe the treatments they needed and their impact on prognosis, and to identify the variables associated with patient outcomes.

Methods

Study Design

This cohort study used data from patients who were admitted in the very first weeks of the pandemic. Data were collected retrospectively as well as prospectively, since the ethical committee approved our project. The quality and quantity of data in the 2 groups were comparable.

Data were collected from electronic and written medical records gathered during the patient’s entire stay in our SICU. Data were entered in a database with limited and controlled access. This study complied with the Declaration of Helsinki and was approved by the local ethics committees (ID: MEDURG10).

Study Population

We studied 88 consecutive patients admitted to the SICU of the Santa Croce e Carle Teaching Hospital, Cuneo, Italy, for COVID-19, from March 8 to May 1, 2020. The diagnosis was based on acute respiratory failure associated with SARS-CoV-2 RNA detection on nasopharyngeal swab or tracheal aspirate and/or typical COVID-19 features on a pulmonary computed tomography (CT) scan.⁶ Exclusion criteria were age younger than 18 years and patient denial of permission to use their data for research purposes (the great majority of patients could actively give consent; for patients who were too sick to do so, family members were asked whether they were aware of any reason why the patient would deny consent).

The past medical history and recent symptoms description were obtained by manually reviewing medical records. Epidemiological exposure was defined as contact with SARS-CoV-2–positive people or staying in an epidemic outbreak area. Initial vital parameters, venous blood tests, arterial blood gas analysis, chest x-ray, as well as the result of the nasopharyngeal swab were gathered from the emergency department (ED) examination. (Additional swabs could be requested when the first one was negative but clinical suspicion for COVID-19 was high.) Upon admission to the SICU, a standardized panel of blood tests was performed, which was repeated the next day and then every 48 hours. Arterial blood gas analysis was performed when clinically indicated, at least twice a day, or following a scheduled time in patients undergoing pronation. Charlson Comorbidity Index⁷ and MuLBSTA score⁸ were calculated based on the collected data.

Imaging

Chest ultrasonography was performed in the ED at the time of hospitalization and once a day in the SICU. Pulmonary high-resolution computed tomography (HRCT) was performed when clinically indicated or when the results of nasopharyngeal swabs and/or x-ray results were discordant with COVID-19 clinical suspicion. Contrast CT was performed when pulmonary embolism was suspected.

Medical Therapy

Hydroxychloroquine, antiviral agents, tocilizumab, and ruxolitinib were used in the early phase of the pandemic, then were dismissed after evidence of no efficacy.^9-11 Steroids and low-molecular-weight heparin were used afterward. Enoxaparin was used at the standard prophylactic dosage, and 70% of the anticoagulant dosage was also adopted in patients with moderate-to-severe COVID-19 and D-dimer values >3 times the normal value.^12-14 Antibiotics were given when a bacterial superinfection was suspected.

Oxygen and Ventilatory Therapy

Oxygen support or noninvasive ventilation were started based on patients’ respiratory efficacy, estimated by respiratory rate and the ratio of partial pressure of arterial oxygen and fraction of inspired oxygen (P/F ratio).^15,16 Oxygen support was delivered through nasal cannula, Venturi mask, or reservoir mask. Noninvasive ventilation was performed by continuous positive airway pressure (CPAP) when the P/F ratio was <250 or the respiratory rate was >25 breaths per minute, using the helmet interface.^5,17 Prone positioning during CPAP^18-20 was adopted in patients meeting the acute respiratory distress syndrome (ARDS) criteria²¹ and having persistence of respiratory distress and P/F <300 after a 1-hour trial of CPAP.

The prone position was maintained based on patient tolerance. P/F ratio was measured before pronation (T0), after 1 hour of prone position (T1), before resupination (T2), and 6 hours after resupination (T3). With the same timing, the patient was asked to rate their comfort in each position, from 0 (lack of comfort) to 10 (optimal comfort). Delta P/F was defined as the difference between P/F at T3 and basal P/F at T0.

Outcomes

Positive outcomes were defined as patient discharge from the SICU or transfer to a lower-intensity care ward for treatment continuation. Negative outcomes were defined as need for transfer to the ICU, transfer to another ward for palliation, or death in the SICU.

