Original Research

Necrotizing infection of the extremity is a rare but potentially lethal diagnosis with a mortality rate in the range of 17% to 35%.^1-4 The plastic surgery service at the Malcom Randall Veterans Affairs Medical Center (MRVAMC) treats all hand emergencies, including upper extremity infection, in the North Florida/South Georgia Veterans Heath System. There has been a well-coordinated emergency hand care system in place for several years that includes specialty templates on the electronic health record, pre-existing urgent clinic appointments, and single service surgical specialty care.⁵ This facilitates a fluid line of communication between primary care, emergency department (ED) providers, and surgical specialties. The objective of the study was to evaluate our identification, treatment, and outcome of these serious infections.

Methods

The MRVAMC Institutional Review Board approved a retrospective review of necrotizing infection of the upper extremity treated at the facility by the plastic surgery service. Surgical cases over a 9-year period (June 5, 2008-June 5, 2017) were identified by CPT (current procedural technology) codes for amputation and/or debridement of the upper extremity. The charts were reviewed for evidence of necrotizing infection by clinical description or pathology report. The patients’ age, sex, etiology, comorbidities from their problem list, vitals, and laboratory results were recorded upon arrival at the hospital. Time from presentation to surgery, treatment, and outcomes were recorded.

Results

Ten patients were treated for necrotizing infection of the upper extremity over a 9-year period; all were men with an average age of 64 years. Etiologies included nail biting, “bug bites,” crush injuries, burns, suspected IV drug use, and unknown. Nine of 10 patients had diabetes mellitus (DM). Most did not show evidence of hemodynamic instability on hospital arrival (Table). One patient was hypotensive with a mean arterial blood pressure < 65 mm Hg, 2 had heart rates > 100 beats/min, 1 patient had a temperature > 38° C, and 7 had elevated white blood cell (WBC) counts ranging from 11 to 24 k/cmm. Two undiagnosed patients with DM (patients 1 and 8) expressed no complaints of pain and presented with blood glucose > 450 mg/dL with hemoglobin A_1c levels > 12%.

Infectious disease and critical care services were involved in the treatment of several cases when requested. A computed tomography (CT) scan was used in 2 of the patients (patients 1 and 4) to assist in the diagnosis (Figure 1). 

The patient with the largest debridement (patient 4) had a CT that was not suspicious for necrotizing infection the day prior to emergent surgery. Patient 3 was found to have a subclavian stenosis on CT angiography early in the postoperative course and was treated with a carotid to subclavian bypass by the vascular service.

Seven patients out of 10 were treated with surgery within 24 hours on hospital arrival. The severity of the pathology was not initially recognized in 2 of the patients earlier in the review. A third patient resisted surgical treatment until the second hospital day. Four patients had from 1 to 3 digital amputations, 2 patients had wrist disarticulations, and 1 had a distal forearm amputation. 

The proximal amputations were patients with DM who went to the operating room within 24 hours of admission. Cultures grew a wide range of microorganisms, including methicillin-resistant Staphylococcus aureus (MRSA), methicillin-susceptible Staphylococcus aureus (MSSA), β-hemolytic Streptococcus, Streptococcus viridans, Klebsiella pneumoniae, and Prevotella.

Antibiotics were managed by critical care, hospitalist, or infectious disease services and adjusted once final cultures were returned (Table). 

The patients all had a minimum of 2 procedures (range 2-5), including debridement and closure (Figures 2A and 2B and 3A and 3B). There were no perioperative deaths.

 

Discussion

Necrotizing infection of the upper extremity is a rare pathology with a substantial risk of amputation and mortality that requires a high index of suspicion and expeditious referral to a hand surgeon. It is well accepted that the key to survival is prompt surgical debridement of all necrotic tissue, ideally within 24 hours of hospital arrival.^2-4,6 Death is usually secondary to sepsis.³ The classic presentation of pain out of proportion to exam, hypotension, erythema, skin necrosis, elevated WBC count, and fever may not be present and can delay diagnosis.^1-4,6

DM is the most common comorbidity, and reviews have found the disease occurs more often in males, both which are consistent with our study.^1-3 Diabetic infections have been found to be more likely to present as necrotizing infection than are nondiabetic infections and be at a higher risk for amputation.⁷ The patients with the wrist disarticulations and forearm amputation had DM. A minor trauma can be a portal for infection, which can be monomicrobial or polymicrobial.^1,4 Once the diagnosis is suspected, prompt resuscitation, surgical debridement, IV antibiotics, and early intensive care are lifesaving. Hyperbaric oxygen is not available at MRVAMC and was not pursued as a transfer request due to its controversial benefit.⁶

There were no perioperative 30-day mortalities over a 9-year period in patients identified as having necrotizing infection of the upper extremity. This is attributed to an aggressive and well-coordinated, multisystem approach involving emergency, surgical, anesthesia, intensive care, and infectious disease services.

The hand trauma triage system in place at MRVAMC was started in 2008 and presented at the 38th Annual VA Surgeons Meeting in New Haven, Connecticut. The process starts at the level of the ED, urgent care or primary care provider and facilitates rapid access to subspecialty care by reducing unnecessary phone calls and appointment wait times.

All hand emergencies are covered by the plastic surgery service rather than the traditional split coverage between orthopedics and plastic surgery. This provides consistency and continuity for the patients and staff. The electronic health record consult template gives specific instructions to contact the on-call plastic surgeon. The resident/fellow gets called if patient is in-house, and faculty is called if the patient is outside the main hospital. The requesting provider gets instructions on treatment and follow-up. Clinic profiles have appointments reserved for urgent consults during the first hour so that patients can be sent to pre-anesthesia clinic or hand therapy, depending on the diagnosis. This triage system increased our hand trauma volume by a multiple of 6 between 2008 and 2012 but cut the appointment wait time > 1 week by half, as a percentage of consults, and did not significantly increase after-hour use of the operating room. The number of faculty and trainees stayed the same.

We did find that speed to diagnosis for necrotizing infection is an area that can be improved on with a higher clinical suspicion. There is a learning curve to the diagnosis and treatment, which can be prolonged when the index cases do not present themselves often and the patients do not appear in distress. This argues for consistency in hand-specific trauma coverage. The patients were most often initially seen by the resident and examined by a faculty member within hours. There were 4 different plastic surgery faculty involved in these cases, and they all included resident participation before, during, and after surgery. Debridement consists of wide local excision to bleeding tissue. Author review of the operative notes found the numbers of trips to the operating room for debridement can be reduced as the surgeon becomes more confident in the diagnosis and management, resulting in less “whittling” and a more definitive debridement, resulting in a faster recovery.

The LRINEC (Laboratory Risk Indicator for Necrotizing Fasciitis) is a tool that helps to distinguish necrotizing infection from other forms of soft tissue infection by using a point system for laboratory values that include C-reactive protein (CRP), white blood count, hemoglobin, sodium, creatinine, and glucose values.⁸ We do not routinely request CRP results, but 1 of the 2 patients (patient 9) who had the full complement of laboratory tests would have met high-risk criteria. The diagnostic accuracy of this tool has been questioned⁹; however, the authors welcome any method that can rapidly and noninvasively assist in getting the patient proper attention.

The patients were not seen for long-term follow-up, but some did return to the main hospital or clinic for other pathology and were pleased to show off their grip strength after a 3-ray amputation (patient 1) and aesthetics after upper arm and forearm debridement and skin graft reconstruction (patient 4, Figure 4).

A single-ray amputation can be expected to result in a loss of grip and pinch strength, about 43.3% and 33.6%, respectively; however, given the alternative of further loss of life or limb, this was considered a reasonable trade-off.¹⁰ One wrist disarticulation and the forearm amputation were seen by amputee clinic for prosthetic fitting many months after the amputations once the wounds were healed and edema had subsided.

 

Conclusion

A well-coordinated multidisciplinary effort was the key to successful identification and treatment of this serious life- and limb-threatening infection at our institution. We did identify room for improvement in making an earlier diagnosis and performing a more aggressive first debridement.

Acknowledgments
This project is the result of work supported with resources and use of facilities at the Malcom Randall VA Medical Center in Gainesville, Florida.

Necrotizing infection of the extremity is a rare but potentially lethal diagnosis with a mortality rate in the range of 17% to 35%.^1-4 The plastic surgery service at the Malcom Randall Veterans Affairs Medical Center (MRVAMC) treats all hand emergencies, including upper extremity infection, in the North Florida/South Georgia Veterans Heath System. There has been a well-coordinated emergency hand care system in place for several years that includes specialty templates on the electronic health record, pre-existing urgent clinic appointments, and single service surgical specialty care.⁵ This facilitates a fluid line of communication between primary care, emergency department (ED) providers, and surgical specialties. The objective of the study was to evaluate our identification, treatment, and outcome of these serious infections.

Methods

The MRVAMC Institutional Review Board approved a retrospective review of necrotizing infection of the upper extremity treated at the facility by the plastic surgery service. Surgical cases over a 9-year period (June 5, 2008-June 5, 2017) were identified by CPT (current procedural technology) codes for amputation and/or debridement of the upper extremity. The charts were reviewed for evidence of necrotizing infection by clinical description or pathology report. The patients’ age, sex, etiology, comorbidities from their problem list, vitals, and laboratory results were recorded upon arrival at the hospital. Time from presentation to surgery, treatment, and outcomes were recorded.

Results

Ten patients were treated for necrotizing infection of the upper extremity over a 9-year period; all were men with an average age of 64 years. Etiologies included nail biting, “bug bites,” crush injuries, burns, suspected IV drug use, and unknown. Nine of 10 patients had diabetes mellitus (DM). Most did not show evidence of hemodynamic instability on hospital arrival (Table). One patient was hypotensive with a mean arterial blood pressure < 65 mm Hg, 2 had heart rates > 100 beats/min, 1 patient had a temperature > 38° C, and 7 had elevated white blood cell (WBC) counts ranging from 11 to 24 k/cmm. Two undiagnosed patients with DM (patients 1 and 8) expressed no complaints of pain and presented with blood glucose > 450 mg/dL with hemoglobin A_1c levels > 12%.

Infectious disease and critical care services were involved in the treatment of several cases when requested. A computed tomography (CT) scan was used in 2 of the patients (patients 1 and 4) to assist in the diagnosis (Figure 1). 

The patient with the largest debridement (patient 4) had a CT that was not suspicious for necrotizing infection the day prior to emergent surgery. Patient 3 was found to have a subclavian stenosis on CT angiography early in the postoperative course and was treated with a carotid to subclavian bypass by the vascular service.

Seven patients out of 10 were treated with surgery within 24 hours on hospital arrival. The severity of the pathology was not initially recognized in 2 of the patients earlier in the review. A third patient resisted surgical treatment until the second hospital day. Four patients had from 1 to 3 digital amputations, 2 patients had wrist disarticulations, and 1 had a distal forearm amputation. 

The proximal amputations were patients with DM who went to the operating room within 24 hours of admission. Cultures grew a wide range of microorganisms, including methicillin-resistant Staphylococcus aureus (MRSA), methicillin-susceptible Staphylococcus aureus (MSSA), β-hemolytic Streptococcus, Streptococcus viridans, Klebsiella pneumoniae, and Prevotella.

Antibiotics were managed by critical care, hospitalist, or infectious disease services and adjusted once final cultures were returned (Table). 

The patients all had a minimum of 2 procedures (range 2-5), including debridement and closure (Figures 2A and 2B and 3A and 3B). There were no perioperative deaths.

 

Discussion

Necrotizing infection of the upper extremity is a rare pathology with a substantial risk of amputation and mortality that requires a high index of suspicion and expeditious referral to a hand surgeon. It is well accepted that the key to survival is prompt surgical debridement of all necrotic tissue, ideally within 24 hours of hospital arrival.^2-4,6 Death is usually secondary to sepsis.³ The classic presentation of pain out of proportion to exam, hypotension, erythema, skin necrosis, elevated WBC count, and fever may not be present and can delay diagnosis.^1-4,6

DM is the most common comorbidity, and reviews have found the disease occurs more often in males, both which are consistent with our study.^1-3 Diabetic infections have been found to be more likely to present as necrotizing infection than are nondiabetic infections and be at a higher risk for amputation.⁷ The patients with the wrist disarticulations and forearm amputation had DM. A minor trauma can be a portal for infection, which can be monomicrobial or polymicrobial.^1,4 Once the diagnosis is suspected, prompt resuscitation, surgical debridement, IV antibiotics, and early intensive care are lifesaving. Hyperbaric oxygen is not available at MRVAMC and was not pursued as a transfer request due to its controversial benefit.⁶

There were no perioperative 30-day mortalities over a 9-year period in patients identified as having necrotizing infection of the upper extremity. This is attributed to an aggressive and well-coordinated, multisystem approach involving emergency, surgical, anesthesia, intensive care, and infectious disease services.

The hand trauma triage system in place at MRVAMC was started in 2008 and presented at the 38th Annual VA Surgeons Meeting in New Haven, Connecticut. The process starts at the level of the ED, urgent care or primary care provider and facilitates rapid access to subspecialty care by reducing unnecessary phone calls and appointment wait times.

All hand emergencies are covered by the plastic surgery service rather than the traditional split coverage between orthopedics and plastic surgery. This provides consistency and continuity for the patients and staff. The electronic health record consult template gives specific instructions to contact the on-call plastic surgeon. The resident/fellow gets called if patient is in-house, and faculty is called if the patient is outside the main hospital. The requesting provider gets instructions on treatment and follow-up. Clinic profiles have appointments reserved for urgent consults during the first hour so that patients can be sent to pre-anesthesia clinic or hand therapy, depending on the diagnosis. This triage system increased our hand trauma volume by a multiple of 6 between 2008 and 2012 but cut the appointment wait time > 1 week by half, as a percentage of consults, and did not significantly increase after-hour use of the operating room. The number of faculty and trainees stayed the same.

We did find that speed to diagnosis for necrotizing infection is an area that can be improved on with a higher clinical suspicion. There is a learning curve to the diagnosis and treatment, which can be prolonged when the index cases do not present themselves often and the patients do not appear in distress. This argues for consistency in hand-specific trauma coverage. The patients were most often initially seen by the resident and examined by a faculty member within hours. There were 4 different plastic surgery faculty involved in these cases, and they all included resident participation before, during, and after surgery. Debridement consists of wide local excision to bleeding tissue. Author review of the operative notes found the numbers of trips to the operating room for debridement can be reduced as the surgeon becomes more confident in the diagnosis and management, resulting in less “whittling” and a more definitive debridement, resulting in a faster recovery.

The LRINEC (Laboratory Risk Indicator for Necrotizing Fasciitis) is a tool that helps to distinguish necrotizing infection from other forms of soft tissue infection by using a point system for laboratory values that include C-reactive protein (CRP), white blood count, hemoglobin, sodium, creatinine, and glucose values.⁸ We do not routinely request CRP results, but 1 of the 2 patients (patient 9) who had the full complement of laboratory tests would have met high-risk criteria. The diagnostic accuracy of this tool has been questioned⁹; however, the authors welcome any method that can rapidly and noninvasively assist in getting the patient proper attention.

The patients were not seen for long-term follow-up, but some did return to the main hospital or clinic for other pathology and were pleased to show off their grip strength after a 3-ray amputation (patient 1) and aesthetics after upper arm and forearm debridement and skin graft reconstruction (patient 4, Figure 4).

A single-ray amputation can be expected to result in a loss of grip and pinch strength, about 43.3% and 33.6%, respectively; however, given the alternative of further loss of life or limb, this was considered a reasonable trade-off.¹⁰ One wrist disarticulation and the forearm amputation were seen by amputee clinic for prosthetic fitting many months after the amputations once the wounds were healed and edema had subsided.

 

Conclusion

A well-coordinated multidisciplinary effort was the key to successful identification and treatment of this serious life- and limb-threatening infection at our institution. We did identify room for improvement in making an earlier diagnosis and performing a more aggressive first debridement.

Acknowledgments
This project is the result of work supported with resources and use of facilities at the Malcom Randall VA Medical Center in Gainesville, Florida.

Necrotizing infection of the extremity is a rare but potentially lethal diagnosis with a mortality rate in the range of 17% to 35%.^1-4 The plastic surgery service at the Malcom Randall Veterans Affairs Medical Center (MRVAMC) treats all hand emergencies, including upper extremity infection, in the North Florida/South Georgia Veterans Heath System. There has been a well-coordinated emergency hand care system in place for several years that includes specialty templates on the electronic health record, pre-existing urgent clinic appointments, and single service surgical specialty care.⁵ This facilitates a fluid line of communication between primary care, emergency department (ED) providers, and surgical specialties. The objective of the study was to evaluate our identification, treatment, and outcome of these serious infections.

Methods

The MRVAMC Institutional Review Board approved a retrospective review of necrotizing infection of the upper extremity treated at the facility by the plastic surgery service. Surgical cases over a 9-year period (June 5, 2008-June 5, 2017) were identified by CPT (current procedural technology) codes for amputation and/or debridement of the upper extremity. The charts were reviewed for evidence of necrotizing infection by clinical description or pathology report. The patients’ age, sex, etiology, comorbidities from their problem list, vitals, and laboratory results were recorded upon arrival at the hospital. Time from presentation to surgery, treatment, and outcomes were recorded.

Results

Ten patients were treated for necrotizing infection of the upper extremity over a 9-year period; all were men with an average age of 64 years. Etiologies included nail biting, “bug bites,” crush injuries, burns, suspected IV drug use, and unknown. Nine of 10 patients had diabetes mellitus (DM). Most did not show evidence of hemodynamic instability on hospital arrival (Table). One patient was hypotensive with a mean arterial blood pressure < 65 mm Hg, 2 had heart rates > 100 beats/min, 1 patient had a temperature > 38° C, and 7 had elevated white blood cell (WBC) counts ranging from 11 to 24 k/cmm. Two undiagnosed patients with DM (patients 1 and 8) expressed no complaints of pain and presented with blood glucose > 450 mg/dL with hemoglobin A_1c levels > 12%.

Infectious disease and critical care services were involved in the treatment of several cases when requested. A computed tomography (CT) scan was used in 2 of the patients (patients 1 and 4) to assist in the diagnosis (Figure 1). 

The patient with the largest debridement (patient 4) had a CT that was not suspicious for necrotizing infection the day prior to emergent surgery. Patient 3 was found to have a subclavian stenosis on CT angiography early in the postoperative course and was treated with a carotid to subclavian bypass by the vascular service.

Seven patients out of 10 were treated with surgery within 24 hours on hospital arrival. The severity of the pathology was not initially recognized in 2 of the patients earlier in the review. A third patient resisted surgical treatment until the second hospital day. Four patients had from 1 to 3 digital amputations, 2 patients had wrist disarticulations, and 1 had a distal forearm amputation. 

The proximal amputations were patients with DM who went to the operating room within 24 hours of admission. Cultures grew a wide range of microorganisms, including methicillin-resistant Staphylococcus aureus (MRSA), methicillin-susceptible Staphylococcus aureus (MSSA), β-hemolytic Streptococcus, Streptococcus viridans, Klebsiella pneumoniae, and Prevotella.

Antibiotics were managed by critical care, hospitalist, or infectious disease services and adjusted once final cultures were returned (Table). 

The patients all had a minimum of 2 procedures (range 2-5), including debridement and closure (Figures 2A and 2B and 3A and 3B). There were no perioperative deaths.

 

Discussion

Necrotizing infection of the upper extremity is a rare pathology with a substantial risk of amputation and mortality that requires a high index of suspicion and expeditious referral to a hand surgeon. It is well accepted that the key to survival is prompt surgical debridement of all necrotic tissue, ideally within 24 hours of hospital arrival.^2-4,6 Death is usually secondary to sepsis.³ The classic presentation of pain out of proportion to exam, hypotension, erythema, skin necrosis, elevated WBC count, and fever may not be present and can delay diagnosis.^1-4,6

DM is the most common comorbidity, and reviews have found the disease occurs more often in males, both which are consistent with our study.^1-3 Diabetic infections have been found to be more likely to present as necrotizing infection than are nondiabetic infections and be at a higher risk for amputation.⁷ The patients with the wrist disarticulations and forearm amputation had DM. A minor trauma can be a portal for infection, which can be monomicrobial or polymicrobial.^1,4 Once the diagnosis is suspected, prompt resuscitation, surgical debridement, IV antibiotics, and early intensive care are lifesaving. Hyperbaric oxygen is not available at MRVAMC and was not pursued as a transfer request due to its controversial benefit.⁶

There were no perioperative 30-day mortalities over a 9-year period in patients identified as having necrotizing infection of the upper extremity. This is attributed to an aggressive and well-coordinated, multisystem approach involving emergency, surgical, anesthesia, intensive care, and infectious disease services.

The hand trauma triage system in place at MRVAMC was started in 2008 and presented at the 38th Annual VA Surgeons Meeting in New Haven, Connecticut. The process starts at the level of the ED, urgent care or primary care provider and facilitates rapid access to subspecialty care by reducing unnecessary phone calls and appointment wait times.

All hand emergencies are covered by the plastic surgery service rather than the traditional split coverage between orthopedics and plastic surgery. This provides consistency and continuity for the patients and staff. The electronic health record consult template gives specific instructions to contact the on-call plastic surgeon. The resident/fellow gets called if patient is in-house, and faculty is called if the patient is outside the main hospital. The requesting provider gets instructions on treatment and follow-up. Clinic profiles have appointments reserved for urgent consults during the first hour so that patients can be sent to pre-anesthesia clinic or hand therapy, depending on the diagnosis. This triage system increased our hand trauma volume by a multiple of 6 between 2008 and 2012 but cut the appointment wait time > 1 week by half, as a percentage of consults, and did not significantly increase after-hour use of the operating room. The number of faculty and trainees stayed the same.

We did find that speed to diagnosis for necrotizing infection is an area that can be improved on with a higher clinical suspicion. There is a learning curve to the diagnosis and treatment, which can be prolonged when the index cases do not present themselves often and the patients do not appear in distress. This argues for consistency in hand-specific trauma coverage. The patients were most often initially seen by the resident and examined by a faculty member within hours. There were 4 different plastic surgery faculty involved in these cases, and they all included resident participation before, during, and after surgery. Debridement consists of wide local excision to bleeding tissue. Author review of the operative notes found the numbers of trips to the operating room for debridement can be reduced as the surgeon becomes more confident in the diagnosis and management, resulting in less “whittling” and a more definitive debridement, resulting in a faster recovery.

The LRINEC (Laboratory Risk Indicator for Necrotizing Fasciitis) is a tool that helps to distinguish necrotizing infection from other forms of soft tissue infection by using a point system for laboratory values that include C-reactive protein (CRP), white blood count, hemoglobin, sodium, creatinine, and glucose values.⁸ We do not routinely request CRP results, but 1 of the 2 patients (patient 9) who had the full complement of laboratory tests would have met high-risk criteria. The diagnostic accuracy of this tool has been questioned⁹; however, the authors welcome any method that can rapidly and noninvasively assist in getting the patient proper attention.

The patients were not seen for long-term follow-up, but some did return to the main hospital or clinic for other pathology and were pleased to show off their grip strength after a 3-ray amputation (patient 1) and aesthetics after upper arm and forearm debridement and skin graft reconstruction (patient 4, Figure 4).

A single-ray amputation can be expected to result in a loss of grip and pinch strength, about 43.3% and 33.6%, respectively; however, given the alternative of further loss of life or limb, this was considered a reasonable trade-off.¹⁰ One wrist disarticulation and the forearm amputation were seen by amputee clinic for prosthetic fitting many months after the amputations once the wounds were healed and edema had subsided.

 

Conclusion

A well-coordinated multidisciplinary effort was the key to successful identification and treatment of this serious life- and limb-threatening infection at our institution. We did identify room for improvement in making an earlier diagnosis and performing a more aggressive first debridement.

Acknowledgments
This project is the result of work supported with resources and use of facilities at the Malcom Randall VA Medical Center in Gainesville, Florida.

Continuous positive airway pressure therapy (CPAP) is the first-line treatment for obstructive sleep apnea (OSA) recommended by the American College of Physicians and the American Academy of Sleep Medicine.^1,2 CPAP reduces the apnea hypopnea index (AHI), improves oxyhemoglobin desaturation, and reduces cortical arousals associated with apneic/hypopneic events.³ Despite being an effective treatment for OSA, a significant limitation of CPAP is treatment adherence. Factors associated with CPAP adherence include disease and patient characteristics, perceived self-efficacy, treatment titration procedure, device technology factors, adverse effects, and psychosocial factors.⁴

Recent studies suggest that insomnia and depression may be associated with OSA. According to a review by Luyster and colleagues, insomnia is present in 39% to 58% of patients with OSA.⁵ Since OSA may disturb sleep by the number of nightly awakenings, OSA may cause or worsen insomnia. Furthermore, insomnia may exacerbate sleep apnea thus impeding the effectiveness of sleep apnea treatment.

In some studies, the presence of insomnia symptoms prior to initiating CPAP treatment has been found to be associated with reduced CPAP adherence. For example, in 2010, Wickwire and colleagues found that there was a negative association with the average nightly minutes of CPAP use for those patients with OSA that reported symptoms of sleep maintenance insomnia.⁶ This was not found for those patients with OSA who reported symptoms of sleep onset insomnia or reported no insomnia at all. In another study by Pieh and colleagues, self-reported insomnia symptoms were predictive of CPAP adherence (defined as < 4 hours use/night) at a 6-month follow-up.⁷ However, results from a separate study indicated that insomnia was not associated with 6-month CPAP adherence.⁸

Depressive symptoms are commonly reported by patients with OSA, and higher rates of depressive symptomatology in patients with OSA have been observed in a number of prevalence studies when compared with the general population.^9,10 Between 15% and 56% of patients with OSA are diagnosed with a depressive disorder compared with 6.6% of the general population.¹¹ OSA may be causally related with depression or coexist as a separate disorder. Apnea severity has been shown to exacerbate depressive symptoms, and treatment with CPAP can improve depressive symptoms.^12,13 Unfortunately, depression has been found to reduce CPAP adherence. For example, Law and colleagues found that depression was independently associated with poorer adherence during home-based auto-PAP titration.¹⁴ Furthermore, in a study by Gurlanick and colleagues, depressive symptoms were independently associated with reduced CPAP adherence in surgical patients with OSA.¹⁵

To the best of our knowledge, the combined impact of both insomnia and depression on CPAP adherence has not been investigated. In military populations this may be especially important as CPAP adherence has been reported to be worse in military patients with posttraumatic stress disorder (PTSD) and other psychiatric disorders, and there are increasing rates of insomnia and OSA in the military.^16,17 We hypothesize that active-duty and retired military patients with self-reported insomnia and depression will have reduced short and long-term CPAP adherence.

 

Methods

This is a retrospective cohort study that reviewed charts of active-duty and retired military members diagnosed with OSA by the Sleep Medicine Clinic at Naval Medical Center San Diego in California using a home sleep test (HST). The HSTs were interpreted by board-certified physicians in sleep medicine. Prior to the HST, all patients completed a sleep questionnaire that included self-reports of daytime sleepiness, using the Epworth Sleepiness Scale (ESS), depression using the Center for Epidemiologic Studies Depression Scale (CES-D) and insomnia using the Insomnia Severity Index (ISI).

The study population included active-duty and veteran patients diagnosed with OSA who chose treatment with a CPAP and attended the sleep clinic’s OSA educational class, which discussed the diagnosis and treatment of OSA. Inclusion criteria were patients aged > 18 years and diagnosed with OSA at the Naval Medical Center San Diego sleep lab between June 2014 and June 2015.

The study population was stratified into 4 groups: (1) those with OSA but no self-reported depression or insomnia; (2) those with OSA and self-reported depression but no insomnia; (3) those with OSA and insomnia but no depression; and (4) those with OSA and self-reported depression and insomnia. Charts were excluded from the review if there were incomplete data or if the patient selected an alternative treatment for OSA, such as an oral appliance. A total of 120 charts were included in the final review. This study was approved by the Naval Medical Center San Diego Institutional Review Board.

Data Collection

Data collected included the individual’s age, sex, minimum oxygen saturation during sleep, body mass index (BMI), height, weight, ESS score at time of diagnosis, date of HST, and date of attendance at the clinic’s OSA group treatment class. Diagnosis of OSA was based on the patient’s ≥ 5 AHI. OSA severity was divided into mild (AHI 5-14), moderate (AHI 15-29), or severe (AHI ≥ 30). A patient with a CES-D score > 14 was considered to have clinically significant depression, and a patient with an ISI score of > 14 was considered to have clinically significant insomnia. ISI is a reliable and valid instrument to quantify perceived insomnia severity.¹⁸ The CES-D was used only as an indicator of symptoms relating to depression, not to clinically diagnose depression. It also has been used extensively to investigate levels of depression without a psychiatric diagnosis.¹⁹

Follow-up CPAP adherence was collected at 3- and 12-month intervals after the date of the patient’s OSA treatment group class and included AHI, median pressure setting, median days used, average time used per night, and percentage of days used for more than 4 hours for the previous 30 days. Data were obtained through Sleep Data and ResMed websites, which receive patient adherence data directly from the patient’s CPAP device. Patients were considered to be adherent with CPAP usage based on the Medicare definition: Use of the CPAP device > 4 hours per night for at least 70% of nights during a 30-day period). The 3-month time frame was used as a short interval because that is when patients are seen in the pulmonary clinic for their initial follow-up appointment. Patients are seen again at 12 months because durable medical equipment supplies must be reordered after 12 months, which requires a patient visit.

 

Statistical Analysis

Linear regression methods were used to characterize any potential relationships between the predictor variables and the target outcome variables associated with CPAP adherence at 3 and 12 months. Scatterplots were produced to assess whether linear structure was sufficient to characterize any detectable relationships, or whether there existed more complex, nonlinear relationships. The best-fitting linear regression line was examined in relation to the confidence bands of the corresponding LOESS line to determine whether a more complicated model structure was needed to capture the relationship.

Standard tests of assumptions required for these methods were also carried out: QQ plots of residuals to test for normality, the Durbin-Watson test for independence of residuals, and the nonconstant variance score test for heteroskedasticity (ie, Breusch-Pagan test). The results of these assumptions tests are reported only in cases in which the assumptions were revealed to be untenable. In cases in which suspicious outlying observations may have biased analyses, robust versions of the corresponding models were constructed. In no cases did the resulting conclusions change; only the results of the original analysis are reported. All analyses were carried out in R (R Foundation, r-project.org). Statistical significance was defined as P < .05.

Results

Our study population was predominately male (90%) with a median age of 41 years (range 22-65) and BMI of 29.8 (range 7.7-57.2)(Table 1). 

Subjects had a median ESS score of 13 (range 1-23), median ISI score of 14.3 (range 0-28), and a median CES-D score of 16 (range 0-42)(Tables 2 and 3). 

Most of the patients were on auto-CPAP (78%) and had mild OSA with an AHI of 11.1 (range 5.1-81.9). Median CPAP use at 3 months was 5 hours and 15 minutes, and the median CPAP use at 12 months was 6 hours and 3 minutes.

Predictors of CPAP Adherence

OSA severity, as measured by the AHI, was the only promising predictor of CPAP use at 3 months (b, 2.128; t₈₀, 2.854; P = .005; adjusted R², 0.081). The severity of self-reported daytime sleepiness prior to a diagnosis of OSA, as measured by the ESS, did not predict 3-month CPAP adherence (b, 0.688; t₇₇, 0.300; P = .765; adjusted R², -0.012). Self-reported depression as measured by the CES-D also did not predict CPAP use at 3 months (b, -0.078; t₈₀, -0.014; P = .941; adjusted R², -0.012). Similarly, self-reported insomnia, as measured by the ISI, did not predict 3-month CPAP adherence (b, 1.765; t₈₀, 0.939; P = .350; adjusted R², -0.001). Furthermore, a model that incorporated both depression and insomnia proved no better at accounting for variation in 3-month CPAP use (R², -0.012). Demographic variables, such as age, sex, or BMI did not predict 3-month CPAP adherence (all Ps > .20). Finally, median CPAP pressure approached statistical significance as a predictor of 3-month CPAP adherence (b, 9.493; t₆₆, 1.881; P = .064; adjusted R², 0.037) (Figure 1).

CPAP Use at 12 months

The results for CPAP use at 12 months mirrored the results for 3 months with one main exception: OSA severity, as measured by the AHI, did not predict CPAP use at 12 months (b, 1.158; t₅₂, 1.245; P = .219; adjusted R², 0.010). Neither adding a quadratic predictor nor log transforming the AHI values produced a better model (R², -0.0007 vs R², 0.0089, respectively). The severity of self-reported daytime sleepiness, as measured by the ESS, did not predict 12-month CPAP adherence (b, -2.201; t₅₀, -0.752; P = .456; adjusted R² = -0.0086). Self-reported depression as measured by the CES-D also did not predict CPAP use at 12 months (b, 0.034, t₅₂, 0.022; P = .983; adjusted R², -0.092). Self-reported insomnia, as measured by the ISI, also did not predict 12-month CPAP adherence (b, 1.765; t₈₀, 0.939; P = .350; adjusted R² = -0.001). Furthermore, a model that incorporated both depression and insomnia proved no better at accounting for variation in 12-month CPAP use, (R², -0.0298). 

Demographic variables, such as age, sex, or BMI failed to predict 12-month CPAP adherence (all Ps > .15). Finally, median CPAP pressure, in contrast to its promising value as a predictor of 3-month CPAP adherence, did not predict CPAP adherence at 12 months (b, -6.516; t₂₀, -1.021; P = .319; adjusted R² = 0.002) (Figure 2).

 

Discussion

Our study did not provide evidence that self-reported depressive and insomnia symptoms, as measured by the CES-D and ISI, can serve as useful predictors of short and long-term CPAP adherence in a sample of active-duty and retired military. OSA severity, as measured by the AHI, was the only promising predictor of CPAP adherence at 3 months.

Insomnia has been shown to improve with the use of CPAP. In a pilot study, Krakow and colleagues investigated the use of CPAP, oral appliances, or bilateral turbinectomy on patients with OSA and chronic insomnia.²⁰ Objective measures of insomnia improved with 1 night of CPAP titration. Björnsdóttir and colleagues evaluated the long-term effects of positive airway pressure (PAP) treatment on 705 adults with middle insomnia.²¹ They found after 2 years of PAP treatment combined with cognitive behavioral therapy for insomnia, patients had reduced symptoms of middle insomnia. It is possible that persistent insomnia is associated with more severe OSA which was not studied in our population.²²

As reported in other studies, it is possible that patients with depressive symptoms can improve with CPAP use, suggesting that depression and CPAP use are not totally unrelated. Edwards and colleagues studied the impact of CPAP on depressive symptoms in men and woman. They found that depressive symptoms are common in OSA and markedly improve with CPAP.²³ Bopparaju and colleagues found a high prevalence of anxiety and depression in patients with OSA but did not influence CPAP adherence.²⁴

The results of this study differ from some previous findings where depression was found to predict CPAP adherence.¹⁰ This may be due in part to differences in the type of instrument used to assess depression. Wells and colleagues found that baseline depressive symptoms did not correlate with CPAP adherence and that patients with greater CPAP adherence had improvement in OSA and depressive symptoms.²⁵ Furthermore, patients with residual OSA symptoms using CPAP had more depressive symptoms, suggesting that it is the improvement in OSA symptoms that may be correlated with the improvement in depressive symptoms. Although soldiers with PTSD may have reduced CPAP adherence, use of CPAP is associated with improvement in PTSD symptoms.^11,26

Limitations

This study had several limitations, including a small sample size. Study patients were also from a single institution, and the majority of patients had mild-to-moderate OSA. A multicenter prospective study with a larger sample size that included more severe patients with OSA may have shown different results. The participants in this study were limited to members from the active-duty and retired military population. The findings in this population may not be transferrable to the general public. Another study limitation was that the ISI and the CES-D were only administered prior to the initiation of CPAP. If the CES-D and ISI were administered at the 3- and 12-month follow-up visits, we could determine whether short and long-term CPAP improved these symptoms or whether there was no association between CPAP adherence with insomnia and depressive symptoms. Another limitation is that we did not have access to information about potential PTSD symptomatology, which has been associated with reduced CPAP adherence and is more common in a military and veteran population.¹¹

 

Conclusion

This study found little evidence that symptoms of depression and insomnia are useful predictors of CPAP adherence, in either short- or long-term use, in an active-duty and retired military sample. Although these were not found to be predictors of CPAP adherence, further research will be necessary to determine whether CPAP adherence improves symptoms of depression and insomnia in military and veteran populations. Apnea severity did predict CPAP adherence in the short term, but not for any length of time beyond 3 months. More research is needed to explore strategies to improve CPAP adherence in military populations.

Continuous positive airway pressure therapy (CPAP) is the first-line treatment for obstructive sleep apnea (OSA) recommended by the American College of Physicians and the American Academy of Sleep Medicine.^1,2 CPAP reduces the apnea hypopnea index (AHI), improves oxyhemoglobin desaturation, and reduces cortical arousals associated with apneic/hypopneic events.³ Despite being an effective treatment for OSA, a significant limitation of CPAP is treatment adherence. Factors associated with CPAP adherence include disease and patient characteristics, perceived self-efficacy, treatment titration procedure, device technology factors, adverse effects, and psychosocial factors.⁴

Recent studies suggest that insomnia and depression may be associated with OSA. According to a review by Luyster and colleagues, insomnia is present in 39% to 58% of patients with OSA.⁵ Since OSA may disturb sleep by the number of nightly awakenings, OSA may cause or worsen insomnia. Furthermore, insomnia may exacerbate sleep apnea thus impeding the effectiveness of sleep apnea treatment.

In some studies, the presence of insomnia symptoms prior to initiating CPAP treatment has been found to be associated with reduced CPAP adherence. For example, in 2010, Wickwire and colleagues found that there was a negative association with the average nightly minutes of CPAP use for those patients with OSA that reported symptoms of sleep maintenance insomnia.⁶ This was not found for those patients with OSA who reported symptoms of sleep onset insomnia or reported no insomnia at all. In another study by Pieh and colleagues, self-reported insomnia symptoms were predictive of CPAP adherence (defined as < 4 hours use/night) at a 6-month follow-up.⁷ However, results from a separate study indicated that insomnia was not associated with 6-month CPAP adherence.⁸

Depressive symptoms are commonly reported by patients with OSA, and higher rates of depressive symptomatology in patients with OSA have been observed in a number of prevalence studies when compared with the general population.^9,10 Between 15% and 56% of patients with OSA are diagnosed with a depressive disorder compared with 6.6% of the general population.¹¹ OSA may be causally related with depression or coexist as a separate disorder. Apnea severity has been shown to exacerbate depressive symptoms, and treatment with CPAP can improve depressive symptoms.^12,13 Unfortunately, depression has been found to reduce CPAP adherence. For example, Law and colleagues found that depression was independently associated with poorer adherence during home-based auto-PAP titration.¹⁴ Furthermore, in a study by Gurlanick and colleagues, depressive symptoms were independently associated with reduced CPAP adherence in surgical patients with OSA.¹⁵

To the best of our knowledge, the combined impact of both insomnia and depression on CPAP adherence has not been investigated. In military populations this may be especially important as CPAP adherence has been reported to be worse in military patients with posttraumatic stress disorder (PTSD) and other psychiatric disorders, and there are increasing rates of insomnia and OSA in the military.^16,17 We hypothesize that active-duty and retired military patients with self-reported insomnia and depression will have reduced short and long-term CPAP adherence.

 

Methods

This is a retrospective cohort study that reviewed charts of active-duty and retired military members diagnosed with OSA by the Sleep Medicine Clinic at Naval Medical Center San Diego in California using a home sleep test (HST). The HSTs were interpreted by board-certified physicians in sleep medicine. Prior to the HST, all patients completed a sleep questionnaire that included self-reports of daytime sleepiness, using the Epworth Sleepiness Scale (ESS), depression using the Center for Epidemiologic Studies Depression Scale (CES-D) and insomnia using the Insomnia Severity Index (ISI).

The study population included active-duty and veteran patients diagnosed with OSA who chose treatment with a CPAP and attended the sleep clinic’s OSA educational class, which discussed the diagnosis and treatment of OSA. Inclusion criteria were patients aged > 18 years and diagnosed with OSA at the Naval Medical Center San Diego sleep lab between June 2014 and June 2015.

The study population was stratified into 4 groups: (1) those with OSA but no self-reported depression or insomnia; (2) those with OSA and self-reported depression but no insomnia; (3) those with OSA and insomnia but no depression; and (4) those with OSA and self-reported depression and insomnia. Charts were excluded from the review if there were incomplete data or if the patient selected an alternative treatment for OSA, such as an oral appliance. A total of 120 charts were included in the final review. This study was approved by the Naval Medical Center San Diego Institutional Review Board.

Data Collection

Data collected included the individual’s age, sex, minimum oxygen saturation during sleep, body mass index (BMI), height, weight, ESS score at time of diagnosis, date of HST, and date of attendance at the clinic’s OSA group treatment class. Diagnosis of OSA was based on the patient’s ≥ 5 AHI. OSA severity was divided into mild (AHI 5-14), moderate (AHI 15-29), or severe (AHI ≥ 30). A patient with a CES-D score > 14 was considered to have clinically significant depression, and a patient with an ISI score of > 14 was considered to have clinically significant insomnia. ISI is a reliable and valid instrument to quantify perceived insomnia severity.¹⁸ The CES-D was used only as an indicator of symptoms relating to depression, not to clinically diagnose depression. It also has been used extensively to investigate levels of depression without a psychiatric diagnosis.¹⁹

Follow-up CPAP adherence was collected at 3- and 12-month intervals after the date of the patient’s OSA treatment group class and included AHI, median pressure setting, median days used, average time used per night, and percentage of days used for more than 4 hours for the previous 30 days. Data were obtained through Sleep Data and ResMed websites, which receive patient adherence data directly from the patient’s CPAP device. Patients were considered to be adherent with CPAP usage based on the Medicare definition: Use of the CPAP device > 4 hours per night for at least 70% of nights during a 30-day period). The 3-month time frame was used as a short interval because that is when patients are seen in the pulmonary clinic for their initial follow-up appointment. Patients are seen again at 12 months because durable medical equipment supplies must be reordered after 12 months, which requires a patient visit.

 

Statistical Analysis

Linear regression methods were used to characterize any potential relationships between the predictor variables and the target outcome variables associated with CPAP adherence at 3 and 12 months. Scatterplots were produced to assess whether linear structure was sufficient to characterize any detectable relationships, or whether there existed more complex, nonlinear relationships. The best-fitting linear regression line was examined in relation to the confidence bands of the corresponding LOESS line to determine whether a more complicated model structure was needed to capture the relationship.

Standard tests of assumptions required for these methods were also carried out: QQ plots of residuals to test for normality, the Durbin-Watson test for independence of residuals, and the nonconstant variance score test for heteroskedasticity (ie, Breusch-Pagan test). The results of these assumptions tests are reported only in cases in which the assumptions were revealed to be untenable. In cases in which suspicious outlying observations may have biased analyses, robust versions of the corresponding models were constructed. In no cases did the resulting conclusions change; only the results of the original analysis are reported. All analyses were carried out in R (R Foundation, r-project.org). Statistical significance was defined as P < .05.