Statistical Analysis

Continuous data are reported as median and interquartile range (IQR); normal distribution of variables was tested using the Shapiro-Wilk test. Categorical variables were reported as absolute number and percentage. The Mann-Whitney test was used to compare continuous variables between groups, and chi-square test with continuity correction was used for categorical variables. The variables that were most significantly associated with a negative outcome on the univariate analysis were included in a stepwise logistic regression analysis, in order to identify independent predictors of patient outcome. Statistical analysis was performed using JASP (JASP Team) software.

Study Population

Of the 88 patients included in the study, 70% were male; the median age was 66 years (IQR, 60-77). In most patients, the diagnosis of COVID-19 was derived from a positive SARS-CoV-2 nasopharyngeal swab. Six patients, however, maintained a negative swab at all determinations but had clinical and imaging features strongly suggesting COVID-19. No patients met the exclusion criteria. Most patients came from the ED (n = 58 [66%]) or general wards (n = 22 [25%]), while few were transferred from the ICU (n = 8 [9%]). The median length of stay in the SICU was 4 days (IQR, 2-7). An epidemiological link to affected persons or a known virus exposure was identifiable in 37 patients (42%).

Clinical, Laboratory, and Imaging Data

The clinical and anthropometric characteristics of patients are shown in Table 1. Hypertension and smoking habits were prevalent in our population, and the median Charlson Comorbidity Index was 3. Most patients experienced fever, dyspnea, and cough during the days before hospitalization.

Laboratory data showed a marked inflammatory milieu in all studied patients, both at baseline and after 24 and 72 hours. Lymphopenia was observed, along with a significant increase of lactate dehydrogenase (LDH), C-reactive protein (CPR), and D-dimer, and a mild increase of procalcitonin. N-terminal pro-brain natriuretic peptide (NT-proBNP) values were also increased, with normal troponin I values (Table 2).

Medical Therapy

Most patients (89%) received hydroxychloroquine, whereas steroids were used in one-third of the population (36%). Immunomodulators (tocilizumab and ruxolitinib) were restricted to 12 patients (14%). Empirical antiviral therapy was introduced in the first 41 patients (47%). Enoxaparin was the default agent for thromboembolism prophylaxis, and 6 patients (7%) received 70% of the anticoagulating dose.

Oxygen and Ventilatory Therapy

Basal median P/F ratio was 253 (IQR, 218-291), and respiratory rate at triage was 20 breaths/min (IQR, 16-28), underlining a moderate-to-severe respiratory insufficiency at presentation. A total of 69 patients (78%) underwent CPAP, with a median positive end-expiratory pressure (PEEP) of 10.0 cm H₂O (IQR, 7.5-10.0) and fraction of inspired oxygen (Fio₂) of 0.40 (IQR, 0.40-0.50). In 37 patients (42%) who received ongoing CPAP, prone positioning was adopted. In this subgroup, respiratory rate was not significantly different from baseline to resupination (24 vs 25 breaths/min). The median P/F improved from 197 (IQR, 154-236) at baseline to 217 (IQR, 180-262) after pronation (the duration of the prone position was variable, depending on patients’ tolerance: 1 to 6 hours or every pronation cycle). The median delta P/F ratio was 39.4 (IQR, –17.0 to 78.0).

Outcomes

A total of 28 patients (32%) had a negative outcome in the SICU: 8 patients (9%) died, having no clinical indication for higher-intensity care; 6 patients (7%) were transferred to general wards for palliation; and 14 patients (16%) needed an upgrade of cure intensity and were transferred to the ICU. Of these 14 patients, 9 died in the ICU. The total in-hospital mortality of COVID-19 patients, including patients transferred from the SICU to general wards in fair condition, was 27% (n = 24). Clinical, laboratory, and therapeutic characteristics between the 2 groups are shown in Table 4.

Patients who had a negative outcome were significantly older and had more comorbidities, as suggested by a significantly higher prevalence of diabetes and higher Charlson Comorbidity scores (reflecting the mortality risk based on age and comorbidities). The median MuLBSTA score, which estimates the 90-day mortality risk from viral pneumonia, was also higher in patients who had a negative outcome (9.33%). Symptom occurrence was not different in patients with a negative outcome (apart from cough, which was less frequent), but these patients underwent hospitalization earlier—since the appearance of their first COVID-19 symptoms—compared to patients who had a positive outcome. No difference was found in antihypertensive therapy with angiotensin-converting enzyme inhibitors or angiotensin-receptor blockers among outcome groups.

More pronounced laboratory abnormalities were found in patients who had a negative outcome, compared to patients who had a positive outcome: lower lymphocytes and higher C-reactive protein (CRP), procalcitonin, D-dimer, LDH, and NT-proBNP. We found no differences in the radiological distribution of pulmonary involvement in patients who had negative or positive outcomes, nor in the adopted medical treatment.

Data showed no difference in CPAP implementation in the 2 groups. However, prone positioning had been more frequently adopted in the group of patients who had a positive outcome, compared with patients who had a negative outcome. No differences of basal P/F were found in patients who had a negative or positive outcome, but the median P/F after 6 hours of prone position was significantly lower in patients who had a negative outcome. The delta P/F ratio did not differ in the 2 groups of patients.

Multivariate Analysis

A logistic regression model was created, including the variables significantly associated with outcomes in the univariate analysis (age, sex, history of diabetes, lymphocytes, CRP, procalcitonin, LDH, NT-proBNP, and D-dimer). In the multivariate analysis, independent predictors of a negative outcome were history of diabetes (odds ratio [OR], 8.22; 95% CI, 1.50-44.70; P =.015), high D-dimer values (OR, 1.28; CI, 1.04-1.57; P = .019), high LDH values (OR, 1.003; CI, 1.000-1.006; P = .039), and low lymphocytes count (OR, 0.996; CI, 0.993-0.999; P = .004).

Discussion

Role of Subintensive Units and Mortality

The novelty of our report is its attempt to investigate the specific group of COVID-19 patients admitted to a SICU. In Italy, SICUs receive acutely ill, spontaneously breathing patients who need (invasive) hemodynamic monitoring, vasoactive medication, renal replacement therapy, chest- tube placement, thrombolysis, and respiratory noninvasive support. The nurse-to-patient ratio is higher than for general wards (usually 1 nurse to every 4 or 5 patients), though lower than for ICUs. In northern Italy, a great number of COVID-19 patients have required this kind of high-intensity care during the pandemic: Noninvasive ventilation support had to be maintained for several days, pronation maneuvers required a high number of people 2 or 3 times a day, and strict monitoring had to be assured. The SICU setting allows patients to buy time as a bridge to progressive reduction of pulmonary involvement, sometimes preventing the need for intubation.

The high prevalence of negative outcomes in the SICU underlines the complexity of COVID-19 patients in this setting. In fact, published data about mortality for patients with severe COVID-19 pneumonia are similar to ours.^22,23

Clinical, Laboratory, and Imaging Data

Our analysis confirmed a high rate of comorbidities in COVID-19 patients²⁴ and their prognostic role with age.^25,26 A marked inflammatory milieu was a negative prognostic indicator, and associated concomitant bacterial superinfection could have led to a worse prognosis (procalcitonin was associated with negative outcomes).²⁷ The cardiovascular system was nevertheless stressed, as suggested by higher values of NT-proBNP in patients with negative outcomes, which could reflect sepsis-related systemic involvement.²⁸

It is known that the pulmonary damage caused by SARS-CoV-2 has a dynamic radiological and clinical course, with early areas of subsegmental consolidation, and bilateral ground-glass opacities predominating later in the course of the disease.²⁹ This could explain why in our population we found no specific radiological pattern leading to a worse outcome.

Medical Therapy

No specific pharmacological therapy was found to be associated with a positive outcome in our study, just like antiviral and immunomodulator therapies failed to demonstrate effectiveness in subsequent pandemic surges. The low statistical power of our study did not allow us to give insight into the effectiveness of steroids and heparin at any dosage.

PEEP Support and Prone Positioning

Continuous positive airway pressure was initiated in the majority of patients and maintained for several days. This was an absolute novelty, because we rarely had to keep patients in helmets for long. This was feasible thanks to the SICU’s high nurse-to-patient ratio and the possibility of providing monitored sedation. Patients who could no longer tolerate CPAP helmets or did not improve with CPAP support were evaluated with anesthetists for programming further management. No initial data on respiratory rate, level of hypoxemia, or oxygen support need (level of PEEP and Fio₂) could discriminate between outcomes.