Results

Our study population was predominately male (90%) with a median age of 41 years (range 22-65) and BMI of 29.8 (range 7.7-57.2)(Table 1). 

Subjects had a median ESS score of 13 (range 1-23), median ISI score of 14.3 (range 0-28), and a median CES-D score of 16 (range 0-42)(Tables 2 and 3). 

Most of the patients were on auto-CPAP (78%) and had mild OSA with an AHI of 11.1 (range 5.1-81.9). Median CPAP use at 3 months was 5 hours and 15 minutes, and the median CPAP use at 12 months was 6 hours and 3 minutes.

Predictors of CPAP Adherence

OSA severity, as measured by the AHI, was the only promising predictor of CPAP use at 3 months (b, 2.128; t₈₀, 2.854; P = .005; adjusted R², 0.081). The severity of self-reported daytime sleepiness prior to a diagnosis of OSA, as measured by the ESS, did not predict 3-month CPAP adherence (b, 0.688; t₇₇, 0.300; P = .765; adjusted R², -0.012). Self-reported depression as measured by the CES-D also did not predict CPAP use at 3 months (b, -0.078; t₈₀, -0.014; P = .941; adjusted R², -0.012). Similarly, self-reported insomnia, as measured by the ISI, did not predict 3-month CPAP adherence (b, 1.765; t₈₀, 0.939; P = .350; adjusted R², -0.001). Furthermore, a model that incorporated both depression and insomnia proved no better at accounting for variation in 3-month CPAP use (R², -0.012). Demographic variables, such as age, sex, or BMI did not predict 3-month CPAP adherence (all Ps > .20). Finally, median CPAP pressure approached statistical significance as a predictor of 3-month CPAP adherence (b, 9.493; t₆₆, 1.881; P = .064; adjusted R², 0.037) (Figure 1).

CPAP Use at 12 months

The results for CPAP use at 12 months mirrored the results for 3 months with one main exception: OSA severity, as measured by the AHI, did not predict CPAP use at 12 months (b, 1.158; t₅₂, 1.245; P = .219; adjusted R², 0.010). Neither adding a quadratic predictor nor log transforming the AHI values produced a better model (R², -0.0007 vs R², 0.0089, respectively). The severity of self-reported daytime sleepiness, as measured by the ESS, did not predict 12-month CPAP adherence (b, -2.201; t₅₀, -0.752; P = .456; adjusted R² = -0.0086). Self-reported depression as measured by the CES-D also did not predict CPAP use at 12 months (b, 0.034, t₅₂, 0.022; P = .983; adjusted R², -0.092). Self-reported insomnia, as measured by the ISI, also did not predict 12-month CPAP adherence (b, 1.765; t₈₀, 0.939; P = .350; adjusted R² = -0.001). Furthermore, a model that incorporated both depression and insomnia proved no better at accounting for variation in 12-month CPAP use, (R², -0.0298). 

Demographic variables, such as age, sex, or BMI failed to predict 12-month CPAP adherence (all Ps > .15). Finally, median CPAP pressure, in contrast to its promising value as a predictor of 3-month CPAP adherence, did not predict CPAP adherence at 12 months (b, -6.516; t₂₀, -1.021; P = .319; adjusted R² = 0.002) (Figure 2).

 

Discussion

Our study did not provide evidence that self-reported depressive and insomnia symptoms, as measured by the CES-D and ISI, can serve as useful predictors of short and long-term CPAP adherence in a sample of active-duty and retired military. OSA severity, as measured by the AHI, was the only promising predictor of CPAP adherence at 3 months.

Insomnia has been shown to improve with the use of CPAP. In a pilot study, Krakow and colleagues investigated the use of CPAP, oral appliances, or bilateral turbinectomy on patients with OSA and chronic insomnia.²⁰ Objective measures of insomnia improved with 1 night of CPAP titration. Björnsdóttir and colleagues evaluated the long-term effects of positive airway pressure (PAP) treatment on 705 adults with middle insomnia.²¹ They found after 2 years of PAP treatment combined with cognitive behavioral therapy for insomnia, patients had reduced symptoms of middle insomnia. It is possible that persistent insomnia is associated with more severe OSA which was not studied in our population.²²

As reported in other studies, it is possible that patients with depressive symptoms can improve with CPAP use, suggesting that depression and CPAP use are not totally unrelated. Edwards and colleagues studied the impact of CPAP on depressive symptoms in men and woman. They found that depressive symptoms are common in OSA and markedly improve with CPAP.²³ Bopparaju and colleagues found a high prevalence of anxiety and depression in patients with OSA but did not influence CPAP adherence.²⁴

The results of this study differ from some previous findings where depression was found to predict CPAP adherence.¹⁰ This may be due in part to differences in the type of instrument used to assess depression. Wells and colleagues found that baseline depressive symptoms did not correlate with CPAP adherence and that patients with greater CPAP adherence had improvement in OSA and depressive symptoms.²⁵ Furthermore, patients with residual OSA symptoms using CPAP had more depressive symptoms, suggesting that it is the improvement in OSA symptoms that may be correlated with the improvement in depressive symptoms. Although soldiers with PTSD may have reduced CPAP adherence, use of CPAP is associated with improvement in PTSD symptoms.^11,26

Limitations

This study had several limitations, including a small sample size. Study patients were also from a single institution, and the majority of patients had mild-to-moderate OSA. A multicenter prospective study with a larger sample size that included more severe patients with OSA may have shown different results. The participants in this study were limited to members from the active-duty and retired military population. The findings in this population may not be transferrable to the general public. Another study limitation was that the ISI and the CES-D were only administered prior to the initiation of CPAP. If the CES-D and ISI were administered at the 3- and 12-month follow-up visits, we could determine whether short and long-term CPAP improved these symptoms or whether there was no association between CPAP adherence with insomnia and depressive symptoms. Another limitation is that we did not have access to information about potential PTSD symptomatology, which has been associated with reduced CPAP adherence and is more common in a military and veteran population.¹¹

 

Conclusion

This study found little evidence that symptoms of depression and insomnia are useful predictors of CPAP adherence, in either short- or long-term use, in an active-duty and retired military sample. Although these were not found to be predictors of CPAP adherence, further research will be necessary to determine whether CPAP adherence improves symptoms of depression and insomnia in military and veteran populations. Apnea severity did predict CPAP adherence in the short term, but not for any length of time beyond 3 months. More research is needed to explore strategies to improve CPAP adherence in military populations.

Continuous positive airway pressure therapy (CPAP) is the first-line treatment for obstructive sleep apnea (OSA) recommended by the American College of Physicians and the American Academy of Sleep Medicine.^1,2 CPAP reduces the apnea hypopnea index (AHI), improves oxyhemoglobin desaturation, and reduces cortical arousals associated with apneic/hypopneic events.³ Despite being an effective treatment for OSA, a significant limitation of CPAP is treatment adherence. Factors associated with CPAP adherence include disease and patient characteristics, perceived self-efficacy, treatment titration procedure, device technology factors, adverse effects, and psychosocial factors.⁴

Recent studies suggest that insomnia and depression may be associated with OSA. According to a review by Luyster and colleagues, insomnia is present in 39% to 58% of patients with OSA.⁵ Since OSA may disturb sleep by the number of nightly awakenings, OSA may cause or worsen insomnia. Furthermore, insomnia may exacerbate sleep apnea thus impeding the effectiveness of sleep apnea treatment.

In some studies, the presence of insomnia symptoms prior to initiating CPAP treatment has been found to be associated with reduced CPAP adherence. For example, in 2010, Wickwire and colleagues found that there was a negative association with the average nightly minutes of CPAP use for those patients with OSA that reported symptoms of sleep maintenance insomnia.⁶ This was not found for those patients with OSA who reported symptoms of sleep onset insomnia or reported no insomnia at all. In another study by Pieh and colleagues, self-reported insomnia symptoms were predictive of CPAP adherence (defined as < 4 hours use/night) at a 6-month follow-up.⁷ However, results from a separate study indicated that insomnia was not associated with 6-month CPAP adherence.⁸

Depressive symptoms are commonly reported by patients with OSA, and higher rates of depressive symptomatology in patients with OSA have been observed in a number of prevalence studies when compared with the general population.^9,10 Between 15% and 56% of patients with OSA are diagnosed with a depressive disorder compared with 6.6% of the general population.¹¹ OSA may be causally related with depression or coexist as a separate disorder. Apnea severity has been shown to exacerbate depressive symptoms, and treatment with CPAP can improve depressive symptoms.^12,13 Unfortunately, depression has been found to reduce CPAP adherence. For example, Law and colleagues found that depression was independently associated with poorer adherence during home-based auto-PAP titration.¹⁴ Furthermore, in a study by Gurlanick and colleagues, depressive symptoms were independently associated with reduced CPAP adherence in surgical patients with OSA.¹⁵

To the best of our knowledge, the combined impact of both insomnia and depression on CPAP adherence has not been investigated. In military populations this may be especially important as CPAP adherence has been reported to be worse in military patients with posttraumatic stress disorder (PTSD) and other psychiatric disorders, and there are increasing rates of insomnia and OSA in the military.^16,17 We hypothesize that active-duty and retired military patients with self-reported insomnia and depression will have reduced short and long-term CPAP adherence.

 

Methods

This is a retrospective cohort study that reviewed charts of active-duty and retired military members diagnosed with OSA by the Sleep Medicine Clinic at Naval Medical Center San Diego in California using a home sleep test (HST). The HSTs were interpreted by board-certified physicians in sleep medicine. Prior to the HST, all patients completed a sleep questionnaire that included self-reports of daytime sleepiness, using the Epworth Sleepiness Scale (ESS), depression using the Center for Epidemiologic Studies Depression Scale (CES-D) and insomnia using the Insomnia Severity Index (ISI).

The study population included active-duty and veteran patients diagnosed with OSA who chose treatment with a CPAP and attended the sleep clinic’s OSA educational class, which discussed the diagnosis and treatment of OSA. Inclusion criteria were patients aged > 18 years and diagnosed with OSA at the Naval Medical Center San Diego sleep lab between June 2014 and June 2015.

The study population was stratified into 4 groups: (1) those with OSA but no self-reported depression or insomnia; (2) those with OSA and self-reported depression but no insomnia; (3) those with OSA and insomnia but no depression; and (4) those with OSA and self-reported depression and insomnia. Charts were excluded from the review if there were incomplete data or if the patient selected an alternative treatment for OSA, such as an oral appliance. A total of 120 charts were included in the final review. This study was approved by the Naval Medical Center San Diego Institutional Review Board.

Data Collection

Data collected included the individual’s age, sex, minimum oxygen saturation during sleep, body mass index (BMI), height, weight, ESS score at time of diagnosis, date of HST, and date of attendance at the clinic’s OSA group treatment class. Diagnosis of OSA was based on the patient’s ≥ 5 AHI. OSA severity was divided into mild (AHI 5-14), moderate (AHI 15-29), or severe (AHI ≥ 30). A patient with a CES-D score > 14 was considered to have clinically significant depression, and a patient with an ISI score of > 14 was considered to have clinically significant insomnia. ISI is a reliable and valid instrument to quantify perceived insomnia severity.¹⁸ The CES-D was used only as an indicator of symptoms relating to depression, not to clinically diagnose depression. It also has been used extensively to investigate levels of depression without a psychiatric diagnosis.¹⁹

Follow-up CPAP adherence was collected at 3- and 12-month intervals after the date of the patient’s OSA treatment group class and included AHI, median pressure setting, median days used, average time used per night, and percentage of days used for more than 4 hours for the previous 30 days. Data were obtained through Sleep Data and ResMed websites, which receive patient adherence data directly from the patient’s CPAP device. Patients were considered to be adherent with CPAP usage based on the Medicare definition: Use of the CPAP device > 4 hours per night for at least 70% of nights during a 30-day period). The 3-month time frame was used as a short interval because that is when patients are seen in the pulmonary clinic for their initial follow-up appointment. Patients are seen again at 12 months because durable medical equipment supplies must be reordered after 12 months, which requires a patient visit.

 

Statistical Analysis

Linear regression methods were used to characterize any potential relationships between the predictor variables and the target outcome variables associated with CPAP adherence at 3 and 12 months. Scatterplots were produced to assess whether linear structure was sufficient to characterize any detectable relationships, or whether there existed more complex, nonlinear relationships. The best-fitting linear regression line was examined in relation to the confidence bands of the corresponding LOESS line to determine whether a more complicated model structure was needed to capture the relationship.

Standard tests of assumptions required for these methods were also carried out: QQ plots of residuals to test for normality, the Durbin-Watson test for independence of residuals, and the nonconstant variance score test for heteroskedasticity (ie, Breusch-Pagan test). The results of these assumptions tests are reported only in cases in which the assumptions were revealed to be untenable. In cases in which suspicious outlying observations may have biased analyses, robust versions of the corresponding models were constructed. In no cases did the resulting conclusions change; only the results of the original analysis are reported. All analyses were carried out in R (R Foundation, r-project.org). Statistical significance was defined as P < .05.

Results

Our study population was predominately male (90%) with a median age of 41 years (range 22-65) and BMI of 29.8 (range 7.7-57.2)(Table 1). 

Subjects had a median ESS score of 13 (range 1-23), median ISI score of 14.3 (range 0-28), and a median CES-D score of 16 (range 0-42)(Tables 2 and 3). 

Most of the patients were on auto-CPAP (78%) and had mild OSA with an AHI of 11.1 (range 5.1-81.9). Median CPAP use at 3 months was 5 hours and 15 minutes, and the median CPAP use at 12 months was 6 hours and 3 minutes.

Predictors of CPAP Adherence

OSA severity, as measured by the AHI, was the only promising predictor of CPAP use at 3 months (b, 2.128; t₈₀, 2.854; P = .005; adjusted R², 0.081). The severity of self-reported daytime sleepiness prior to a diagnosis of OSA, as measured by the ESS, did not predict 3-month CPAP adherence (b, 0.688; t₇₇, 0.300; P = .765; adjusted R², -0.012). Self-reported depression as measured by the CES-D also did not predict CPAP use at 3 months (b, -0.078; t₈₀, -0.014; P = .941; adjusted R², -0.012). Similarly, self-reported insomnia, as measured by the ISI, did not predict 3-month CPAP adherence (b, 1.765; t₈₀, 0.939; P = .350; adjusted R², -0.001). Furthermore, a model that incorporated both depression and insomnia proved no better at accounting for variation in 3-month CPAP use (R², -0.012). Demographic variables, such as age, sex, or BMI did not predict 3-month CPAP adherence (all Ps > .20). Finally, median CPAP pressure approached statistical significance as a predictor of 3-month CPAP adherence (b, 9.493; t₆₆, 1.881; P = .064; adjusted R², 0.037) (Figure 1).

CPAP Use at 12 months

The results for CPAP use at 12 months mirrored the results for 3 months with one main exception: OSA severity, as measured by the AHI, did not predict CPAP use at 12 months (b, 1.158; t₅₂, 1.245; P = .219; adjusted R², 0.010). Neither adding a quadratic predictor nor log transforming the AHI values produced a better model (R², -0.0007 vs R², 0.0089, respectively). The severity of self-reported daytime sleepiness, as measured by the ESS, did not predict 12-month CPAP adherence (b, -2.201; t₅₀, -0.752; P = .456; adjusted R² = -0.0086). Self-reported depression as measured by the CES-D also did not predict CPAP use at 12 months (b, 0.034, t₅₂, 0.022; P = .983; adjusted R², -0.092). Self-reported insomnia, as measured by the ISI, also did not predict 12-month CPAP adherence (b, 1.765; t₈₀, 0.939; P = .350; adjusted R² = -0.001). Furthermore, a model that incorporated both depression and insomnia proved no better at accounting for variation in 12-month CPAP use, (R², -0.0298). 

Demographic variables, such as age, sex, or BMI failed to predict 12-month CPAP adherence (all Ps > .15). Finally, median CPAP pressure, in contrast to its promising value as a predictor of 3-month CPAP adherence, did not predict CPAP adherence at 12 months (b, -6.516; t₂₀, -1.021; P = .319; adjusted R² = 0.002) (Figure 2).

 

Discussion

Our study did not provide evidence that self-reported depressive and insomnia symptoms, as measured by the CES-D and ISI, can serve as useful predictors of short and long-term CPAP adherence in a sample of active-duty and retired military. OSA severity, as measured by the AHI, was the only promising predictor of CPAP adherence at 3 months.

Insomnia has been shown to improve with the use of CPAP. In a pilot study, Krakow and colleagues investigated the use of CPAP, oral appliances, or bilateral turbinectomy on patients with OSA and chronic insomnia.²⁰ Objective measures of insomnia improved with 1 night of CPAP titration. Björnsdóttir and colleagues evaluated the long-term effects of positive airway pressure (PAP) treatment on 705 adults with middle insomnia.²¹ They found after 2 years of PAP treatment combined with cognitive behavioral therapy for insomnia, patients had reduced symptoms of middle insomnia. It is possible that persistent insomnia is associated with more severe OSA which was not studied in our population.²²

As reported in other studies, it is possible that patients with depressive symptoms can improve with CPAP use, suggesting that depression and CPAP use are not totally unrelated. Edwards and colleagues studied the impact of CPAP on depressive symptoms in men and woman. They found that depressive symptoms are common in OSA and markedly improve with CPAP.²³ Bopparaju and colleagues found a high prevalence of anxiety and depression in patients with OSA but did not influence CPAP adherence.²⁴

The results of this study differ from some previous findings where depression was found to predict CPAP adherence.¹⁰ This may be due in part to differences in the type of instrument used to assess depression. Wells and colleagues found that baseline depressive symptoms did not correlate with CPAP adherence and that patients with greater CPAP adherence had improvement in OSA and depressive symptoms.²⁵ Furthermore, patients with residual OSA symptoms using CPAP had more depressive symptoms, suggesting that it is the improvement in OSA symptoms that may be correlated with the improvement in depressive symptoms. Although soldiers with PTSD may have reduced CPAP adherence, use of CPAP is associated with improvement in PTSD symptoms.^11,26

Limitations

This study had several limitations, including a small sample size. Study patients were also from a single institution, and the majority of patients had mild-to-moderate OSA. A multicenter prospective study with a larger sample size that included more severe patients with OSA may have shown different results. The participants in this study were limited to members from the active-duty and retired military population. The findings in this population may not be transferrable to the general public. Another study limitation was that the ISI and the CES-D were only administered prior to the initiation of CPAP. If the CES-D and ISI were administered at the 3- and 12-month follow-up visits, we could determine whether short and long-term CPAP improved these symptoms or whether there was no association between CPAP adherence with insomnia and depressive symptoms. Another limitation is that we did not have access to information about potential PTSD symptomatology, which has been associated with reduced CPAP adherence and is more common in a military and veteran population.¹¹

 

Conclusion

This study found little evidence that symptoms of depression and insomnia are useful predictors of CPAP adherence, in either short- or long-term use, in an active-duty and retired military sample. Although these were not found to be predictors of CPAP adherence, further research will be necessary to determine whether CPAP adherence improves symptoms of depression and insomnia in military and veteran populations. Apnea severity did predict CPAP adherence in the short term, but not for any length of time beyond 3 months. More research is needed to explore strategies to improve CPAP adherence in military populations.

Historically, the Veterans Health Administration (VHA) has excelled at improving veterans’ access to health care and enhancing foundational services, such as prosthetics and other veteran-centric services, and this continues to be the VHA’s top priority.¹ Travel distance and time are often barriers to accessing health care for many veterans.^2-11 For veterans with disabilities who must overcome additional physical, cognitive, and emotional obstacles to access vital rehabilitation services, these geographic obstacles are magnified. Further compounding the challenge is that rehabilitation therapies frequently require multiple encounters. Telerehabilitation is a promising solution for veterans in need of rehabilitation to regain optimal functioning. This alternative mode of service delivery can help veterans overcome geographic access barriers by delivering health care directly to veterans in their homes or nearby community-based outpatient clinics.^12,13

A growing body of evidence supports telerehabilitation. In a 2017 systematic review and meta-analysis, Cottrell and colleagues reviewed and analyzed data from 13 studies that met their inclusion criteria; specifically, their meta-analytic sample comprised adults aged ≥ 18 years presenting with any diagnosed primary musculoskeletal condition; treatment interventions via a real-time telerehabilitation medium, trials that had a comparison group with the same condition; provided clinical outcomes data, and included published randomized and nonrandomized controlled trials.¹⁴ Based on their aggregated results, they concluded that real-time telerehabilitation was effective in improving physical function (standardized mean difference [SMD], 0.63; 95% CI, 0.92-2.33; I², 93%), and reducing pain (SMD, 0.66; 95% CI, −0.27- .60; I², 96%) in patients with any diagnosed primary musculoskeletal condition.¹⁴

Two other systematic reviews conducted by Pietrzak and colleagues and Agostini and colleagues also demonstrated the clinical effectiveness of telerehabilitation.^15,16 Clinical effectiveness was defined as changes in health, functional status, and satisfaction with the telerehabilitation services delivered. The studies examined in the review included those that provided online self-management and education in addition to exercise via teleconferencing in real time.

Pietrzak and colleagues found that Internet-based osteoarthritis self-management interventions significantly improved 4 of 6 health status measures reviewed (ie, pain, fatigue, activity limitation, health distress, disability, and self‐reported global health).¹⁵ User acceptance and satisfaction were high (≥ 70% satisfied) in all studies meeting the inclusion criteria.

Agostini and colleagues found that telerehabilitation was more effective than other modes of delivering rehabilitation to regain motor function in cardiac (SMD, 0.24; 95% CI, 0.04-0.43) and total knee arthroplasty (Timed Up and Go test: SMD, −5.17; 95% CI, −9.79- −0.55) patients.¹⁶ Some evidence from VHA and non-VHA studies also support the use of telerehabilitation to reduce health care costs,^17-19 improve treatment adherence,^12,20 and enhance patient physical, cognitive and mobility function, as well as patient satisfaction and health-related quality of life.^13,21-24

Since the first recorded use of telehealth in 1959, the application of technology to deliver health care, including rehabilitation services, has increased exponentially.¹⁴ In fiscal year (FY) 2017 alone, the VA provided > 2 million episodes of care for > 700,000 veterans using telehealth services.²⁵

Although the process for accessing telerehabilitation may vary throughout the VA, typically a few common factors make a veteran eligible for this mode of rehabilitation care delivery: Veterans must meet criteria for a specific program (eg, amputation, occupational therapy, and physical therapy) and receive VA care from a VA medical facility or clinic that offers telehealth services. Care providers must believe that the veteran would benefit from telerehabilitation (eg, limited mobility and long-distance travel to the facility) and that they would be able to receive an appropriate consult. The veteran must meet the following requirements: (1) willingness to consent to a visit via telehealth; (2) access to required equipment/e-mail; and (3) a caregiver to assist if they are unable to complete a visit independently.

In this article, we provide an overview of the growth of telerehabilitation in the VHA. Data are presented for specific telerehabilitation programs over time and by rurality.

 

Methods

The VHA Support Service Center works with VHA program offices and field users to provide field-focused business, clinical, and special topic reports. An online portal provides access to these customizable reports organized as data cubes, which represent data dimensions (ie, clinic type) and measures (ie, number of unique patients). For this study, we used the Connected Care, Telehealth, Call Centers Clinical Video Telehealth/Store and Forward Telehealth data cube clinical stop codes to identify the numbers of telerehabilitation veteran users and encounters across time. The following telerehabilitation clinic-stop codes were selected: 197 (polytrauma/traumatic brain injury [TBI]–individuals), 201 (Physical Medicine and Rehabilitation [PM&R] Service), 205 (physical therapy), 206 (occupational therapy), 211 (PM&R amputation clinic), 418 (amputation clinic), 214 (kinesiotherapy), and 240 (PM&R assistive technology clinic). Data for total unique patients served and the total number of encounters were extracted at the national level and by rurality from FY 2012 to FY 2017, providing the past 5 years of VHA telerehabilitation data.

It is important to note that in FY 2015, the VHA changed its definition of rurality to a rural-urban commuting areas (RUCA)-based system (www.ruralhealth.va.gov/rural-definition.asp). Prior to FY 2015, the VHA used the US Census Bureau (CB) urbanized area definitions. According to CB, an urbanized area contains a central city and surrounding area that totals > 50,000 in population. It also includes places outside of urbanized areas with populations > 2,500. Rural areas are defined as all other areas. VHA added a third category, highly rural, which is defined as areas that had < 7 people per square mile. In the RUCA system, each census tract defined by the CB is given a score. The VHA definitions are as follows:

• Urban (U)—census tracts with RUCA scores of 1.0 or 1.1. These tracts are determined by the CB as being in an urban core and having the majority of their workers commute within that same core (1.0). If 30% to 49% commute to an even larger urban core, then the code is 1.1;
• Rural (R)—all tracts not receiving scores in the urban or highly rural tiers; and
• Highly rural (H)—tracts with a RUCA score of 10.0. These are the most remote occupied land areas. Less than 10% of workers travel to CB-defined urbanized areas or urban clusters.

In addition, VHA recently added an “I” category to complement “U,” “R,” and “H.” The “I” value is assigned to veterans living on the US insular islands (ie, territories): Guam, American Samoa, Northern Marianas, and US Virgin Islands. For the analysis by rurality in this study, we excluded veterans living in the insular islands and those of unknown rurality (< 1.0% of patients and encounters). Further, because the numbers of highly rural veterans were relatively small (< 2% of patients and encounters), the rural and highly rural categories were combined and compared with urban-dwelling veterans.

Results

Overall, the workload for telerehabilitation nearly quadrupled over the 5-year period (Table 1 and Figure 1). 

In FY 2012, there were 4,397 unique individuals receiving telerehabilitation in the selected telerehabilitation clinics. By FY 2017, this number had grown to 16,319 veterans. 

Similar increases were seen for total encounters, growing from 6,643 in FY 2012 to 22,179 in FY 2017 (Figure 2). The rate of the increase for the number of unique patients seen and telerehabilitation encounter totals across years were higher from FY 2012 to FY 2015 than from FY 2015 to FY 2017.

 

Interesting trends were seen by clinic type. Some clinics increased substantially, whereas others showed only moderate increases, and in 1 case (PM&R Service), a decrease. For example, there is significant growth in the number of patients and encounters involving physical therapy through telerehabilitation. This telerehabilitation clinic increased its workload from 1,676 patients with 3,016 encounters in FY 2012 to 9,136 patients with 11,834 encounters in FY 2017, accounting for 62.6% of total growth in patients and 56.8% of total growth in encounters.

Other clinics showing substantial growth over time included occupational therapy and polytrauma/TBI-individual secondary evaluation. Kinesiotherapy telerehabilitation was almost nonexistent in the VHA during FY 2012, with only 23 patients having 23 encounters. By FY 2017, there were 563 patients with 624 kinesiotherapy telerehabilitation encounters, equating to staggering increases in 5 years: 2,348% for patients and 2,613% for encounters. Similarly, the Physical Medicine and Rehabilitation Assistive Technology clinics had very low numbers in FY 2012 (patients, 2; encounters, 3) and increased over time; albeit, at a slow rate.

Trends by Rurality

Trends by rural location of patients and encounters must be interpreted with caution because of the changing rural definition between FY 2014 and FY 2015 (Tables 2 and 3; Figures 3 and 4). 

Nevertheless, the number of veterans seen and encounters performed via telerehabilitation increased in both urban and rural settings during the time under investigation. 

Under both the legacy and RUCA definitions of rural, the percentage increase was greater for rural veterans than that for urban veterans.

The increased total number of patients seen between FY 2012 and FY 2014 (old definition) was 225% for rural veterans vs 134% for urban veterans. Between FY 2015 and FY 2017 (new definition), the increase was lower for both groups (rural, 13.4%; urban, 7.3%), but rural veterans still increased at a higher rate than did urban dwellers.

Discussion

Our primary aim was to provide data on the growth of telerehabilitation in the VHA over the past 5 years. Our secondary aim was to examine growth in the use of telerehabilitation by rurality. Specifically, we provided an overview of telerehabilitation growth in terms of unique patients and overall encounters in the VHA by rurality from FY 2012 to FY 2014 and FY 2015 to FY 2017 using the following programs: Polytrauma/TBI, PM&R Service, physical therapy, occupational therapy, PM&R amputation clinic, amputation clinic, kinesiotherapy, and PM&R assistive technology clinic. Our findings demonstrated a noteworthy increase in telerehabilitation encounters and unique patients over time for these programs. These findings were consistent with the overall trend of continued growth and expansion of telehealth within the VHA.

Our findings reveal an upward trend in the total number of rural encounters and rural unique patients despite the change in the VA’s definition of rurality in FY 2015. To our knowledge, urban and rural use of telerehabilitation has not been examined previously. Under both definitions of rurality, encounters and unique patients show an important increase over time, and by year-end 2017, more than half of all patients and encounters were attributed to rural patients (53.7% and 53.9%, respectively). Indeed, the upward trend may have been more pronounced if the rural definition had not changed in FY 2015. Our early VHA stroke patients study on the difference between rural-urban patients and taxonomies showed that the RUCA definition was more likely to reduce the number of rural patients by 8.5% than the early definition used by the VHA.²⁶

It is notable that although the use of tele-delivery of rehabilitation has continually increased, the rate of this increase was steeper from FY 2012 to FY 2014 than FY 2015 to FY 2017. For the programs under consideration in this study, the total number of rural patients/encounters increased throughout the observed periods. However, urban patients and encounters increased through FY 2016 and experienced a slight decrease in FY 2017.

The appearance of a slower rate of increase may be due to a rapid initial rate of increase through early adopters and “crossing the diffusion chasm,” a well-documented process of slower diffusion between the time of invention to penetration that often characterizes the spread of successful telehealth innovations.²⁷ Integrating technology into care delivery innovation requires the integration of technical, clinical, and administrative processes and can take time to scale successfully.²⁸

With an emphasis on increasing access to rehabilitation services, the VHA can expect to see a continuing increase in both the number and the percentage of telerehabilitation rural patients and encounters. The VHA has several telerehabilitation initiatives underway through the VHA’s Physical Medicine and Rehabilitation Telerehabilitation Enterprise Wide Initiative (TREWI) and Rural Veterans Telerehabilitation Initiative. These projects demonstrate the feasibility of this delivery approach and facilitate integration of this modality in clinical workflows. However, to sustain these efforts, facilities will need more infrastructure and personnel resources dedicated to the delivery of services.

In an ongoing evaluation of the TREWI, several factors seem to influence the uptake of the VHA Office of Rural Health TREWI programs. These factors are the presence or absence of a local site champion; the quality of hospital leadership support; the quality of past relationships between telerehabilitation sending sites and receiving sites; barriers to getting a telehealth service agreement in place; the availability of space; administrative know-how on setting up clinics appropriately; time involved to bring on staff; contracting issues; equipment availability and installation; cultural issues in embracing technologic innovation; training burden; hassle factors; and limited funds. Although early adopters may be able to negotiate and push through many of the barriers associated with the diffusion of telerehabilitation, the numerous barriers may slow its larger systemwide diffusion.

Telerehabilitation is a promising mode to deliver care to rural veterans who otherwise may not have access to this type of specialty care. Therefore, the identification of elements that foster telerehabilitation growth in future investigations can assist policy makers and key stakeholders in optimally leveraging program resources for maximal productivity. Future studies investigating the drivers of increases in telerehabilitation growth by rurality are warranted. Furthermore, more research is needed to examine telerehabilitation growth quality of care outcomes (eg, patient and provider satisfaction) to ensure that care is not only timely and accessible, but of high quality.

 

Conclusion

Disparities between rural and urban veterans compel a mode of expanding delivery of care. The VHA has embraced the use of telehealth modalities to extend its reach of rehabilitation services to veterans with disability and rehabilitation needs. Growth in telerehabilitation rural patient encounters increases access to rehabilitative care, reduces patient and caregiver travel burden, and helps ensure treatment adherence. Telerehabilitation utilization (unique patients and total encounters) is growing more rapidly for rural veterans than for their urban counterparts. Overall, telerehabilitation is filling a gap for rural veterans, as well as veterans in general with challenges in accessibility to health care. In order to make full use of the telerehabilitation services across its health care system, VA health care facilities may need to expand their effort in telerehabilitation dissemination and education among providers and veterans, particularly among providers who are less familiar with telerehabilitation services and among veterans who live in rural or highly rural areas and need special rehabilitation care.

Historically, the Veterans Health Administration (VHA) has excelled at improving veterans’ access to health care and enhancing foundational services, such as prosthetics and other veteran-centric services, and this continues to be the VHA’s top priority.¹ Travel distance and time are often barriers to accessing health care for many veterans.^2-11 For veterans with disabilities who must overcome additional physical, cognitive, and emotional obstacles to access vital rehabilitation services, these geographic obstacles are magnified. Further compounding the challenge is that rehabilitation therapies frequently require multiple encounters. Telerehabilitation is a promising solution for veterans in need of rehabilitation to regain optimal functioning. This alternative mode of service delivery can help veterans overcome geographic access barriers by delivering health care directly to veterans in their homes or nearby community-based outpatient clinics.^12,13

A growing body of evidence supports telerehabilitation. In a 2017 systematic review and meta-analysis, Cottrell and colleagues reviewed and analyzed data from 13 studies that met their inclusion criteria; specifically, their meta-analytic sample comprised adults aged ≥ 18 years presenting with any diagnosed primary musculoskeletal condition; treatment interventions via a real-time telerehabilitation medium, trials that had a comparison group with the same condition; provided clinical outcomes data, and included published randomized and nonrandomized controlled trials.¹⁴ Based on their aggregated results, they concluded that real-time telerehabilitation was effective in improving physical function (standardized mean difference [SMD], 0.63; 95% CI, 0.92-2.33; I², 93%), and reducing pain (SMD, 0.66; 95% CI, −0.27- .60; I², 96%) in patients with any diagnosed primary musculoskeletal condition.¹⁴

Two other systematic reviews conducted by Pietrzak and colleagues and Agostini and colleagues also demonstrated the clinical effectiveness of telerehabilitation.^15,16 Clinical effectiveness was defined as changes in health, functional status, and satisfaction with the telerehabilitation services delivered. The studies examined in the review included those that provided online self-management and education in addition to exercise via teleconferencing in real time.

Pietrzak and colleagues found that Internet-based osteoarthritis self-management interventions significantly improved 4 of 6 health status measures reviewed (ie, pain, fatigue, activity limitation, health distress, disability, and self‐reported global health).¹⁵ User acceptance and satisfaction were high (≥ 70% satisfied) in all studies meeting the inclusion criteria.

Agostini and colleagues found that telerehabilitation was more effective than other modes of delivering rehabilitation to regain motor function in cardiac (SMD, 0.24; 95% CI, 0.04-0.43) and total knee arthroplasty (Timed Up and Go test: SMD, −5.17; 95% CI, −9.79- −0.55) patients.¹⁶ Some evidence from VHA and non-VHA studies also support the use of telerehabilitation to reduce health care costs,^17-19 improve treatment adherence,^12,20 and enhance patient physical, cognitive and mobility function, as well as patient satisfaction and health-related quality of life.^13,21-24

Since the first recorded use of telehealth in 1959, the application of technology to deliver health care, including rehabilitation services, has increased exponentially.¹⁴ In fiscal year (FY) 2017 alone, the VA provided > 2 million episodes of care for > 700,000 veterans using telehealth services.²⁵

Although the process for accessing telerehabilitation may vary throughout the VA, typically a few common factors make a veteran eligible for this mode of rehabilitation care delivery: Veterans must meet criteria for a specific program (eg, amputation, occupational therapy, and physical therapy) and receive VA care from a VA medical facility or clinic that offers telehealth services. Care providers must believe that the veteran would benefit from telerehabilitation (eg, limited mobility and long-distance travel to the facility) and that they would be able to receive an appropriate consult. The veteran must meet the following requirements: (1) willingness to consent to a visit via telehealth; (2) access to required equipment/e-mail; and (3) a caregiver to assist if they are unable to complete a visit independently.

In this article, we provide an overview of the growth of telerehabilitation in the VHA. Data are presented for specific telerehabilitation programs over time and by rurality.

 

Methods

The VHA Support Service Center works with VHA program offices and field users to provide field-focused business, clinical, and special topic reports. An online portal provides access to these customizable reports organized as data cubes, which represent data dimensions (ie, clinic type) and measures (ie, number of unique patients). For this study, we used the Connected Care, Telehealth, Call Centers Clinical Video Telehealth/Store and Forward Telehealth data cube clinical stop codes to identify the numbers of telerehabilitation veteran users and encounters across time. The following telerehabilitation clinic-stop codes were selected: 197 (polytrauma/traumatic brain injury [TBI]–individuals), 201 (Physical Medicine and Rehabilitation [PM&R] Service), 205 (physical therapy), 206 (occupational therapy), 211 (PM&R amputation clinic), 418 (amputation clinic), 214 (kinesiotherapy), and 240 (PM&R assistive technology clinic). Data for total unique patients served and the total number of encounters were extracted at the national level and by rurality from FY 2012 to FY 2017, providing the past 5 years of VHA telerehabilitation data.

It is important to note that in FY 2015, the VHA changed its definition of rurality to a rural-urban commuting areas (RUCA)-based system (www.ruralhealth.va.gov/rural-definition.asp). Prior to FY 2015, the VHA used the US Census Bureau (CB) urbanized area definitions. According to CB, an urbanized area contains a central city and surrounding area that totals > 50,000 in population. It also includes places outside of urbanized areas with populations > 2,500. Rural areas are defined as all other areas. VHA added a third category, highly rural, which is defined as areas that had < 7 people per square mile. In the RUCA system, each census tract defined by the CB is given a score. The VHA definitions are as follows:

• Urban (U)—census tracts with RUCA scores of 1.0 or 1.1. These tracts are determined by the CB as being in an urban core and having the majority of their workers commute within that same core (1.0). If 30% to 49% commute to an even larger urban core, then the code is 1.1;
• Rural (R)—all tracts not receiving scores in the urban or highly rural tiers; and
• Highly rural (H)—tracts with a RUCA score of 10.0. These are the most remote occupied land areas. Less than 10% of workers travel to CB-defined urbanized areas or urban clusters.

In addition, VHA recently added an “I” category to complement “U,” “R,” and “H.” The “I” value is assigned to veterans living on the US insular islands (ie, territories): Guam, American Samoa, Northern Marianas, and US Virgin Islands. For the analysis by rurality in this study, we excluded veterans living in the insular islands and those of unknown rurality (< 1.0% of patients and encounters). Further, because the numbers of highly rural veterans were relatively small (< 2% of patients and encounters), the rural and highly rural categories were combined and compared with urban-dwelling veterans.

Results

Overall, the workload for telerehabilitation nearly quadrupled over the 5-year period (Table 1 and Figure 1). 

In FY 2012, there were 4,397 unique individuals receiving telerehabilitation in the selected telerehabilitation clinics. By FY 2017, this number had grown to 16,319 veterans. 

Similar increases were seen for total encounters, growing from 6,643 in FY 2012 to 22,179 in FY 2017 (Figure 2). The rate of the increase for the number of unique patients seen and telerehabilitation encounter totals across years were higher from FY 2012 to FY 2015 than from FY 2015 to FY 2017.

 

Interesting trends were seen by clinic type. Some clinics increased substantially, whereas others showed only moderate increases, and in 1 case (PM&R Service), a decrease. For example, there is significant growth in the number of patients and encounters involving physical therapy through telerehabilitation. This telerehabilitation clinic increased its workload from 1,676 patients with 3,016 encounters in FY 2012 to 9,136 patients with 11,834 encounters in FY 2017, accounting for 62.6% of total growth in patients and 56.8% of total growth in encounters.

Other clinics showing substantial growth over time included occupational therapy and polytrauma/TBI-individual secondary evaluation. Kinesiotherapy telerehabilitation was almost nonexistent in the VHA during FY 2012, with only 23 patients having 23 encounters. By FY 2017, there were 563 patients with 624 kinesiotherapy telerehabilitation encounters, equating to staggering increases in 5 years: 2,348% for patients and 2,613% for encounters. Similarly, the Physical Medicine and Rehabilitation Assistive Technology clinics had very low numbers in FY 2012 (patients, 2; encounters, 3) and increased over time; albeit, at a slow rate.

Trends by Rurality

Trends by rural location of patients and encounters must be interpreted with caution because of the changing rural definition between FY 2014 and FY 2015 (Tables 2 and 3; Figures 3 and 4). 

Nevertheless, the number of veterans seen and encounters performed via telerehabilitation increased in both urban and rural settings during the time under investigation. 

Under both the legacy and RUCA definitions of rural, the percentage increase was greater for rural veterans than that for urban veterans.

The increased total number of patients seen between FY 2012 and FY 2014 (old definition) was 225% for rural veterans vs 134% for urban veterans. Between FY 2015 and FY 2017 (new definition), the increase was lower for both groups (rural, 13.4%; urban, 7.3%), but rural veterans still increased at a higher rate than did urban dwellers.

Discussion

Our primary aim was to provide data on the growth of telerehabilitation in the VHA over the past 5 years. Our secondary aim was to examine growth in the use of telerehabilitation by rurality. Specifically, we provided an overview of telerehabilitation growth in terms of unique patients and overall encounters in the VHA by rurality from FY 2012 to FY 2014 and FY 2015 to FY 2017 using the following programs: Polytrauma/TBI, PM&R Service, physical therapy, occupational therapy, PM&R amputation clinic, amputation clinic, kinesiotherapy, and PM&R assistive technology clinic. Our findings demonstrated a noteworthy increase in telerehabilitation encounters and unique patients over time for these programs. These findings were consistent with the overall trend of continued growth and expansion of telehealth within the VHA.

Our findings reveal an upward trend in the total number of rural encounters and rural unique patients despite the change in the VA’s definition of rurality in FY 2015. To our knowledge, urban and rural use of telerehabilitation has not been examined previously. Under both definitions of rurality, encounters and unique patients show an important increase over time, and by year-end 2017, more than half of all patients and encounters were attributed to rural patients (53.7% and 53.9%, respectively). Indeed, the upward trend may have been more pronounced if the rural definition had not changed in FY 2015. Our early VHA stroke patients study on the difference between rural-urban patients and taxonomies showed that the RUCA definition was more likely to reduce the number of rural patients by 8.5% than the early definition used by the VHA.²⁶

It is notable that although the use of tele-delivery of rehabilitation has continually increased, the rate of this increase was steeper from FY 2012 to FY 2014 than FY 2015 to FY 2017. For the programs under consideration in this study, the total number of rural patients/encounters increased throughout the observed periods. However, urban patients and encounters increased through FY 2016 and experienced a slight decrease in FY 2017.