Prone positioning during CPAP was implemented in 42% of our study population: P/F ratio amelioration after prone positioning was highly variable, ranging from very good P/F ratio improvements to few responses or no response. No significantly greater delta P/F ratio was seen after the first prone positioning cycle in patients who had a positive outcome, probably due to the small size of our population, but we observed a clear positive trend. Interestingly, patients showing a negative outcome had a lower percentage of long-term responses to prone positioning: 6 hours after resupination, they lost the benefit of prone positioning in terms of P/F ratio amelioration. Similarly, a greater number of patients tolerating prone positioning had a positive outcome. These data give insight on the possible benefits of prone positioning in a noninvasively supported cohort of patients, which has been mentioned in previous studies.^30,31

Outcomes and Variables Associated With Negative Outcomes

After correction for age and sex, we found in multiple regression analysis that higher D-dimer and LDH values, lymphopenia, and history of diabetes were independently associated with a worse outcome. Although our results had low statistical significance, we consider the trend of the obtained odds ratios important from a clinical point of view. These results could lead to greater attention being placed on COVID-19 patients who present with these characteristics upon their arrival to the ED because they have increased risk of death or intensive care need. Clinicians should consider SICU admission for these patients in order to guarantee closer monitoring and possibly more aggressive ventilatory treatments, earlier pronation, or earlier transfer to the ICU.

Limitations

The major limitation to our study is undoubtedly its statistical power, due to its relatively low patient population. Particularly, the small number of patients who underwent pronation did not allow speculation about the efficacy of this technique, although preliminary data seem promising. However, ours is among the first studies regarding patients with COVID-19 admitted to a SICU, and these preliminary data truthfully describe the Italian, and perhaps international, experience with the first surge of the pandemic.

Conclusions

Our data highlight the primary role of the SICU in COVID-19 in adequately treating critically ill patients who have high care needs different from intubation, and who require noninvasive ventilation for prolonged times as well as frequent pronation cycles. This setting of care may represent a valid, reliable, and effective option for critically ill respiratory patients. History of diabetes, lymphopenia, and high D-dimer and LDH values are independently associated with negative outcomes, and patients presenting with these characteristics should be strictly monitored.

Acknowledgments: The authors thank the Informatica System S.R.L., as well as Allessando Mendolia for the pro bono creation of the ISCovidCollect data collecting app.

Corresponding author: Sara Abram, MD, via Coppino, 12100 Cuneo, Italy; sara.abram84@gmail.com.

Disclosures: None.

From the Department of Emergency Medicine, Santa Croce e Carle Hospital, Cuneo, Italy (Drs. Abram, Tosello, Emanuele Bernardi, Allione, Cavalot, Dutto, Corsini, Martini, Sciolla, Sara Bernardi, and Lauria). From the School of Emergency Medicine, University of Turin, Turin, Italy (Drs. Paglietta and Giamello).

Objective: This retrospective and prospective cohort study was designed to describe the characteristics, treatments, and outcomes of patients with SARS-CoV-2 infection (COVID-19) admitted to subintensive care units (SICU) and to identify the variables associated with outcomes. SICUs have been extremely stressed during the pandemic, but most data regarding critically ill COVID-19 patients come from intensive care units (ICUs). Studies about COVID-19 patients in SICUs are lacking.

Setting and participants: The study included 88 COVID-19 patients admitted to our SICU in Cuneo, Italy, between March and May 2020.

Measurements: Clinical and ventilatory data were collected, and patients were divided by outcome. Multivariable logistic regression analysis examined the variables associated with negative outcomes (transfer to the ICU, palliation, or death in a SICU).