The appearance of a slower rate of increase may be due to a rapid initial rate of increase through early adopters and “crossing the diffusion chasm,” a well-documented process of slower diffusion between the time of invention to penetration that often characterizes the spread of successful telehealth innovations.²⁷ Integrating technology into care delivery innovation requires the integration of technical, clinical, and administrative processes and can take time to scale successfully.²⁸

With an emphasis on increasing access to rehabilitation services, the VHA can expect to see a continuing increase in both the number and the percentage of telerehabilitation rural patients and encounters. The VHA has several telerehabilitation initiatives underway through the VHA’s Physical Medicine and Rehabilitation Telerehabilitation Enterprise Wide Initiative (TREWI) and Rural Veterans Telerehabilitation Initiative. These projects demonstrate the feasibility of this delivery approach and facilitate integration of this modality in clinical workflows. However, to sustain these efforts, facilities will need more infrastructure and personnel resources dedicated to the delivery of services.

In an ongoing evaluation of the TREWI, several factors seem to influence the uptake of the VHA Office of Rural Health TREWI programs. These factors are the presence or absence of a local site champion; the quality of hospital leadership support; the quality of past relationships between telerehabilitation sending sites and receiving sites; barriers to getting a telehealth service agreement in place; the availability of space; administrative know-how on setting up clinics appropriately; time involved to bring on staff; contracting issues; equipment availability and installation; cultural issues in embracing technologic innovation; training burden; hassle factors; and limited funds. Although early adopters may be able to negotiate and push through many of the barriers associated with the diffusion of telerehabilitation, the numerous barriers may slow its larger systemwide diffusion.

Telerehabilitation is a promising mode to deliver care to rural veterans who otherwise may not have access to this type of specialty care. Therefore, the identification of elements that foster telerehabilitation growth in future investigations can assist policy makers and key stakeholders in optimally leveraging program resources for maximal productivity. Future studies investigating the drivers of increases in telerehabilitation growth by rurality are warranted. Furthermore, more research is needed to examine telerehabilitation growth quality of care outcomes (eg, patient and provider satisfaction) to ensure that care is not only timely and accessible, but of high quality.

 

Conclusion

Disparities between rural and urban veterans compel a mode of expanding delivery of care. The VHA has embraced the use of telehealth modalities to extend its reach of rehabilitation services to veterans with disability and rehabilitation needs. Growth in telerehabilitation rural patient encounters increases access to rehabilitative care, reduces patient and caregiver travel burden, and helps ensure treatment adherence. Telerehabilitation utilization (unique patients and total encounters) is growing more rapidly for rural veterans than for their urban counterparts. Overall, telerehabilitation is filling a gap for rural veterans, as well as veterans in general with challenges in accessibility to health care. In order to make full use of the telerehabilitation services across its health care system, VA health care facilities may need to expand their effort in telerehabilitation dissemination and education among providers and veterans, particularly among providers who are less familiar with telerehabilitation services and among veterans who live in rural or highly rural areas and need special rehabilitation care.

Historically, the Veterans Health Administration (VHA) has excelled at improving veterans’ access to health care and enhancing foundational services, such as prosthetics and other veteran-centric services, and this continues to be the VHA’s top priority.¹ Travel distance and time are often barriers to accessing health care for many veterans.^2-11 For veterans with disabilities who must overcome additional physical, cognitive, and emotional obstacles to access vital rehabilitation services, these geographic obstacles are magnified. Further compounding the challenge is that rehabilitation therapies frequently require multiple encounters. Telerehabilitation is a promising solution for veterans in need of rehabilitation to regain optimal functioning. This alternative mode of service delivery can help veterans overcome geographic access barriers by delivering health care directly to veterans in their homes or nearby community-based outpatient clinics.^12,13

A growing body of evidence supports telerehabilitation. In a 2017 systematic review and meta-analysis, Cottrell and colleagues reviewed and analyzed data from 13 studies that met their inclusion criteria; specifically, their meta-analytic sample comprised adults aged ≥ 18 years presenting with any diagnosed primary musculoskeletal condition; treatment interventions via a real-time telerehabilitation medium, trials that had a comparison group with the same condition; provided clinical outcomes data, and included published randomized and nonrandomized controlled trials.¹⁴ Based on their aggregated results, they concluded that real-time telerehabilitation was effective in improving physical function (standardized mean difference [SMD], 0.63; 95% CI, 0.92-2.33; I², 93%), and reducing pain (SMD, 0.66; 95% CI, −0.27- .60; I², 96%) in patients with any diagnosed primary musculoskeletal condition.¹⁴

Two other systematic reviews conducted by Pietrzak and colleagues and Agostini and colleagues also demonstrated the clinical effectiveness of telerehabilitation.^15,16 Clinical effectiveness was defined as changes in health, functional status, and satisfaction with the telerehabilitation services delivered. The studies examined in the review included those that provided online self-management and education in addition to exercise via teleconferencing in real time.

Pietrzak and colleagues found that Internet-based osteoarthritis self-management interventions significantly improved 4 of 6 health status measures reviewed (ie, pain, fatigue, activity limitation, health distress, disability, and self‐reported global health).¹⁵ User acceptance and satisfaction were high (≥ 70% satisfied) in all studies meeting the inclusion criteria.

Agostini and colleagues found that telerehabilitation was more effective than other modes of delivering rehabilitation to regain motor function in cardiac (SMD, 0.24; 95% CI, 0.04-0.43) and total knee arthroplasty (Timed Up and Go test: SMD, −5.17; 95% CI, −9.79- −0.55) patients.¹⁶ Some evidence from VHA and non-VHA studies also support the use of telerehabilitation to reduce health care costs,^17-19 improve treatment adherence,^12,20 and enhance patient physical, cognitive and mobility function, as well as patient satisfaction and health-related quality of life.^13,21-24

Since the first recorded use of telehealth in 1959, the application of technology to deliver health care, including rehabilitation services, has increased exponentially.¹⁴ In fiscal year (FY) 2017 alone, the VA provided > 2 million episodes of care for > 700,000 veterans using telehealth services.²⁵

Although the process for accessing telerehabilitation may vary throughout the VA, typically a few common factors make a veteran eligible for this mode of rehabilitation care delivery: Veterans must meet criteria for a specific program (eg, amputation, occupational therapy, and physical therapy) and receive VA care from a VA medical facility or clinic that offers telehealth services. Care providers must believe that the veteran would benefit from telerehabilitation (eg, limited mobility and long-distance travel to the facility) and that they would be able to receive an appropriate consult. The veteran must meet the following requirements: (1) willingness to consent to a visit via telehealth; (2) access to required equipment/e-mail; and (3) a caregiver to assist if they are unable to complete a visit independently.

In this article, we provide an overview of the growth of telerehabilitation in the VHA. Data are presented for specific telerehabilitation programs over time and by rurality.

 

Methods

The VHA Support Service Center works with VHA program offices and field users to provide field-focused business, clinical, and special topic reports. An online portal provides access to these customizable reports organized as data cubes, which represent data dimensions (ie, clinic type) and measures (ie, number of unique patients). For this study, we used the Connected Care, Telehealth, Call Centers Clinical Video Telehealth/Store and Forward Telehealth data cube clinical stop codes to identify the numbers of telerehabilitation veteran users and encounters across time. The following telerehabilitation clinic-stop codes were selected: 197 (polytrauma/traumatic brain injury [TBI]–individuals), 201 (Physical Medicine and Rehabilitation [PM&R] Service), 205 (physical therapy), 206 (occupational therapy), 211 (PM&R amputation clinic), 418 (amputation clinic), 214 (kinesiotherapy), and 240 (PM&R assistive technology clinic). Data for total unique patients served and the total number of encounters were extracted at the national level and by rurality from FY 2012 to FY 2017, providing the past 5 years of VHA telerehabilitation data.

It is important to note that in FY 2015, the VHA changed its definition of rurality to a rural-urban commuting areas (RUCA)-based system (www.ruralhealth.va.gov/rural-definition.asp). Prior to FY 2015, the VHA used the US Census Bureau (CB) urbanized area definitions. According to CB, an urbanized area contains a central city and surrounding area that totals > 50,000 in population. It also includes places outside of urbanized areas with populations > 2,500. Rural areas are defined as all other areas. VHA added a third category, highly rural, which is defined as areas that had < 7 people per square mile. In the RUCA system, each census tract defined by the CB is given a score. The VHA definitions are as follows:

• Urban (U)—census tracts with RUCA scores of 1.0 or 1.1. These tracts are determined by the CB as being in an urban core and having the majority of their workers commute within that same core (1.0). If 30% to 49% commute to an even larger urban core, then the code is 1.1;
• Rural (R)—all tracts not receiving scores in the urban or highly rural tiers; and
• Highly rural (H)—tracts with a RUCA score of 10.0. These are the most remote occupied land areas. Less than 10% of workers travel to CB-defined urbanized areas or urban clusters.

In addition, VHA recently added an “I” category to complement “U,” “R,” and “H.” The “I” value is assigned to veterans living on the US insular islands (ie, territories): Guam, American Samoa, Northern Marianas, and US Virgin Islands. For the analysis by rurality in this study, we excluded veterans living in the insular islands and those of unknown rurality (< 1.0% of patients and encounters). Further, because the numbers of highly rural veterans were relatively small (< 2% of patients and encounters), the rural and highly rural categories were combined and compared with urban-dwelling veterans.

Results

Overall, the workload for telerehabilitation nearly quadrupled over the 5-year period (Table 1 and Figure 1). 

In FY 2012, there were 4,397 unique individuals receiving telerehabilitation in the selected telerehabilitation clinics. By FY 2017, this number had grown to 16,319 veterans. 

Similar increases were seen for total encounters, growing from 6,643 in FY 2012 to 22,179 in FY 2017 (Figure 2). The rate of the increase for the number of unique patients seen and telerehabilitation encounter totals across years were higher from FY 2012 to FY 2015 than from FY 2015 to FY 2017.

 

Interesting trends were seen by clinic type. Some clinics increased substantially, whereas others showed only moderate increases, and in 1 case (PM&R Service), a decrease. For example, there is significant growth in the number of patients and encounters involving physical therapy through telerehabilitation. This telerehabilitation clinic increased its workload from 1,676 patients with 3,016 encounters in FY 2012 to 9,136 patients with 11,834 encounters in FY 2017, accounting for 62.6% of total growth in patients and 56.8% of total growth in encounters.

Other clinics showing substantial growth over time included occupational therapy and polytrauma/TBI-individual secondary evaluation. Kinesiotherapy telerehabilitation was almost nonexistent in the VHA during FY 2012, with only 23 patients having 23 encounters. By FY 2017, there were 563 patients with 624 kinesiotherapy telerehabilitation encounters, equating to staggering increases in 5 years: 2,348% for patients and 2,613% for encounters. Similarly, the Physical Medicine and Rehabilitation Assistive Technology clinics had very low numbers in FY 2012 (patients, 2; encounters, 3) and increased over time; albeit, at a slow rate.

Trends by Rurality

Trends by rural location of patients and encounters must be interpreted with caution because of the changing rural definition between FY 2014 and FY 2015 (Tables 2 and 3; Figures 3 and 4). 

Nevertheless, the number of veterans seen and encounters performed via telerehabilitation increased in both urban and rural settings during the time under investigation. 

Under both the legacy and RUCA definitions of rural, the percentage increase was greater for rural veterans than that for urban veterans.

The increased total number of patients seen between FY 2012 and FY 2014 (old definition) was 225% for rural veterans vs 134% for urban veterans. Between FY 2015 and FY 2017 (new definition), the increase was lower for both groups (rural, 13.4%; urban, 7.3%), but rural veterans still increased at a higher rate than did urban dwellers.

Discussion

Our primary aim was to provide data on the growth of telerehabilitation in the VHA over the past 5 years. Our secondary aim was to examine growth in the use of telerehabilitation by rurality. Specifically, we provided an overview of telerehabilitation growth in terms of unique patients and overall encounters in the VHA by rurality from FY 2012 to FY 2014 and FY 2015 to FY 2017 using the following programs: Polytrauma/TBI, PM&R Service, physical therapy, occupational therapy, PM&R amputation clinic, amputation clinic, kinesiotherapy, and PM&R assistive technology clinic. Our findings demonstrated a noteworthy increase in telerehabilitation encounters and unique patients over time for these programs. These findings were consistent with the overall trend of continued growth and expansion of telehealth within the VHA.

Our findings reveal an upward trend in the total number of rural encounters and rural unique patients despite the change in the VA’s definition of rurality in FY 2015. To our knowledge, urban and rural use of telerehabilitation has not been examined previously. Under both definitions of rurality, encounters and unique patients show an important increase over time, and by year-end 2017, more than half of all patients and encounters were attributed to rural patients (53.7% and 53.9%, respectively). Indeed, the upward trend may have been more pronounced if the rural definition had not changed in FY 2015. Our early VHA stroke patients study on the difference between rural-urban patients and taxonomies showed that the RUCA definition was more likely to reduce the number of rural patients by 8.5% than the early definition used by the VHA.²⁶

It is notable that although the use of tele-delivery of rehabilitation has continually increased, the rate of this increase was steeper from FY 2012 to FY 2014 than FY 2015 to FY 2017. For the programs under consideration in this study, the total number of rural patients/encounters increased throughout the observed periods. However, urban patients and encounters increased through FY 2016 and experienced a slight decrease in FY 2017.

The appearance of a slower rate of increase may be due to a rapid initial rate of increase through early adopters and “crossing the diffusion chasm,” a well-documented process of slower diffusion between the time of invention to penetration that often characterizes the spread of successful telehealth innovations.²⁷ Integrating technology into care delivery innovation requires the integration of technical, clinical, and administrative processes and can take time to scale successfully.²⁸

With an emphasis on increasing access to rehabilitation services, the VHA can expect to see a continuing increase in both the number and the percentage of telerehabilitation rural patients and encounters. The VHA has several telerehabilitation initiatives underway through the VHA’s Physical Medicine and Rehabilitation Telerehabilitation Enterprise Wide Initiative (TREWI) and Rural Veterans Telerehabilitation Initiative. These projects demonstrate the feasibility of this delivery approach and facilitate integration of this modality in clinical workflows. However, to sustain these efforts, facilities will need more infrastructure and personnel resources dedicated to the delivery of services.

In an ongoing evaluation of the TREWI, several factors seem to influence the uptake of the VHA Office of Rural Health TREWI programs. These factors are the presence or absence of a local site champion; the quality of hospital leadership support; the quality of past relationships between telerehabilitation sending sites and receiving sites; barriers to getting a telehealth service agreement in place; the availability of space; administrative know-how on setting up clinics appropriately; time involved to bring on staff; contracting issues; equipment availability and installation; cultural issues in embracing technologic innovation; training burden; hassle factors; and limited funds. Although early adopters may be able to negotiate and push through many of the barriers associated with the diffusion of telerehabilitation, the numerous barriers may slow its larger systemwide diffusion.

Telerehabilitation is a promising mode to deliver care to rural veterans who otherwise may not have access to this type of specialty care. Therefore, the identification of elements that foster telerehabilitation growth in future investigations can assist policy makers and key stakeholders in optimally leveraging program resources for maximal productivity. Future studies investigating the drivers of increases in telerehabilitation growth by rurality are warranted. Furthermore, more research is needed to examine telerehabilitation growth quality of care outcomes (eg, patient and provider satisfaction) to ensure that care is not only timely and accessible, but of high quality.

 

Conclusion

Disparities between rural and urban veterans compel a mode of expanding delivery of care. The VHA has embraced the use of telehealth modalities to extend its reach of rehabilitation services to veterans with disability and rehabilitation needs. Growth in telerehabilitation rural patient encounters increases access to rehabilitative care, reduces patient and caregiver travel burden, and helps ensure treatment adherence. Telerehabilitation utilization (unique patients and total encounters) is growing more rapidly for rural veterans than for their urban counterparts. Overall, telerehabilitation is filling a gap for rural veterans, as well as veterans in general with challenges in accessibility to health care. In order to make full use of the telerehabilitation services across its health care system, VA health care facilities may need to expand their effort in telerehabilitation dissemination and education among providers and veterans, particularly among providers who are less familiar with telerehabilitation services and among veterans who live in rural or highly rural areas and need special rehabilitation care.

Knee osteoarthritis (OA) affects almost 9.3 million adults in the US and accounts for $27 billion in annual health care expenses.^1,2 Due to the increasing cost of health care and an aging population, there has been renewed interest in establishing criteria for nonarthroplasty treatment of knee OA.

In 2013, using the RAND/UCLA Appropriateness method, the American Academy of Orthopaedic Surgeons (AAOS) developed an appropriate use criteria (AUC) for nonarthroplasty management of primary OA of the knee, based on orthopaedic literature and expert opinion.³ Interventions such as activity modification, weight loss, prescribed physical therapy, nonsteroidal anti-inflammatory drugs, tramadol, prescribed oral or transcutaneous opioids, acetaminophen, intra-articular corticosteroids, hinged or unloading knee braces, arthroscopic partial menisectomy or loose body removal, and realignment osteotomy were assessed. An algorithm was developed for 576 patients scenarios that incorporated patient-specific, prognostic/predictor variables to assign designations of “appropriate,” “may be appropriate,” or “rarely appropriate,” to treatment interventions.^4,5 An online version of the algorithm (orthoguidelines.org) is available for physicians and surgeons to judge appropriateness of nonarthroplasty treatments; however, it is not intended to mandate candidacy for treatment or intervention.

Clinical evaluation of the AAOS AUC is necessary to determine how treatment recommendations correlate with current practice. A recent examination of the AAOS Appropriateness System for Surgical Management of Knee OA found that prognostic/predictor variables, such as patient age, OA severity, and pattern of knee OA involvement were more heavily weighted when determining arthroplasty appropriateness than was pain severity or functional loss.⁶ Furthermore, non-AAOS AUC prognostic/predictor variables, such as race and gender, have been linked to disparities in utilization of knee OA interventions.^7-9 Such disparities can be costly not just from a patient perceptive, but also employer and societal perspectives.¹⁰

The Department of Veterans Affairs (VA) health care system represents a model of equal-access-to care system in the US that is ideal for examination of issues about health care utilization and any disparities within the AAOS AUC model and has previously been used to assess utilization of total knee arthroplasty.⁹ The aim of this study was to characterize utilization of the AAOS AUC for nonarthroplasty treatment of knee OA in a VA patient population. We asked the following questions: (1) What variables are predictive of receiving a greater number of AAOS AUC evaluated nonarthroplasty treatments? (2) What variables are predictive of receiving “rarely appropriate” AAOS AUC evaluated nonarthroplasty treatment? (3) What factors are predictive of duration of nonarthroplasty care until total knee arthroplasty (TKA)?

Methods

The institutional review board at the Louis Stokes Cleveland VA Medical Center in Ohio approved a retrospective chart review of nonarthroplasty treatments utilized by patients presenting to its orthopaedic section who subsequently underwent knee arthroplasty between 2013 and 2016. Eligibility criteria included patients aged ≥ 30 years with a diagnosis of unilateral or bilateral primary knee OA. Patients with posttraumatic OA, inflammatory arthritis, and a history of infectious arthritis or Charcot arthropathy of the knee were excluded. Patients with a body mass index (BMI) > 40 or a hemoglobin A_1c > 8.0 at presentation were excluded as nonarthroplasty care was the recommended course of treatment above these thresholds.

 

Data collected included race, gender, duration of nonarthroplasty treatment, BMI, and Kellgren-Lawrence classification of knee OA at time of presentation for symptomatic knee OA.¹¹ All AAOS AUC-evaluated nonarthroplasty treatments utilized prior to arthroplasty intervention also were recorded (Table 1). 

Indications and classifications for each subject were entered into the AAOS AUC online algorithm, and every AAOS AUC evaluated treatment utilized was assigned a rating of appropriate, may be appropriate, or rarely appropriate, based on the algorithm results for that clinical scenario (Table 2). 

Information regarding anti-inflammatory, analgesic, and prescribed oral or transcutaneous opioid use for chronic knee pain during the period of nonoperative management of knee OA prior to TKA was obtained by review of medication lists and reconciliation with orthopaedic consultation notes in the electronic health record. Peri-operative anti-inflammatory, analgesic, and prescribed oral or transcutaneous opioid use did not constitute an AUC intervention.

Statistical Analysis

Statistical analysis was completed with GraphPad Software Prism 7.0a (La Jolla, CA) and Mathworks MatLab R2016b software (Natick, MA). Univariate analysis with Student t tests with Welch corrections in the setting of unequal variance, Mann-Whitney nonparametric tests, and Fisher exact test were generated in the appropriate setting. Multivariable analyses also were conducted. For continuous outcomes, stepwise multiple linear regression was used to generate predictive models; for binary outcomes, binomial logistic regression was used.

Factors analyzed in regression modeling for the total number of AAOS AUC evaluated nonarthroplasty treatments utilized and the likelihood of receiving a rarely appropriate treatment included gender, race, function-limiting pain, range of motion (ROM), ligamentous instability, arthritis pattern, limb alignment, mechanical symptoms, BMI, age, and Kellgren-Lawrence grade. Factors analyzed in timing of TKA included the above variables plus the total number of AUC interventions, whether the patient received an inappropriate intervention, and average appropriateness of the interventions received. Residual analysis with Cook’s distance was used to identify outliers in regression. Observations with Cook’s distance > 3 times the mean Cook’s distance were identified as potential outliers, and models were adjusted accordingly. All statistical analyses were 2-tailed. Statistical significance was set to P ≤ .05 for all outputs.

Results

In the study, 97.8% of participants identified as male, and the mean age was 62.8 years (Table 3). 

The study group was predominantly white (70.3%). All participants had a diagnosis of primary OA. The majority of patients were aged 51 to 70 years (68.1%) and presented with pain occurring following short-distance ambulation (79.1%) but without mechanical symptoms (80.2%). On examination, the majority of patients were found to have full knee ROM (53.8%), no ligamentous instability (97.8%), and normal limb alignment (60.4%). Radiographically, patients most often had multicompartmental disease (69.2%) with evidence of severe joint-space narrowing (63.7%), resulting in a plurality of patients having a Kellgren-Lawrence arthritis grade of 3 (46.2%) (Table 4).

Appropriate Use Criteria Interventions

Patients received a mean of 5.2 AAOS AUC evaluated interventions before undergoing arthroplasty management at a mean of 32.3 months (range 2-181 months) from initial presentation. The majority of these interventions were classified as either appropriate or may be appropriate, according to the AUC definitions (95.1%). Self-management and physical therapy programs were widely utilized (100% and 90.1%, respectively), with all use of these interventions classified as appropriate.

 

Hinged or unloader knee braces were utilized in about half the study patients; this intervention was classified as rarely appropriate in 4.4% of these patients. Medical therapy was also widely used, with all use of NSAIDs, acetaminophen, and tramadol classified as appropriate or may be appropriate. Oral or transcutaneous opioid medications were prescribed in 14.3% of patients, with 92.3% of this use classified as rarely appropriate. Although the opioid medication prescribing provider was not specifically evaluated, there were no instances in which the orthopaedic service provided an oral or transcutaneous opioid prescriptions. Procedural interventions, with the exception of corticosteroid injections, were uncommon; no patient received realignment osteotomy, and only 12.1% of patients underwent arthroscopy. The use of arthroscopy was deemed rarely appropriate in 72.7% of these cases.

Factors Associated With AAOS AUC Intervention Use

There was no difference in the number of AAOS AUC evaluated interventions received based on BMI (mean [SD] BMI < 35, 5.2 [1.0] vs BMI ≥ 35, 5.3 [1.1], P = .49), age (mean [SD] aged < 60 years, 5.4 [1.0] vs aged ≥ 60 years, 5.1 [1.2], P = .23), or Kellgren-Lawrence arthritic grade (mean [SD] grade ≤ 2, 5.5 [1.0] vs grade > 2, 5.1 [1.1], P = .06). These variables also were not associated with receiving a rarely appropriate intervention (mean [SD] BMI < 35, 0.27 [0.5] vs BMI > 35, 0.2 [0.4], P = .81; aged > 60 years, 0.3 [0.5] vs aged < 60 years, 0.2 [0.4], P = .26; Kellgren-Lawrence grade < 2, 0.4 [0.6] vs grade > 2, 0.2 [0.4], P = .1).

Regression modeling to predict total number of AAOS AUC evaluated interventions received produced a significant model (R² = 0.111, P = .006). The presence of ligamentous instability (β coefficient, -1.61) and the absence of mechanical symptoms (β coefficient, -0.67) were negative predictors of number of AUC interventions received. Variance inflation factors were 1.014 and 1.012, respectively. Likewise, regression modeling to identify factors predictive of receiving a rarely appropriate intervention also produced a significant model (pseudo R²= 0.06, P = .025), with lower Kellgren-Lawrence grade the only significant predictor of receiving a rarely appropriate intervention (odds ratio [OR] 0.54; 95% CI, 0.42 -0.72, per unit increase).

Timing from presentation to arthroplasty intervention was also evaluated. Age was a negative predictor (β coefficient -1.61), while positive predictors were reduced ROM (β coefficient 15.72) and having more AUC interventions (β coefficient 7.31) (model R²= 0.29, P = < .001). Age was the most significant predictor. Variance inflations factors were 1.02, 1.01, and 1.03, respectively. Receiving a rarely appropriate intervention was not associated with TKA timing.

Discussion

This single-center retrospective study examined the utilization of AAOS AUC-evaluated nonarthroplasty interventions for symptomatic knee OA prior to TKA. The aims of this study were to validate the AAOS AUC in a clinical setting and identify predictors of AAOS AUC utilization. In particular, this study focused on the number of interventions utilized prior to knee arthroplasty, whether interventions receiving a designation of rarely appropriate were used, and the duration of nonarthroplasty treatment.

 

Patients with knee instability used fewer total AAOS AUC evaluated interventions prior to TKA. Subjective instability has been reported as high as 27% in patients with OA and has been associated with fear of falling, poor balance confidence, activity limitations, and lower Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) physical function scores.¹² However, it has not been found to correlate with knee laxity.¹³ Nevertheless, significant functional impairment with the risk of falling may reduce the number of nonarthroplasty interventions attempted. On the other hand, the presence of mechanical symptoms resulted in greater utilization of nonarthroplasty interventions. This is likely due to the greater utilization of arthroscopic partial menisectomy or loose body removal in this group of patients. Despite its inclusion as an AAOS AUC evaluated intervention, arthroscopy remains a contentious treatment for symptomatic knee pain in the setting of OA.^14,15

For every unit decrease in Kellgren-Lawrence OA grade, patients were 54% more likely to receive a rarely appropriate intervention prior to knee arthroplasty. This is supported by the recent literature examining the AAOS AUC for surgical management of knee OA. Riddle and colleagues developed a classification tree to determine the contributions of various prognostic variables in final classifications of the 864 clinical vignettes used to develop the appropriateness algorithm and found that OA severity was strongly favored, with only 4 of the 432 vignettes with severe knee OA judged as rarely appropriate for surgical intervention.⁶

Our findings, too, may be explained by an AAOS AUC system that too heavily weighs radiographic severity of knee OA, resulting in more frequent rarely appropriate interventions in patients with less severe arthritis, including nonarthroplasty treatments. It is likely that rarely appropriate interventions were attempted in this subset of our study cohort based on patient’s subjective symptoms and functional status, both of which have been shown to be discordant with radiographic severity of knee OA.¹⁶

Oral or transcutaneous prescribed opioid medications were the most frequent intervention that received a rarely appropriate designation. Patients with preoperative opioid use undergoing TKA have been shown to have a greater risk for postoperative complications and longer hospital stay, particularly those patients aged < 75 years. Younger age, use of more interventions, and decreased knee ROM at presentation were predictive of longer duration of nonarthroplasty treatment. The use of more AAOS AUC evaluated interventions in these patients suggests that the AAOS AUC model may effectively be used to manage symptomatic OA, increasing the time from presentation to knee arthroplasty.

Interestingly, the use of rarely appropriate interventions did not affect TKA timing, as would be expected in a clinically effective nonarthroplasty treatment model. The reasons for rarely appropriate nonsurgical interventions are complex and require further investigation. One possible explanation is that decreased ROM was a marker for mechanical symptoms that necessitated additional intervention in the form of knee arthroscopy, delaying time to TKA.

Limitations

There are several limitations of this study. First, the small sample size (N = 90) requires acknowledgment; however, this limitation reflects the difficulty in following patients for years prior to an operative intervention. Second, the study population consists of veterans using the VA system and may not be reflective of the general population, differing with respect to gender, racial, and socioeconomic factors. Nevertheless, studies examining TKA utilization found, aside from racial and ethnic variability, patient gender and age do not affect arthroplasty utilization rate in the VA system.¹⁷

 

Additional limitations stem from the retrospective nature of this study. While the Computerized Patient Record System and centralized care of the VA system allows for review of all physical therapy consultations, orthotic consultations, and medications within the VA system, any treatments and intervention delivered by non-VA providers were not captured. Furthermore, the ability to assess for confounding variables limiting the prescription of certain medications, such as chronic kidney disease with NSAIDs or liver disease with acetaminophen, was limited by our study design.

Although our study suffers from selection bias with respect to examination of nonarthroplasty treatment in patients who have ultimately undergone TKA, we feel that this subset of patients with symptomatic knee OA represents the majority of patients evaluated for knee OA by orthopaedic surgeons in the clinic setting. It should be noted that although realignment osteotomies were sometimes indicated as appropriate by AAOS AUC model in our study population, this intervention was never performed due to patient and surgeon preference. Additionally, although it is not an AAOS AUC evaluated intervention, viscosupplementation was sporadically used during the study period; however, it is now off formulary at the investigation institution.

Conclusion

Our study suggests that patients without knee instability use more nonarthroplasty treatments over a longer period before TKA, and those patients with less severe knee OA are at risk of receiving an intervention judged to be rarely appropriate by the AAOS AUC. Such interventions do not affect timing of TKA. Nonarthroplasty care should be individualized to patients’ needs, and the decision to proceed with arthroplasty should be considered only after exhausting appropriate conservative measures. We recommend that providers use the AAOS AUC, especially when treating younger patients with less severe knee OA, particularly if considering opiate therapy or knee arthroscopy.

Acknowledgments
The authors would like to acknowledge Patrick Getty, MD, for his surgical care of some of the study patients. This material is the result of work supported with resources and the use of facilities at the Louis Stokes Cleveland VA Medical Center in Ohio.

Knee osteoarthritis (OA) affects almost 9.3 million adults in the US and accounts for $27 billion in annual health care expenses.^1,2 Due to the increasing cost of health care and an aging population, there has been renewed interest in establishing criteria for nonarthroplasty treatment of knee OA.

In 2013, using the RAND/UCLA Appropriateness method, the American Academy of Orthopaedic Surgeons (AAOS) developed an appropriate use criteria (AUC) for nonarthroplasty management of primary OA of the knee, based on orthopaedic literature and expert opinion.³ Interventions such as activity modification, weight loss, prescribed physical therapy, nonsteroidal anti-inflammatory drugs, tramadol, prescribed oral or transcutaneous opioids, acetaminophen, intra-articular corticosteroids, hinged or unloading knee braces, arthroscopic partial menisectomy or loose body removal, and realignment osteotomy were assessed. An algorithm was developed for 576 patients scenarios that incorporated patient-specific, prognostic/predictor variables to assign designations of “appropriate,” “may be appropriate,” or “rarely appropriate,” to treatment interventions.^4,5 An online version of the algorithm (orthoguidelines.org) is available for physicians and surgeons to judge appropriateness of nonarthroplasty treatments; however, it is not intended to mandate candidacy for treatment or intervention.

Clinical evaluation of the AAOS AUC is necessary to determine how treatment recommendations correlate with current practice. A recent examination of the AAOS Appropriateness System for Surgical Management of Knee OA found that prognostic/predictor variables, such as patient age, OA severity, and pattern of knee OA involvement were more heavily weighted when determining arthroplasty appropriateness than was pain severity or functional loss.⁶ Furthermore, non-AAOS AUC prognostic/predictor variables, such as race and gender, have been linked to disparities in utilization of knee OA interventions.^7-9 Such disparities can be costly not just from a patient perceptive, but also employer and societal perspectives.¹⁰

The Department of Veterans Affairs (VA) health care system represents a model of equal-access-to care system in the US that is ideal for examination of issues about health care utilization and any disparities within the AAOS AUC model and has previously been used to assess utilization of total knee arthroplasty.⁹ The aim of this study was to characterize utilization of the AAOS AUC for nonarthroplasty treatment of knee OA in a VA patient population. We asked the following questions: (1) What variables are predictive of receiving a greater number of AAOS AUC evaluated nonarthroplasty treatments? (2) What variables are predictive of receiving “rarely appropriate” AAOS AUC evaluated nonarthroplasty treatment? (3) What factors are predictive of duration of nonarthroplasty care until total knee arthroplasty (TKA)?

Methods

The institutional review board at the Louis Stokes Cleveland VA Medical Center in Ohio approved a retrospective chart review of nonarthroplasty treatments utilized by patients presenting to its orthopaedic section who subsequently underwent knee arthroplasty between 2013 and 2016. Eligibility criteria included patients aged ≥ 30 years with a diagnosis of unilateral or bilateral primary knee OA. Patients with posttraumatic OA, inflammatory arthritis, and a history of infectious arthritis or Charcot arthropathy of the knee were excluded. Patients with a body mass index (BMI) > 40 or a hemoglobin A_1c > 8.0 at presentation were excluded as nonarthroplasty care was the recommended course of treatment above these thresholds.

 

Data collected included race, gender, duration of nonarthroplasty treatment, BMI, and Kellgren-Lawrence classification of knee OA at time of presentation for symptomatic knee OA.¹¹ All AAOS AUC-evaluated nonarthroplasty treatments utilized prior to arthroplasty intervention also were recorded (Table 1). 

Indications and classifications for each subject were entered into the AAOS AUC online algorithm, and every AAOS AUC evaluated treatment utilized was assigned a rating of appropriate, may be appropriate, or rarely appropriate, based on the algorithm results for that clinical scenario (Table 2). 

Information regarding anti-inflammatory, analgesic, and prescribed oral or transcutaneous opioid use for chronic knee pain during the period of nonoperative management of knee OA prior to TKA was obtained by review of medication lists and reconciliation with orthopaedic consultation notes in the electronic health record. Peri-operative anti-inflammatory, analgesic, and prescribed oral or transcutaneous opioid use did not constitute an AUC intervention.

Statistical Analysis

Statistical analysis was completed with GraphPad Software Prism 7.0a (La Jolla, CA) and Mathworks MatLab R2016b software (Natick, MA). Univariate analysis with Student t tests with Welch corrections in the setting of unequal variance, Mann-Whitney nonparametric tests, and Fisher exact test were generated in the appropriate setting. Multivariable analyses also were conducted. For continuous outcomes, stepwise multiple linear regression was used to generate predictive models; for binary outcomes, binomial logistic regression was used.

Factors analyzed in regression modeling for the total number of AAOS AUC evaluated nonarthroplasty treatments utilized and the likelihood of receiving a rarely appropriate treatment included gender, race, function-limiting pain, range of motion (ROM), ligamentous instability, arthritis pattern, limb alignment, mechanical symptoms, BMI, age, and Kellgren-Lawrence grade. Factors analyzed in timing of TKA included the above variables plus the total number of AUC interventions, whether the patient received an inappropriate intervention, and average appropriateness of the interventions received. Residual analysis with Cook’s distance was used to identify outliers in regression. Observations with Cook’s distance > 3 times the mean Cook’s distance were identified as potential outliers, and models were adjusted accordingly. All statistical analyses were 2-tailed. Statistical significance was set to P ≤ .05 for all outputs.

Results

In the study, 97.8% of participants identified as male, and the mean age was 62.8 years (Table 3). 

The study group was predominantly white (70.3%). All participants had a diagnosis of primary OA. The majority of patients were aged 51 to 70 years (68.1%) and presented with pain occurring following short-distance ambulation (79.1%) but without mechanical symptoms (80.2%). On examination, the majority of patients were found to have full knee ROM (53.8%), no ligamentous instability (97.8%), and normal limb alignment (60.4%). Radiographically, patients most often had multicompartmental disease (69.2%) with evidence of severe joint-space narrowing (63.7%), resulting in a plurality of patients having a Kellgren-Lawrence arthritis grade of 3 (46.2%) (Table 4).

Appropriate Use Criteria Interventions

Patients received a mean of 5.2 AAOS AUC evaluated interventions before undergoing arthroplasty management at a mean of 32.3 months (range 2-181 months) from initial presentation. The majority of these interventions were classified as either appropriate or may be appropriate, according to the AUC definitions (95.1%). Self-management and physical therapy programs were widely utilized (100% and 90.1%, respectively), with all use of these interventions classified as appropriate.

 

Hinged or unloader knee braces were utilized in about half the study patients; this intervention was classified as rarely appropriate in 4.4% of these patients. Medical therapy was also widely used, with all use of NSAIDs, acetaminophen, and tramadol classified as appropriate or may be appropriate. Oral or transcutaneous opioid medications were prescribed in 14.3% of patients, with 92.3% of this use classified as rarely appropriate. Although the opioid medication prescribing provider was not specifically evaluated, there were no instances in which the orthopaedic service provided an oral or transcutaneous opioid prescriptions. Procedural interventions, with the exception of corticosteroid injections, were uncommon; no patient received realignment osteotomy, and only 12.1% of patients underwent arthroscopy. The use of arthroscopy was deemed rarely appropriate in 72.7% of these cases.

Factors Associated With AAOS AUC Intervention Use

There was no difference in the number of AAOS AUC evaluated interventions received based on BMI (mean [SD] BMI < 35, 5.2 [1.0] vs BMI ≥ 35, 5.3 [1.1], P = .49), age (mean [SD] aged < 60 years, 5.4 [1.0] vs aged ≥ 60 years, 5.1 [1.2], P = .23), or Kellgren-Lawrence arthritic grade (mean [SD] grade ≤ 2, 5.5 [1.0] vs grade > 2, 5.1 [1.1], P = .06). These variables also were not associated with receiving a rarely appropriate intervention (mean [SD] BMI < 35, 0.27 [0.5] vs BMI > 35, 0.2 [0.4], P = .81; aged > 60 years, 0.3 [0.5] vs aged < 60 years, 0.2 [0.4], P = .26; Kellgren-Lawrence grade < 2, 0.4 [0.6] vs grade > 2, 0.2 [0.4], P = .1).

Regression modeling to predict total number of AAOS AUC evaluated interventions received produced a significant model (R² = 0.111, P = .006). The presence of ligamentous instability (β coefficient, -1.61) and the absence of mechanical symptoms (β coefficient, -0.67) were negative predictors of number of AUC interventions received. Variance inflation factors were 1.014 and 1.012, respectively. Likewise, regression modeling to identify factors predictive of receiving a rarely appropriate intervention also produced a significant model (pseudo R²= 0.06, P = .025), with lower Kellgren-Lawrence grade the only significant predictor of receiving a rarely appropriate intervention (odds ratio [OR] 0.54; 95% CI, 0.42 -0.72, per unit increase).

Timing from presentation to arthroplasty intervention was also evaluated. Age was a negative predictor (β coefficient -1.61), while positive predictors were reduced ROM (β coefficient 15.72) and having more AUC interventions (β coefficient 7.31) (model R²= 0.29, P = < .001). Age was the most significant predictor. Variance inflations factors were 1.02, 1.01, and 1.03, respectively. Receiving a rarely appropriate intervention was not associated with TKA timing.

Discussion

This single-center retrospective study examined the utilization of AAOS AUC-evaluated nonarthroplasty interventions for symptomatic knee OA prior to TKA. The aims of this study were to validate the AAOS AUC in a clinical setting and identify predictors of AAOS AUC utilization. In particular, this study focused on the number of interventions utilized prior to knee arthroplasty, whether interventions receiving a designation of rarely appropriate were used, and the duration of nonarthroplasty treatment.

 

Patients with knee instability used fewer total AAOS AUC evaluated interventions prior to TKA. Subjective instability has been reported as high as 27% in patients with OA and has been associated with fear of falling, poor balance confidence, activity limitations, and lower Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) physical function scores.¹² However, it has not been found to correlate with knee laxity.¹³ Nevertheless, significant functional impairment with the risk of falling may reduce the number of nonarthroplasty interventions attempted. On the other hand, the presence of mechanical symptoms resulted in greater utilization of nonarthroplasty interventions. This is likely due to the greater utilization of arthroscopic partial menisectomy or loose body removal in this group of patients. Despite its inclusion as an AAOS AUC evaluated intervention, arthroscopy remains a contentious treatment for symptomatic knee pain in the setting of OA.^14,15

For every unit decrease in Kellgren-Lawrence OA grade, patients were 54% more likely to receive a rarely appropriate intervention prior to knee arthroplasty. This is supported by the recent literature examining the AAOS AUC for surgical management of knee OA. Riddle and colleagues developed a classification tree to determine the contributions of various prognostic variables in final classifications of the 864 clinical vignettes used to develop the appropriateness algorithm and found that OA severity was strongly favored, with only 4 of the 432 vignettes with severe knee OA judged as rarely appropriate for surgical intervention.⁶

Our findings, too, may be explained by an AAOS AUC system that too heavily weighs radiographic severity of knee OA, resulting in more frequent rarely appropriate interventions in patients with less severe arthritis, including nonarthroplasty treatments. It is likely that rarely appropriate interventions were attempted in this subset of our study cohort based on patient’s subjective symptoms and functional status, both of which have been shown to be discordant with radiographic severity of knee OA.¹⁶

Oral or transcutaneous prescribed opioid medications were the most frequent intervention that received a rarely appropriate designation. Patients with preoperative opioid use undergoing TKA have been shown to have a greater risk for postoperative complications and longer hospital stay, particularly those patients aged < 75 years. Younger age, use of more interventions, and decreased knee ROM at presentation were predictive of longer duration of nonarthroplasty treatment. The use of more AAOS AUC evaluated interventions in these patients suggests that the AAOS AUC model may effectively be used to manage symptomatic OA, increasing the time from presentation to knee arthroplasty.

Interestingly, the use of rarely appropriate interventions did not affect TKA timing, as would be expected in a clinically effective nonarthroplasty treatment model. The reasons for rarely appropriate nonsurgical interventions are complex and require further investigation. One possible explanation is that decreased ROM was a marker for mechanical symptoms that necessitated additional intervention in the form of knee arthroscopy, delaying time to TKA.

Limitations

There are several limitations of this study. First, the small sample size (N = 90) requires acknowledgment; however, this limitation reflects the difficulty in following patients for years prior to an operative intervention. Second, the study population consists of veterans using the VA system and may not be reflective of the general population, differing with respect to gender, racial, and socioeconomic factors. Nevertheless, studies examining TKA utilization found, aside from racial and ethnic variability, patient gender and age do not affect arthroplasty utilization rate in the VA system.¹⁷

 

Additional limitations stem from the retrospective nature of this study. While the Computerized Patient Record System and centralized care of the VA system allows for review of all physical therapy consultations, orthotic consultations, and medications within the VA system, any treatments and intervention delivered by non-VA providers were not captured. Furthermore, the ability to assess for confounding variables limiting the prescription of certain medications, such as chronic kidney disease with NSAIDs or liver disease with acetaminophen, was limited by our study design.

Although our study suffers from selection bias with respect to examination of nonarthroplasty treatment in patients who have ultimately undergone TKA, we feel that this subset of patients with symptomatic knee OA represents the majority of patients evaluated for knee OA by orthopaedic surgeons in the clinic setting. It should be noted that although realignment osteotomies were sometimes indicated as appropriate by AAOS AUC model in our study population, this intervention was never performed due to patient and surgeon preference. Additionally, although it is not an AAOS AUC evaluated intervention, viscosupplementation was sporadically used during the study period; however, it is now off formulary at the investigation institution.