Results: A total of 60 patients (68%) had a positive outcome, and 28 patients (32%) had a negative outcome; 69 patients (78%) underwent continuous positive airway pressure (CPAP). Pronation (n = 37 [42%]) had been more frequently adopted in patients who had a positive outcome vs a negative outcome (n = 30 [50%] vs n = 7 [25%]; P = .048), and the median (interquartile range) Pao₂/Fio₂ ratio after 6 hours of prone positioning was lower in patients who had a negative outcome vs a positive outcome (144 [140-168] vs 249 [195-268], P = .006). Independent predictors of a negative outcome were diabetes (odds ratio [OR], 8.22; 95% CI, 1.50-44.70; P = .015), higher D-dimer (OR, 1.28; 95% CI, 1.04-1.57; P = .019), higher lactate dehydrogenase level (OR, 1.003; 95% CI, 1.000-1.006; P = .039), and lower lymphocytes count (OR, 0.996; 95% CI, 0.993-0.999; P = .004).

Conclusion: SICUs have a fundamental role in the treatment of critically ill patients with COVID-19, who require long-term CPAP and pronation cycles. Diabetes, lymphopenia, and high D-dimer and LDH levels are associated with negative outcomes.

Keywords: emergency medicine, noninvasive ventilation, prone position, continuous positive airway pressure.

The COVID-19 pandemic has led to large increases in hospital admissions. Subintensive care units (SICUs) are among the wards most under pressure worldwide,¹ dealing with the increased number of critically ill patients who need noninvasive ventilation, as well as serving as the best alternative to overfilled intensive care units (ICUs). In Italy, SICUs are playing a fundamental role in the management of COVID-19 patients, providing early treatment of respiratory failure by continuous noninvasive ventilation in order to reduce the need for intubation.^2-5 Nevertheless, the great majority of available data about critically ill COVID-19 patients comes from ICUs. Full studies about outcomes of patients in SICUs are lacking and need to be conducted.

We sought to evaluate the characteristics and outcomes of patients admitted to our SICU for COVID-19 to describe the treatments they needed and their impact on prognosis, and to identify the variables associated with patient outcomes.

Methods

Study Design

This cohort study used data from patients who were admitted in the very first weeks of the pandemic. Data were collected retrospectively as well as prospectively, since the ethical committee approved our project. The quality and quantity of data in the 2 groups were comparable.

Data were collected from electronic and written medical records gathered during the patient’s entire stay in our SICU. Data were entered in a database with limited and controlled access. This study complied with the Declaration of Helsinki and was approved by the local ethics committees (ID: MEDURG10).

Study Population

We studied 88 consecutive patients admitted to the SICU of the Santa Croce e Carle Teaching Hospital, Cuneo, Italy, for COVID-19, from March 8 to May 1, 2020. The diagnosis was based on acute respiratory failure associated with SARS-CoV-2 RNA detection on nasopharyngeal swab or tracheal aspirate and/or typical COVID-19 features on a pulmonary computed tomography (CT) scan.⁶ Exclusion criteria were age younger than 18 years and patient denial of permission to use their data for research purposes (the great majority of patients could actively give consent; for patients who were too sick to do so, family members were asked whether they were aware of any reason why the patient would deny consent).

The past medical history and recent symptoms description were obtained by manually reviewing medical records. Epidemiological exposure was defined as contact with SARS-CoV-2–positive people or staying in an epidemic outbreak area. Initial vital parameters, venous blood tests, arterial blood gas analysis, chest x-ray, as well as the result of the nasopharyngeal swab were gathered from the emergency department (ED) examination. (Additional swabs could be requested when the first one was negative but clinical suspicion for COVID-19 was high.) Upon admission to the SICU, a standardized panel of blood tests was performed, which was repeated the next day and then every 48 hours. Arterial blood gas analysis was performed when clinically indicated, at least twice a day, or following a scheduled time in patients undergoing pronation. Charlson Comorbidity Index⁷ and MuLBSTA score⁸ were calculated based on the collected data.

Imaging

Chest ultrasonography was performed in the ED at the time of hospitalization and once a day in the SICU. Pulmonary high-resolution computed tomography (HRCT) was performed when clinically indicated or when the results of nasopharyngeal swabs and/or x-ray results were discordant with COVID-19 clinical suspicion. Contrast CT was performed when pulmonary embolism was suspected.

Medical Therapy

Hydroxychloroquine, antiviral agents, tocilizumab, and ruxolitinib were used in the early phase of the pandemic, then were dismissed after evidence of no efficacy.^9-11 Steroids and low-molecular-weight heparin were used afterward. Enoxaparin was used at the standard prophylactic dosage, and 70% of the anticoagulant dosage was also adopted in patients with moderate-to-severe COVID-19 and D-dimer values >3 times the normal value.^12-14 Antibiotics were given when a bacterial superinfection was suspected.