Conclusion

Our study suggests that patients without knee instability use more nonarthroplasty treatments over a longer period before TKA, and those patients with less severe knee OA are at risk of receiving an intervention judged to be rarely appropriate by the AAOS AUC. Such interventions do not affect timing of TKA. Nonarthroplasty care should be individualized to patients’ needs, and the decision to proceed with arthroplasty should be considered only after exhausting appropriate conservative measures. We recommend that providers use the AAOS AUC, especially when treating younger patients with less severe knee OA, particularly if considering opiate therapy or knee arthroscopy.

Acknowledgments
The authors would like to acknowledge Patrick Getty, MD, for his surgical care of some of the study patients. This material is the result of work supported with resources and the use of facilities at the Louis Stokes Cleveland VA Medical Center in Ohio.

Knee osteoarthritis (OA) affects almost 9.3 million adults in the US and accounts for $27 billion in annual health care expenses.^1,2 Due to the increasing cost of health care and an aging population, there has been renewed interest in establishing criteria for nonarthroplasty treatment of knee OA.

In 2013, using the RAND/UCLA Appropriateness method, the American Academy of Orthopaedic Surgeons (AAOS) developed an appropriate use criteria (AUC) for nonarthroplasty management of primary OA of the knee, based on orthopaedic literature and expert opinion.³ Interventions such as activity modification, weight loss, prescribed physical therapy, nonsteroidal anti-inflammatory drugs, tramadol, prescribed oral or transcutaneous opioids, acetaminophen, intra-articular corticosteroids, hinged or unloading knee braces, arthroscopic partial menisectomy or loose body removal, and realignment osteotomy were assessed. An algorithm was developed for 576 patients scenarios that incorporated patient-specific, prognostic/predictor variables to assign designations of “appropriate,” “may be appropriate,” or “rarely appropriate,” to treatment interventions.^4,5 An online version of the algorithm (orthoguidelines.org) is available for physicians and surgeons to judge appropriateness of nonarthroplasty treatments; however, it is not intended to mandate candidacy for treatment or intervention.

Clinical evaluation of the AAOS AUC is necessary to determine how treatment recommendations correlate with current practice. A recent examination of the AAOS Appropriateness System for Surgical Management of Knee OA found that prognostic/predictor variables, such as patient age, OA severity, and pattern of knee OA involvement were more heavily weighted when determining arthroplasty appropriateness than was pain severity or functional loss.⁶ Furthermore, non-AAOS AUC prognostic/predictor variables, such as race and gender, have been linked to disparities in utilization of knee OA interventions.^7-9 Such disparities can be costly not just from a patient perceptive, but also employer and societal perspectives.¹⁰

The Department of Veterans Affairs (VA) health care system represents a model of equal-access-to care system in the US that is ideal for examination of issues about health care utilization and any disparities within the AAOS AUC model and has previously been used to assess utilization of total knee arthroplasty.⁹ The aim of this study was to characterize utilization of the AAOS AUC for nonarthroplasty treatment of knee OA in a VA patient population. We asked the following questions: (1) What variables are predictive of receiving a greater number of AAOS AUC evaluated nonarthroplasty treatments? (2) What variables are predictive of receiving “rarely appropriate” AAOS AUC evaluated nonarthroplasty treatment? (3) What factors are predictive of duration of nonarthroplasty care until total knee arthroplasty (TKA)?

Methods

The institutional review board at the Louis Stokes Cleveland VA Medical Center in Ohio approved a retrospective chart review of nonarthroplasty treatments utilized by patients presenting to its orthopaedic section who subsequently underwent knee arthroplasty between 2013 and 2016. Eligibility criteria included patients aged ≥ 30 years with a diagnosis of unilateral or bilateral primary knee OA. Patients with posttraumatic OA, inflammatory arthritis, and a history of infectious arthritis or Charcot arthropathy of the knee were excluded. Patients with a body mass index (BMI) > 40 or a hemoglobin A_1c > 8.0 at presentation were excluded as nonarthroplasty care was the recommended course of treatment above these thresholds.

 

Data collected included race, gender, duration of nonarthroplasty treatment, BMI, and Kellgren-Lawrence classification of knee OA at time of presentation for symptomatic knee OA.¹¹ All AAOS AUC-evaluated nonarthroplasty treatments utilized prior to arthroplasty intervention also were recorded (Table 1). 

Indications and classifications for each subject were entered into the AAOS AUC online algorithm, and every AAOS AUC evaluated treatment utilized was assigned a rating of appropriate, may be appropriate, or rarely appropriate, based on the algorithm results for that clinical scenario (Table 2). 

Information regarding anti-inflammatory, analgesic, and prescribed oral or transcutaneous opioid use for chronic knee pain during the period of nonoperative management of knee OA prior to TKA was obtained by review of medication lists and reconciliation with orthopaedic consultation notes in the electronic health record. Peri-operative anti-inflammatory, analgesic, and prescribed oral or transcutaneous opioid use did not constitute an AUC intervention.

Statistical Analysis

Statistical analysis was completed with GraphPad Software Prism 7.0a (La Jolla, CA) and Mathworks MatLab R2016b software (Natick, MA). Univariate analysis with Student t tests with Welch corrections in the setting of unequal variance, Mann-Whitney nonparametric tests, and Fisher exact test were generated in the appropriate setting. Multivariable analyses also were conducted. For continuous outcomes, stepwise multiple linear regression was used to generate predictive models; for binary outcomes, binomial logistic regression was used.

Factors analyzed in regression modeling for the total number of AAOS AUC evaluated nonarthroplasty treatments utilized and the likelihood of receiving a rarely appropriate treatment included gender, race, function-limiting pain, range of motion (ROM), ligamentous instability, arthritis pattern, limb alignment, mechanical symptoms, BMI, age, and Kellgren-Lawrence grade. Factors analyzed in timing of TKA included the above variables plus the total number of AUC interventions, whether the patient received an inappropriate intervention, and average appropriateness of the interventions received. Residual analysis with Cook’s distance was used to identify outliers in regression. Observations with Cook’s distance > 3 times the mean Cook’s distance were identified as potential outliers, and models were adjusted accordingly. All statistical analyses were 2-tailed. Statistical significance was set to P ≤ .05 for all outputs.

Results

In the study, 97.8% of participants identified as male, and the mean age was 62.8 years (Table 3). 

The study group was predominantly white (70.3%). All participants had a diagnosis of primary OA. The majority of patients were aged 51 to 70 years (68.1%) and presented with pain occurring following short-distance ambulation (79.1%) but without mechanical symptoms (80.2%). On examination, the majority of patients were found to have full knee ROM (53.8%), no ligamentous instability (97.8%), and normal limb alignment (60.4%). Radiographically, patients most often had multicompartmental disease (69.2%) with evidence of severe joint-space narrowing (63.7%), resulting in a plurality of patients having a Kellgren-Lawrence arthritis grade of 3 (46.2%) (Table 4).

Appropriate Use Criteria Interventions

Patients received a mean of 5.2 AAOS AUC evaluated interventions before undergoing arthroplasty management at a mean of 32.3 months (range 2-181 months) from initial presentation. The majority of these interventions were classified as either appropriate or may be appropriate, according to the AUC definitions (95.1%). Self-management and physical therapy programs were widely utilized (100% and 90.1%, respectively), with all use of these interventions classified as appropriate.

 

Hinged or unloader knee braces were utilized in about half the study patients; this intervention was classified as rarely appropriate in 4.4% of these patients. Medical therapy was also widely used, with all use of NSAIDs, acetaminophen, and tramadol classified as appropriate or may be appropriate. Oral or transcutaneous opioid medications were prescribed in 14.3% of patients, with 92.3% of this use classified as rarely appropriate. Although the opioid medication prescribing provider was not specifically evaluated, there were no instances in which the orthopaedic service provided an oral or transcutaneous opioid prescriptions. Procedural interventions, with the exception of corticosteroid injections, were uncommon; no patient received realignment osteotomy, and only 12.1% of patients underwent arthroscopy. The use of arthroscopy was deemed rarely appropriate in 72.7% of these cases.

Factors Associated With AAOS AUC Intervention Use

There was no difference in the number of AAOS AUC evaluated interventions received based on BMI (mean [SD] BMI < 35, 5.2 [1.0] vs BMI ≥ 35, 5.3 [1.1], P = .49), age (mean [SD] aged < 60 years, 5.4 [1.0] vs aged ≥ 60 years, 5.1 [1.2], P = .23), or Kellgren-Lawrence arthritic grade (mean [SD] grade ≤ 2, 5.5 [1.0] vs grade > 2, 5.1 [1.1], P = .06). These variables also were not associated with receiving a rarely appropriate intervention (mean [SD] BMI < 35, 0.27 [0.5] vs BMI > 35, 0.2 [0.4], P = .81; aged > 60 years, 0.3 [0.5] vs aged < 60 years, 0.2 [0.4], P = .26; Kellgren-Lawrence grade < 2, 0.4 [0.6] vs grade > 2, 0.2 [0.4], P = .1).

Regression modeling to predict total number of AAOS AUC evaluated interventions received produced a significant model (R² = 0.111, P = .006). The presence of ligamentous instability (β coefficient, -1.61) and the absence of mechanical symptoms (β coefficient, -0.67) were negative predictors of number of AUC interventions received. Variance inflation factors were 1.014 and 1.012, respectively. Likewise, regression modeling to identify factors predictive of receiving a rarely appropriate intervention also produced a significant model (pseudo R²= 0.06, P = .025), with lower Kellgren-Lawrence grade the only significant predictor of receiving a rarely appropriate intervention (odds ratio [OR] 0.54; 95% CI, 0.42 -0.72, per unit increase).

Timing from presentation to arthroplasty intervention was also evaluated. Age was a negative predictor (β coefficient -1.61), while positive predictors were reduced ROM (β coefficient 15.72) and having more AUC interventions (β coefficient 7.31) (model R²= 0.29, P = < .001). Age was the most significant predictor. Variance inflations factors were 1.02, 1.01, and 1.03, respectively. Receiving a rarely appropriate intervention was not associated with TKA timing.

Discussion

This single-center retrospective study examined the utilization of AAOS AUC-evaluated nonarthroplasty interventions for symptomatic knee OA prior to TKA. The aims of this study were to validate the AAOS AUC in a clinical setting and identify predictors of AAOS AUC utilization. In particular, this study focused on the number of interventions utilized prior to knee arthroplasty, whether interventions receiving a designation of rarely appropriate were used, and the duration of nonarthroplasty treatment.

 

Patients with knee instability used fewer total AAOS AUC evaluated interventions prior to TKA. Subjective instability has been reported as high as 27% in patients with OA and has been associated with fear of falling, poor balance confidence, activity limitations, and lower Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) physical function scores.¹² However, it has not been found to correlate with knee laxity.¹³ Nevertheless, significant functional impairment with the risk of falling may reduce the number of nonarthroplasty interventions attempted. On the other hand, the presence of mechanical symptoms resulted in greater utilization of nonarthroplasty interventions. This is likely due to the greater utilization of arthroscopic partial menisectomy or loose body removal in this group of patients. Despite its inclusion as an AAOS AUC evaluated intervention, arthroscopy remains a contentious treatment for symptomatic knee pain in the setting of OA.^14,15

For every unit decrease in Kellgren-Lawrence OA grade, patients were 54% more likely to receive a rarely appropriate intervention prior to knee arthroplasty. This is supported by the recent literature examining the AAOS AUC for surgical management of knee OA. Riddle and colleagues developed a classification tree to determine the contributions of various prognostic variables in final classifications of the 864 clinical vignettes used to develop the appropriateness algorithm and found that OA severity was strongly favored, with only 4 of the 432 vignettes with severe knee OA judged as rarely appropriate for surgical intervention.⁶

Our findings, too, may be explained by an AAOS AUC system that too heavily weighs radiographic severity of knee OA, resulting in more frequent rarely appropriate interventions in patients with less severe arthritis, including nonarthroplasty treatments. It is likely that rarely appropriate interventions were attempted in this subset of our study cohort based on patient’s subjective symptoms and functional status, both of which have been shown to be discordant with radiographic severity of knee OA.¹⁶

Oral or transcutaneous prescribed opioid medications were the most frequent intervention that received a rarely appropriate designation. Patients with preoperative opioid use undergoing TKA have been shown to have a greater risk for postoperative complications and longer hospital stay, particularly those patients aged < 75 years. Younger age, use of more interventions, and decreased knee ROM at presentation were predictive of longer duration of nonarthroplasty treatment. The use of more AAOS AUC evaluated interventions in these patients suggests that the AAOS AUC model may effectively be used to manage symptomatic OA, increasing the time from presentation to knee arthroplasty.

Interestingly, the use of rarely appropriate interventions did not affect TKA timing, as would be expected in a clinically effective nonarthroplasty treatment model. The reasons for rarely appropriate nonsurgical interventions are complex and require further investigation. One possible explanation is that decreased ROM was a marker for mechanical symptoms that necessitated additional intervention in the form of knee arthroscopy, delaying time to TKA.

Limitations

There are several limitations of this study. First, the small sample size (N = 90) requires acknowledgment; however, this limitation reflects the difficulty in following patients for years prior to an operative intervention. Second, the study population consists of veterans using the VA system and may not be reflective of the general population, differing with respect to gender, racial, and socioeconomic factors. Nevertheless, studies examining TKA utilization found, aside from racial and ethnic variability, patient gender and age do not affect arthroplasty utilization rate in the VA system.¹⁷

 

Additional limitations stem from the retrospective nature of this study. While the Computerized Patient Record System and centralized care of the VA system allows for review of all physical therapy consultations, orthotic consultations, and medications within the VA system, any treatments and intervention delivered by non-VA providers were not captured. Furthermore, the ability to assess for confounding variables limiting the prescription of certain medications, such as chronic kidney disease with NSAIDs or liver disease with acetaminophen, was limited by our study design.

Although our study suffers from selection bias with respect to examination of nonarthroplasty treatment in patients who have ultimately undergone TKA, we feel that this subset of patients with symptomatic knee OA represents the majority of patients evaluated for knee OA by orthopaedic surgeons in the clinic setting. It should be noted that although realignment osteotomies were sometimes indicated as appropriate by AAOS AUC model in our study population, this intervention was never performed due to patient and surgeon preference. Additionally, although it is not an AAOS AUC evaluated intervention, viscosupplementation was sporadically used during the study period; however, it is now off formulary at the investigation institution.

Conclusion

Our study suggests that patients without knee instability use more nonarthroplasty treatments over a longer period before TKA, and those patients with less severe knee OA are at risk of receiving an intervention judged to be rarely appropriate by the AAOS AUC. Such interventions do not affect timing of TKA. Nonarthroplasty care should be individualized to patients’ needs, and the decision to proceed with arthroplasty should be considered only after exhausting appropriate conservative measures. We recommend that providers use the AAOS AUC, especially when treating younger patients with less severe knee OA, particularly if considering opiate therapy or knee arthroscopy.

Acknowledgments
The authors would like to acknowledge Patrick Getty, MD, for his surgical care of some of the study patients. This material is the result of work supported with resources and the use of facilities at the Louis Stokes Cleveland VA Medical Center in Ohio.

Psoriasis is a chronic skin condition affecting approximately 2% to 3% of the population.^1,2 Beyond cutaneous manifestations, psoriasis is a systemic inflammatory state that is associated with an increased risk for cardiovascular disease, including obesity,^3,4 type 2 diabetes mellitus,^5,6 hypertension,⁵ dyslipidemia,^3,7 metabolic syndrome,⁷ atherosclerosis,⁸ peripheral vascular disease,⁹ coronary artery calcification,¹⁰ myocardial infarction,^11-13 stroke,^9,14 and cardiac death.^15,16

Psoriasis also has been associated with inflammatory bowel disease (IBD), possibly because of similar autoimmune mechanisms in the pathogenesis of both diseases.^17,18 However, there is no literature regarding the risk for acute gastrointestinal pathologies such as appendicitis, cholecystitis, or diverticulitis in patients with psoriasis.

The primary objective of this study was to examine if patients with psoriasis are at increased risk for appendicitis, cholecystitis, or diverticulitis compared to the general population. The secondary objective was to determine if patients with severe psoriasis (ie, patients treated with phototherapy or systemic therapy) are at a higher risk for these conditions compared to patients with mild psoriasis.

Methods

Patients and Tools
A descriptive, population-based cohort study design with controls from a matched cohort was used to ascertain the effect of psoriasis status on patients’ risk for appendicitis, cholecystitis, or diverticulitis. Our cohort was selected using administrative data from Kaiser Permanente Southern California (KPSC) during the study period (January 1, 2004, through December 31, 2016).

Kaiser Permanente Southern California is a large integrated health maintenance organization that includes approximately 4 million patients as of December 31, 2016, and includes roughly 20% of the region’s population. The geographic area served extends from Bakersfield in the lower California Central Valley to San Diego on the border with Mexico. Membership demographics, socioeconomic status, and ethnicity composition are representative of California.

Patients were included if they had a diagnosis of psoriasis (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] code 696.1; International Classification of Diseases, Tenth Revision, Clinical Modification [ICD-10-CM] codes L40.0, L40.4, L40.8, or L40.9) for at least 3 visits between January 1, 2004, and December 31, 2016. Patients were not excluded if they also had a diagnosis of psoriatic arthritis (ICD-9-CM code 696.0; ICD-10-CM code L40.5x). Patients also must have been continuously enrolled for at least 1 year before and 1 year after the index date, which was defined as the date of the third psoriasis diagnosis.

Each patient with psoriasis was assigned to 1 of 2 cohorts: (1) severe psoriasis: patients who received UVB phototherapy, psoralen plus UVA phototherapy, methotrexate, acitretin, cyclosporine, apremilast, etanercept, adalimumab, infliximab, ustekinumab, efalizumab, alefacept, secukinumab, or ixekizumab during the study period; and (2) mild psoriasis: patients who had a diagnosis of psoriasis who did not receive one of these therapies during the study period.

Patients were excluded if they had a history of appendicitis, cholecystitis, or diverticulitis at any time before the index date. Only patients older than 18 years were included.

Patients with psoriasis were frequency matched (1:5) with healthy patients, also from the KPSC network. Individuals were matched by age, sex, and ethnicity.

Statistical Analysis
Baseline characteristics were described with means and SD for continuous variables as well as percentages for categorical variables. Chi-square tests for categorical variables and the Mann-Whitney U Test for continuous variables were used to compare the patients’ characteristics by psoriasis status. Cox proportional hazards regression models were used to examine the risk for appendicitis, cholecystitis, or diverticulitis among patients with and without psoriasis and among patients with mild and severe psoriasis. Proportionality assumption was validated using Pearson product moment correlation between the scaled Schoenfeld residuals and log transformed time for each covariate.

Results were presented as crude (unadjusted) hazard ratios (HRs) and adjusted HRs, where confounding factors (ie, age, sex, ethnicity, body mass index [BMI], alcohol use, smoking status, income, education, and membership length) were adjusted. All tests were performed with SAS EG 5.1 and R software. P<.05 was considered statistically significant. Results are reported with the 95% confidence interval (CI), when appropriate.

 

Results

A total of 1,690,214 KPSC patients were eligible for the study; 10,307 (0.6%) met diagnostic and inclusion criteria for the psoriasis cohort. Patients with psoriasis had a significantly higher mean BMI (29.9 vs 28.7; P<.0001) as well as higher mean rates of alcohol use (56% vs 53%; P<.0001) and smoking (47% vs 38%; P<.01) compared to controls. Psoriasis patients had a shorter average duration of membership within the Kaiser network (P=.0001) compared to controls.

A total of 7416 patients met criteria for mild psoriasis and 2891 patients met criteria for severe psoriasis (eTable). Patients with severe psoriasis were significantly younger and had significantly higher mean BMI compared to patients with mild psoriasis (P<.0001 and P=.0001, respectively). No significant difference in rates of alcohol or tobacco use was detected among patients with mild and severe psoriasis.

Appendicitis
The prevalence of appendicitis was not significantly different between patients with and without psoriasis or between patients with mild and severe psoriasis, though the incidence rate was slightly higher among patients with psoriasis (0.80 per 1000 patient-years compared to 0.62 per 1000 patient-years among patients without psoriasis)(Table 1). However, there was not a significant difference in risk for appendicitis between healthy patients, patients with severe psoriasis, and patients with mild psoriasis after adjusting for potential confounding factors (Table 2). Interestingly, patients with severe psoriasis who had a diagnosis of appendicitis had a significantly shorter time to diagnosis of appendicitis compared to patients with mild psoriasis (7.4 years vs 8.1 years; P<.0001).

Cholecystitis
Psoriasis patients also did not have an increased prevalence of cholecystitis compared to healthy patients. However, patients with severe psoriasis had a significantly higher prevalence of cholecystitis compared to patients with mild psoriasis (P=.0038). Overall, patients with psoriasis had a slightly higher incidence rate (1.72 per 1000 patient-years) compared to healthy patients (1.46 per 1000 patient-years). Moreover, the time to diagnosis of cholecystitis was significantly shorter for patients with severe psoriasis than for patients with mild psoriasis (7.4 years vs 8.1 years; P<.0001). Mild psoriasis was associated with a significantly increased risk (HR, 1.33; 95% CI, 1.09-1.63; P<.01) for cholecystitis compared to individuals without psoriasis in both the crude and adjusted models (Table 2). There was no difference between mild psoriasis patients and severe psoriasis patients in risk for cholecystitis.