Oxygen and Ventilatory Therapy

Oxygen support or noninvasive ventilation were started based on patients’ respiratory efficacy, estimated by respiratory rate and the ratio of partial pressure of arterial oxygen and fraction of inspired oxygen (P/F ratio).^15,16 Oxygen support was delivered through nasal cannula, Venturi mask, or reservoir mask. Noninvasive ventilation was performed by continuous positive airway pressure (CPAP) when the P/F ratio was <250 or the respiratory rate was >25 breaths per minute, using the helmet interface.^5,17 Prone positioning during CPAP^18-20 was adopted in patients meeting the acute respiratory distress syndrome (ARDS) criteria²¹ and having persistence of respiratory distress and P/F <300 after a 1-hour trial of CPAP.

The prone position was maintained based on patient tolerance. P/F ratio was measured before pronation (T0), after 1 hour of prone position (T1), before resupination (T2), and 6 hours after resupination (T3). With the same timing, the patient was asked to rate their comfort in each position, from 0 (lack of comfort) to 10 (optimal comfort). Delta P/F was defined as the difference between P/F at T3 and basal P/F at T0.

Outcomes

Positive outcomes were defined as patient discharge from the SICU or transfer to a lower-intensity care ward for treatment continuation. Negative outcomes were defined as need for transfer to the ICU, transfer to another ward for palliation, or death in the SICU.

Statistical Analysis

Continuous data are reported as median and interquartile range (IQR); normal distribution of variables was tested using the Shapiro-Wilk test. Categorical variables were reported as absolute number and percentage. The Mann-Whitney test was used to compare continuous variables between groups, and chi-square test with continuity correction was used for categorical variables. The variables that were most significantly associated with a negative outcome on the univariate analysis were included in a stepwise logistic regression analysis, in order to identify independent predictors of patient outcome. Statistical analysis was performed using JASP (JASP Team) software.

Study Population

Of the 88 patients included in the study, 70% were male; the median age was 66 years (IQR, 60-77). In most patients, the diagnosis of COVID-19 was derived from a positive SARS-CoV-2 nasopharyngeal swab. Six patients, however, maintained a negative swab at all determinations but had clinical and imaging features strongly suggesting COVID-19. No patients met the exclusion criteria. Most patients came from the ED (n = 58 [66%]) or general wards (n = 22 [25%]), while few were transferred from the ICU (n = 8 [9%]). The median length of stay in the SICU was 4 days (IQR, 2-7). An epidemiological link to affected persons or a known virus exposure was identifiable in 37 patients (42%).

Clinical, Laboratory, and Imaging Data

The clinical and anthropometric characteristics of patients are shown in Table 1. Hypertension and smoking habits were prevalent in our population, and the median Charlson Comorbidity Index was 3. Most patients experienced fever, dyspnea, and cough during the days before hospitalization.

Laboratory data showed a marked inflammatory milieu in all studied patients, both at baseline and after 24 and 72 hours. Lymphopenia was observed, along with a significant increase of lactate dehydrogenase (LDH), C-reactive protein (CPR), and D-dimer, and a mild increase of procalcitonin. N-terminal pro-brain natriuretic peptide (NT-proBNP) values were also increased, with normal troponin I values (Table 2).

Medical Therapy

Most patients (89%) received hydroxychloroquine, whereas steroids were used in one-third of the population (36%). Immunomodulators (tocilizumab and ruxolitinib) were restricted to 12 patients (14%). Empirical antiviral therapy was introduced in the first 41 patients (47%). Enoxaparin was the default agent for thromboembolism prophylaxis, and 6 patients (7%) received 70% of the anticoagulating dose.

Oxygen and Ventilatory Therapy

Basal median P/F ratio was 253 (IQR, 218-291), and respiratory rate at triage was 20 breaths/min (IQR, 16-28), underlining a moderate-to-severe respiratory insufficiency at presentation. A total of 69 patients (78%) underwent CPAP, with a median positive end-expiratory pressure (PEEP) of 10.0 cm H₂O (IQR, 7.5-10.0) and fraction of inspired oxygen (Fio₂) of 0.40 (IQR, 0.40-0.50). In 37 patients (42%) who received ongoing CPAP, prone positioning was adopted. In this subgroup, respiratory rate was not significantly different from baseline to resupination (24 vs 25 breaths/min). The median P/F improved from 197 (IQR, 154-236) at baseline to 217 (IQR, 180-262) after pronation (the duration of the prone position was variable, depending on patients’ tolerance: 1 to 6 hours or every pronation cycle). The median delta P/F ratio was 39.4 (IQR, –17.0 to 78.0).

Outcomes

A total of 28 patients (32%) had a negative outcome in the SICU: 8 patients (9%) died, having no clinical indication for higher-intensity care; 6 patients (7%) were transferred to general wards for palliation; and 14 patients (16%) needed an upgrade of cure intensity and were transferred to the ICU. Of these 14 patients, 9 died in the ICU. The total in-hospital mortality of COVID-19 patients, including patients transferred from the SICU to general wards in fair condition, was 27% (n = 24). Clinical, laboratory, and therapeutic characteristics between the 2 groups are shown in Table 4.

Patients who had a negative outcome were significantly older and had more comorbidities, as suggested by a significantly higher prevalence of diabetes and higher Charlson Comorbidity scores (reflecting the mortality risk based on age and comorbidities). The median MuLBSTA score, which estimates the 90-day mortality risk from viral pneumonia, was also higher in patients who had a negative outcome (9.33%). Symptom occurrence was not different in patients with a negative outcome (apart from cough, which was less frequent), but these patients underwent hospitalization earlier—since the appearance of their first COVID-19 symptoms—compared to patients who had a positive outcome. No difference was found in antihypertensive therapy with angiotensin-converting enzyme inhibitors or angiotensin-receptor blockers among outcome groups.

More pronounced laboratory abnormalities were found in patients who had a negative outcome, compared to patients who had a positive outcome: lower lymphocytes and higher C-reactive protein (CRP), procalcitonin, D-dimer, LDH, and NT-proBNP. We found no differences in the radiological distribution of pulmonary involvement in patients who had negative or positive outcomes, nor in the adopted medical treatment.

Data showed no difference in CPAP implementation in the 2 groups. However, prone positioning had been more frequently adopted in the group of patients who had a positive outcome, compared with patients who had a negative outcome. No differences of basal P/F were found in patients who had a negative or positive outcome, but the median P/F after 6 hours of prone position was significantly lower in patients who had a negative outcome. The delta P/F ratio did not differ in the 2 groups of patients.

Multivariate Analysis

A logistic regression model was created, including the variables significantly associated with outcomes in the univariate analysis (age, sex, history of diabetes, lymphocytes, CRP, procalcitonin, LDH, NT-proBNP, and D-dimer). In the multivariate analysis, independent predictors of a negative outcome were history of diabetes (odds ratio [OR], 8.22; 95% CI, 1.50-44.70; P =.015), high D-dimer values (OR, 1.28; CI, 1.04-1.57; P = .019), high LDH values (OR, 1.003; CI, 1.000-1.006; P = .039), and low lymphocytes count (OR, 0.996; CI, 0.993-0.999; P = .004).

Discussion

Role of Subintensive Units and Mortality

The novelty of our report is its attempt to investigate the specific group of COVID-19 patients admitted to a SICU. In Italy, SICUs receive acutely ill, spontaneously breathing patients who need (invasive) hemodynamic monitoring, vasoactive medication, renal replacement therapy, chest- tube placement, thrombolysis, and respiratory noninvasive support. The nurse-to-patient ratio is higher than for general wards (usually 1 nurse to every 4 or 5 patients), though lower than for ICUs. In northern Italy, a great number of COVID-19 patients have required this kind of high-intensity care during the pandemic: Noninvasive ventilation support had to be maintained for several days, pronation maneuvers required a high number of people 2 or 3 times a day, and strict monitoring had to be assured. The SICU setting allows patients to buy time as a bridge to progressive reduction of pulmonary involvement, sometimes preventing the need for intubation.

The high prevalence of negative outcomes in the SICU underlines the complexity of COVID-19 patients in this setting. In fact, published data about mortality for patients with severe COVID-19 pneumonia are similar to ours.^22,23

Clinical, Laboratory, and Imaging Data

Our analysis confirmed a high rate of comorbidities in COVID-19 patients²⁴ and their prognostic role with age.^25,26 A marked inflammatory milieu was a negative prognostic indicator, and associated concomitant bacterial superinfection could have led to a worse prognosis (procalcitonin was associated with negative outcomes).²⁷ The cardiovascular system was nevertheless stressed, as suggested by higher values of NT-proBNP in patients with negative outcomes, which could reflect sepsis-related systemic involvement.²⁸

It is known that the pulmonary damage caused by SARS-CoV-2 has a dynamic radiological and clinical course, with early areas of subsegmental consolidation, and bilateral ground-glass opacities predominating later in the course of the disease.²⁹ This could explain why in our population we found no specific radiological pattern leading to a worse outcome.

Medical Therapy

No specific pharmacological therapy was found to be associated with a positive outcome in our study, just like antiviral and immunomodulator therapies failed to demonstrate effectiveness in subsequent pandemic surges. The low statistical power of our study did not allow us to give insight into the effectiveness of steroids and heparin at any dosage.

PEEP Support and Prone Positioning

Continuous positive airway pressure was initiated in the majority of patients and maintained for several days. This was an absolute novelty, because we rarely had to keep patients in helmets for long. This was feasible thanks to the SICU’s high nurse-to-patient ratio and the possibility of providing monitored sedation. Patients who could no longer tolerate CPAP helmets or did not improve with CPAP support were evaluated with anesthetists for programming further management. No initial data on respiratory rate, level of hypoxemia, or oxygen support need (level of PEEP and Fio₂) could discriminate between outcomes.

Prone positioning during CPAP was implemented in 42% of our study population: P/F ratio amelioration after prone positioning was highly variable, ranging from very good P/F ratio improvements to few responses or no response. No significantly greater delta P/F ratio was seen after the first prone positioning cycle in patients who had a positive outcome, probably due to the small size of our population, but we observed a clear positive trend. Interestingly, patients showing a negative outcome had a lower percentage of long-term responses to prone positioning: 6 hours after resupination, they lost the benefit of prone positioning in terms of P/F ratio amelioration. Similarly, a greater number of patients tolerating prone positioning had a positive outcome. These data give insight on the possible benefits of prone positioning in a noninvasively supported cohort of patients, which has been mentioned in previous studies.^30,31

Outcomes and Variables Associated With Negative Outcomes

After correction for age and sex, we found in multiple regression analysis that higher D-dimer and LDH values, lymphopenia, and history of diabetes were independently associated with a worse outcome. Although our results had low statistical significance, we consider the trend of the obtained odds ratios important from a clinical point of view. These results could lead to greater attention being placed on COVID-19 patients who present with these characteristics upon their arrival to the ED because they have increased risk of death or intensive care need. Clinicians should consider SICU admission for these patients in order to guarantee closer monitoring and possibly more aggressive ventilatory treatments, earlier pronation, or earlier transfer to the ICU.

Limitations

The major limitation to our study is undoubtedly its statistical power, due to its relatively low patient population. Particularly, the small number of patients who underwent pronation did not allow speculation about the efficacy of this technique, although preliminary data seem promising. However, ours is among the first studies regarding patients with COVID-19 admitted to a SICU, and these preliminary data truthfully describe the Italian, and perhaps international, experience with the first surge of the pandemic.

Conclusions

Our data highlight the primary role of the SICU in COVID-19 in adequately treating critically ill patients who have high care needs different from intubation, and who require noninvasive ventilation for prolonged times as well as frequent pronation cycles. This setting of care may represent a valid, reliable, and effective option for critically ill respiratory patients. History of diabetes, lymphopenia, and high D-dimer and LDH values are independently associated with negative outcomes, and patients presenting with these characteristics should be strictly monitored.

Acknowledgments: The authors thank the Informatica System S.R.L., as well as Allessando Mendolia for the pro bono creation of the ISCovidCollect data collecting app.

Corresponding author: Sara Abram, MD, via Coppino, 12100 Cuneo, Italy; sara.abram84@gmail.com.

Disclosures: None.