Diverticulitis
Patients with psoriasis had a significantly greater prevalence of diverticulitis compared to the control cohort (5.1% vs 4.2%; P<.0001). There was no difference in prevalence between the severe psoriasis group and the mild psoriasis group (P=.96), but the time to diagnosis of diverticulitis was shorter in the severe psoriasis group than in the mild psoriasis group (7.2 years vs 7.9 years; P<.0001). Psoriasis patients had an incidence rate of diverticulitis of 6.61 per 1000 patient-years compared to 5.38 per 1000 patient-years in the control group. Psoriasis conferred a higher risk for diverticulitis in both the crude and adjusted models (HR, 1.23; 95% CI, 1.11-1.35 [P<.001] and HR, 1.16; 95% CI, 1.05-1.29; [P<.01], respectively)(Table 3); however, when stratified by disease severity, only patients with severe psoriasis were found to be at higher risk (HR, 1.26; 95% CI, 1.15-1.61; P<.001 for the adjusted model).

 

Comment

The objective of this study was to examine the background risks for specific gastrointestinal pathologies in a large cohort of patients with psoriasis compared to the general population. After adjusting for measured confounders, patients with severe psoriasis had a significantly higher risk of diverticulitis compared to the general population. Although more patients with severe psoriasis developed appendicitis or cholecystitis, the difference was not significant.

The pathogenesis of diverticulosis and diverticulitis has been thought to be related to increased intracolonic pressure and decreased dietary fiber intake, leading to formation of diverticula in the colon.¹⁹ Our study did not correct for differences in diet between the 2 groups, making it a possible confounding variable. Studies evaluating dietary habits of psoriatic patients have found that adult males with psoriasis might consume less fiber compared to healthy patients,²⁰ and psoriasis patients also might consume less whole-grain fiber.²¹ Furthermore, fiber deficiency also might affect gut flora, causing low-grade chronic inflammation,¹⁸ which also has been supported by response to anti-inflammatory medications such as mesalazine.²² Given the autoimmune association between psoriasis and IBD, it is possible that psoriasis also might create an environment of chronic inflammation in the gut, predisposing patients with psoriasis to diverticulitis. However, further research is needed to better evaluate this possibility.

Our study also does not address any potential effects on outcomes of specific treatments for psoriasis. Brandl et al²³ found that patients on immunosuppressive therapy for autoimmune diseases had longer hospital and intensive care unit stays, higher rates of emergency operations, and higher mortality while hospitalized. Because our results suggest that patients with severe psoriasis, who are therefore more likely to require treatment with an immunomodulator, are at higher risk for diverticulitis, these patients also might be at risk for poorer outcomes.

There is no literature evaluating the relationship between psoriasis and appendicitis. Our study found a slightly lower incidence rate compared to the national trend (9.38 per 10,000 patient-years in the United States in 2008) in both healthy patients and psoriasis patients.²⁴ Of note, this statistic includes children, whereas our study did not, which might in part account for the lower rate. However, Cheluvappa et al²⁵ hypothesized a relationship between appendicitis and subsequent appendectomy at a young age and protection against IBD. They also found that the mechanism for protection involves downregulation of the helper T cell (T_H17) pathway,²⁵ which also has been found to play a role in psoriasis pathogenesis.^26,27 Although our results suggest that the risk for appendicitis is not increased for patients with psoriasis, further research might be able to determine if appendicitis and subsequent appendectomy also can offer protection against development of psoriasis.

We found that patients with severe psoriasis had a higher incidence rate of cholecystitis compared to patients with mild psoriasis. Egeberg et al²⁸ found an increased risk for cholelithiasis among patients with psoriasis, which may contribute to a higher rate of cholecystitis. Although both acute and chronic cholecystitis were incorporated in this study, a Russian study found that chronic cholecystitis may be a predictor of progression of psoriasis.²⁹ Moreover, patients with severe psoriasis had a shorter duration to diagnosis of cholecystitis than patients with mild psoriasis. It is possible that patients with severe psoriasis are in a state of greater chronic inflammation than those with mild psoriasis, and therefore, when combined with other risk factors for cholecystitis, may progress to disease more quickly. Alternatively, this finding could be treatment related, as there have been reported cases of cholecystitis related to etanercept use in patients treated for psoriasis and juvenile polyarticular rheumatoid arthritis.^30,31 The relationship is not yet well defined, however, and further research is necessary to evaluate this association.

Study Strengths
Key strengths of this study include the large sample size and diversity of the patient population. Kaiser Permanente Southern California membership generally is representative of the broader community, making our results fairly generalizable to populations with health insurance. Use of a matched control cohort allows the results to be more specific to the disease of interest, and the population-based design minimizes bias.

Study Limitations
This study has several limitations. Although the cohorts were categorized based on type of treatment received, exact therapies were not specified. As a retrospective study, it is difficult to control for potential confounding variables that are not included in the electronic medical record. The results of this study also demonstrated significantly shorter durations to diagnosis of all 3 conditions, indicating that surveillance bias may be present.

Conclusion

Patients with psoriasis may be at an increased risk for diverticulitis compared to patients without psoriasis, which could be due to the chronic inflammatory state induced by psoriasis. Therefore, it may be beneficial for clinicians to evaluate psoriasis patients for other risk factors for diverticulitis and subsequently provide counseling to these patients to minimize their risk for diverticulitis. Psoriasis patients do not appear to be at an increased risk for appendicitis or cholecystitis compared to controls; however, further research is needed for confirmation.

Psoriasis is a chronic skin condition affecting approximately 2% to 3% of the population.^1,2 Beyond cutaneous manifestations, psoriasis is a systemic inflammatory state that is associated with an increased risk for cardiovascular disease, including obesity,^3,4 type 2 diabetes mellitus,^5,6 hypertension,⁵ dyslipidemia,^3,7 metabolic syndrome,⁷ atherosclerosis,⁸ peripheral vascular disease,⁹ coronary artery calcification,¹⁰ myocardial infarction,^11-13 stroke,^9,14 and cardiac death.^15,16

Psoriasis also has been associated with inflammatory bowel disease (IBD), possibly because of similar autoimmune mechanisms in the pathogenesis of both diseases.^17,18 However, there is no literature regarding the risk for acute gastrointestinal pathologies such as appendicitis, cholecystitis, or diverticulitis in patients with psoriasis.

The primary objective of this study was to examine if patients with psoriasis are at increased risk for appendicitis, cholecystitis, or diverticulitis compared to the general population. The secondary objective was to determine if patients with severe psoriasis (ie, patients treated with phototherapy or systemic therapy) are at a higher risk for these conditions compared to patients with mild psoriasis.

Methods

Patients and Tools
A descriptive, population-based cohort study design with controls from a matched cohort was used to ascertain the effect of psoriasis status on patients’ risk for appendicitis, cholecystitis, or diverticulitis. Our cohort was selected using administrative data from Kaiser Permanente Southern California (KPSC) during the study period (January 1, 2004, through December 31, 2016).

Kaiser Permanente Southern California is a large integrated health maintenance organization that includes approximately 4 million patients as of December 31, 2016, and includes roughly 20% of the region’s population. The geographic area served extends from Bakersfield in the lower California Central Valley to San Diego on the border with Mexico. Membership demographics, socioeconomic status, and ethnicity composition are representative of California.

Patients were included if they had a diagnosis of psoriasis (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] code 696.1; International Classification of Diseases, Tenth Revision, Clinical Modification [ICD-10-CM] codes L40.0, L40.4, L40.8, or L40.9) for at least 3 visits between January 1, 2004, and December 31, 2016. Patients were not excluded if they also had a diagnosis of psoriatic arthritis (ICD-9-CM code 696.0; ICD-10-CM code L40.5x). Patients also must have been continuously enrolled for at least 1 year before and 1 year after the index date, which was defined as the date of the third psoriasis diagnosis.

Each patient with psoriasis was assigned to 1 of 2 cohorts: (1) severe psoriasis: patients who received UVB phototherapy, psoralen plus UVA phototherapy, methotrexate, acitretin, cyclosporine, apremilast, etanercept, adalimumab, infliximab, ustekinumab, efalizumab, alefacept, secukinumab, or ixekizumab during the study period; and (2) mild psoriasis: patients who had a diagnosis of psoriasis who did not receive one of these therapies during the study period.

Patients were excluded if they had a history of appendicitis, cholecystitis, or diverticulitis at any time before the index date. Only patients older than 18 years were included.

Patients with psoriasis were frequency matched (1:5) with healthy patients, also from the KPSC network. Individuals were matched by age, sex, and ethnicity.

Statistical Analysis
Baseline characteristics were described with means and SD for continuous variables as well as percentages for categorical variables. Chi-square tests for categorical variables and the Mann-Whitney U Test for continuous variables were used to compare the patients’ characteristics by psoriasis status. Cox proportional hazards regression models were used to examine the risk for appendicitis, cholecystitis, or diverticulitis among patients with and without psoriasis and among patients with mild and severe psoriasis. Proportionality assumption was validated using Pearson product moment correlation between the scaled Schoenfeld residuals and log transformed time for each covariate.

Results were presented as crude (unadjusted) hazard ratios (HRs) and adjusted HRs, where confounding factors (ie, age, sex, ethnicity, body mass index [BMI], alcohol use, smoking status, income, education, and membership length) were adjusted. All tests were performed with SAS EG 5.1 and R software. P<.05 was considered statistically significant. Results are reported with the 95% confidence interval (CI), when appropriate.

 

Results

A total of 1,690,214 KPSC patients were eligible for the study; 10,307 (0.6%) met diagnostic and inclusion criteria for the psoriasis cohort. Patients with psoriasis had a significantly higher mean BMI (29.9 vs 28.7; P<.0001) as well as higher mean rates of alcohol use (56% vs 53%; P<.0001) and smoking (47% vs 38%; P<.01) compared to controls. Psoriasis patients had a shorter average duration of membership within the Kaiser network (P=.0001) compared to controls.

A total of 7416 patients met criteria for mild psoriasis and 2891 patients met criteria for severe psoriasis (eTable). Patients with severe psoriasis were significantly younger and had significantly higher mean BMI compared to patients with mild psoriasis (P<.0001 and P=.0001, respectively). No significant difference in rates of alcohol or tobacco use was detected among patients with mild and severe psoriasis.

Appendicitis
The prevalence of appendicitis was not significantly different between patients with and without psoriasis or between patients with mild and severe psoriasis, though the incidence rate was slightly higher among patients with psoriasis (0.80 per 1000 patient-years compared to 0.62 per 1000 patient-years among patients without psoriasis)(Table 1). However, there was not a significant difference in risk for appendicitis between healthy patients, patients with severe psoriasis, and patients with mild psoriasis after adjusting for potential confounding factors (Table 2). Interestingly, patients with severe psoriasis who had a diagnosis of appendicitis had a significantly shorter time to diagnosis of appendicitis compared to patients with mild psoriasis (7.4 years vs 8.1 years; P<.0001).

Cholecystitis
Psoriasis patients also did not have an increased prevalence of cholecystitis compared to healthy patients. However, patients with severe psoriasis had a significantly higher prevalence of cholecystitis compared to patients with mild psoriasis (P=.0038). Overall, patients with psoriasis had a slightly higher incidence rate (1.72 per 1000 patient-years) compared to healthy patients (1.46 per 1000 patient-years). Moreover, the time to diagnosis of cholecystitis was significantly shorter for patients with severe psoriasis than for patients with mild psoriasis (7.4 years vs 8.1 years; P<.0001). Mild psoriasis was associated with a significantly increased risk (HR, 1.33; 95% CI, 1.09-1.63; P<.01) for cholecystitis compared to individuals without psoriasis in both the crude and adjusted models (Table 2). There was no difference between mild psoriasis patients and severe psoriasis patients in risk for cholecystitis.

Diverticulitis
Patients with psoriasis had a significantly greater prevalence of diverticulitis compared to the control cohort (5.1% vs 4.2%; P<.0001). There was no difference in prevalence between the severe psoriasis group and the mild psoriasis group (P=.96), but the time to diagnosis of diverticulitis was shorter in the severe psoriasis group than in the mild psoriasis group (7.2 years vs 7.9 years; P<.0001). Psoriasis patients had an incidence rate of diverticulitis of 6.61 per 1000 patient-years compared to 5.38 per 1000 patient-years in the control group. Psoriasis conferred a higher risk for diverticulitis in both the crude and adjusted models (HR, 1.23; 95% CI, 1.11-1.35 [P<.001] and HR, 1.16; 95% CI, 1.05-1.29; [P<.01], respectively)(Table 3); however, when stratified by disease severity, only patients with severe psoriasis were found to be at higher risk (HR, 1.26; 95% CI, 1.15-1.61; P<.001 for the adjusted model).

 

Comment

The objective of this study was to examine the background risks for specific gastrointestinal pathologies in a large cohort of patients with psoriasis compared to the general population. After adjusting for measured confounders, patients with severe psoriasis had a significantly higher risk of diverticulitis compared to the general population. Although more patients with severe psoriasis developed appendicitis or cholecystitis, the difference was not significant.

The pathogenesis of diverticulosis and diverticulitis has been thought to be related to increased intracolonic pressure and decreased dietary fiber intake, leading to formation of diverticula in the colon.¹⁹ Our study did not correct for differences in diet between the 2 groups, making it a possible confounding variable. Studies evaluating dietary habits of psoriatic patients have found that adult males with psoriasis might consume less fiber compared to healthy patients,²⁰ and psoriasis patients also might consume less whole-grain fiber.²¹ Furthermore, fiber deficiency also might affect gut flora, causing low-grade chronic inflammation,¹⁸ which also has been supported by response to anti-inflammatory medications such as mesalazine.²² Given the autoimmune association between psoriasis and IBD, it is possible that psoriasis also might create an environment of chronic inflammation in the gut, predisposing patients with psoriasis to diverticulitis. However, further research is needed to better evaluate this possibility.

Our study also does not address any potential effects on outcomes of specific treatments for psoriasis. Brandl et al²³ found that patients on immunosuppressive therapy for autoimmune diseases had longer hospital and intensive care unit stays, higher rates of emergency operations, and higher mortality while hospitalized. Because our results suggest that patients with severe psoriasis, who are therefore more likely to require treatment with an immunomodulator, are at higher risk for diverticulitis, these patients also might be at risk for poorer outcomes.

There is no literature evaluating the relationship between psoriasis and appendicitis. Our study found a slightly lower incidence rate compared to the national trend (9.38 per 10,000 patient-years in the United States in 2008) in both healthy patients and psoriasis patients.²⁴ Of note, this statistic includes children, whereas our study did not, which might in part account for the lower rate. However, Cheluvappa et al²⁵ hypothesized a relationship between appendicitis and subsequent appendectomy at a young age and protection against IBD. They also found that the mechanism for protection involves downregulation of the helper T cell (T_H17) pathway,²⁵ which also has been found to play a role in psoriasis pathogenesis.^26,27 Although our results suggest that the risk for appendicitis is not increased for patients with psoriasis, further research might be able to determine if appendicitis and subsequent appendectomy also can offer protection against development of psoriasis.

We found that patients with severe psoriasis had a higher incidence rate of cholecystitis compared to patients with mild psoriasis. Egeberg et al²⁸ found an increased risk for cholelithiasis among patients with psoriasis, which may contribute to a higher rate of cholecystitis. Although both acute and chronic cholecystitis were incorporated in this study, a Russian study found that chronic cholecystitis may be a predictor of progression of psoriasis.²⁹ Moreover, patients with severe psoriasis had a shorter duration to diagnosis of cholecystitis than patients with mild psoriasis. It is possible that patients with severe psoriasis are in a state of greater chronic inflammation than those with mild psoriasis, and therefore, when combined with other risk factors for cholecystitis, may progress to disease more quickly. Alternatively, this finding could be treatment related, as there have been reported cases of cholecystitis related to etanercept use in patients treated for psoriasis and juvenile polyarticular rheumatoid arthritis.^30,31 The relationship is not yet well defined, however, and further research is necessary to evaluate this association.

Study Strengths
Key strengths of this study include the large sample size and diversity of the patient population. Kaiser Permanente Southern California membership generally is representative of the broader community, making our results fairly generalizable to populations with health insurance. Use of a matched control cohort allows the results to be more specific to the disease of interest, and the population-based design minimizes bias.

Study Limitations
This study has several limitations. Although the cohorts were categorized based on type of treatment received, exact therapies were not specified. As a retrospective study, it is difficult to control for potential confounding variables that are not included in the electronic medical record. The results of this study also demonstrated significantly shorter durations to diagnosis of all 3 conditions, indicating that surveillance bias may be present.

Conclusion

Patients with psoriasis may be at an increased risk for diverticulitis compared to patients without psoriasis, which could be due to the chronic inflammatory state induced by psoriasis. Therefore, it may be beneficial for clinicians to evaluate psoriasis patients for other risk factors for diverticulitis and subsequently provide counseling to these patients to minimize their risk for diverticulitis. Psoriasis patients do not appear to be at an increased risk for appendicitis or cholecystitis compared to controls; however, further research is needed for confirmation.

Psoriasis is a chronic skin condition affecting approximately 2% to 3% of the population.^1,2 Beyond cutaneous manifestations, psoriasis is a systemic inflammatory state that is associated with an increased risk for cardiovascular disease, including obesity,^3,4 type 2 diabetes mellitus,^5,6 hypertension,⁵ dyslipidemia,^3,7 metabolic syndrome,⁷ atherosclerosis,⁸ peripheral vascular disease,⁹ coronary artery calcification,¹⁰ myocardial infarction,^11-13 stroke,^9,14 and cardiac death.^15,16

Psoriasis also has been associated with inflammatory bowel disease (IBD), possibly because of similar autoimmune mechanisms in the pathogenesis of both diseases.^17,18 However, there is no literature regarding the risk for acute gastrointestinal pathologies such as appendicitis, cholecystitis, or diverticulitis in patients with psoriasis.

The primary objective of this study was to examine if patients with psoriasis are at increased risk for appendicitis, cholecystitis, or diverticulitis compared to the general population. The secondary objective was to determine if patients with severe psoriasis (ie, patients treated with phototherapy or systemic therapy) are at a higher risk for these conditions compared to patients with mild psoriasis.

Methods

Patients and Tools
A descriptive, population-based cohort study design with controls from a matched cohort was used to ascertain the effect of psoriasis status on patients’ risk for appendicitis, cholecystitis, or diverticulitis. Our cohort was selected using administrative data from Kaiser Permanente Southern California (KPSC) during the study period (January 1, 2004, through December 31, 2016).

Kaiser Permanente Southern California is a large integrated health maintenance organization that includes approximately 4 million patients as of December 31, 2016, and includes roughly 20% of the region’s population. The geographic area served extends from Bakersfield in the lower California Central Valley to San Diego on the border with Mexico. Membership demographics, socioeconomic status, and ethnicity composition are representative of California.

Patients were included if they had a diagnosis of psoriasis (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] code 696.1; International Classification of Diseases, Tenth Revision, Clinical Modification [ICD-10-CM] codes L40.0, L40.4, L40.8, or L40.9) for at least 3 visits between January 1, 2004, and December 31, 2016. Patients were not excluded if they also had a diagnosis of psoriatic arthritis (ICD-9-CM code 696.0; ICD-10-CM code L40.5x). Patients also must have been continuously enrolled for at least 1 year before and 1 year after the index date, which was defined as the date of the third psoriasis diagnosis.

Each patient with psoriasis was assigned to 1 of 2 cohorts: (1) severe psoriasis: patients who received UVB phototherapy, psoralen plus UVA phototherapy, methotrexate, acitretin, cyclosporine, apremilast, etanercept, adalimumab, infliximab, ustekinumab, efalizumab, alefacept, secukinumab, or ixekizumab during the study period; and (2) mild psoriasis: patients who had a diagnosis of psoriasis who did not receive one of these therapies during the study period.

Patients were excluded if they had a history of appendicitis, cholecystitis, or diverticulitis at any time before the index date. Only patients older than 18 years were included.

Patients with psoriasis were frequency matched (1:5) with healthy patients, also from the KPSC network. Individuals were matched by age, sex, and ethnicity.

Statistical Analysis
Baseline characteristics were described with means and SD for continuous variables as well as percentages for categorical variables. Chi-square tests for categorical variables and the Mann-Whitney U Test for continuous variables were used to compare the patients’ characteristics by psoriasis status. Cox proportional hazards regression models were used to examine the risk for appendicitis, cholecystitis, or diverticulitis among patients with and without psoriasis and among patients with mild and severe psoriasis. Proportionality assumption was validated using Pearson product moment correlation between the scaled Schoenfeld residuals and log transformed time for each covariate.

Results were presented as crude (unadjusted) hazard ratios (HRs) and adjusted HRs, where confounding factors (ie, age, sex, ethnicity, body mass index [BMI], alcohol use, smoking status, income, education, and membership length) were adjusted. All tests were performed with SAS EG 5.1 and R software. P<.05 was considered statistically significant. Results are reported with the 95% confidence interval (CI), when appropriate.

 

Results

A total of 1,690,214 KPSC patients were eligible for the study; 10,307 (0.6%) met diagnostic and inclusion criteria for the psoriasis cohort. Patients with psoriasis had a significantly higher mean BMI (29.9 vs 28.7; P<.0001) as well as higher mean rates of alcohol use (56% vs 53%; P<.0001) and smoking (47% vs 38%; P<.01) compared to controls. Psoriasis patients had a shorter average duration of membership within the Kaiser network (P=.0001) compared to controls.

A total of 7416 patients met criteria for mild psoriasis and 2891 patients met criteria for severe psoriasis (eTable). Patients with severe psoriasis were significantly younger and had significantly higher mean BMI compared to patients with mild psoriasis (P<.0001 and P=.0001, respectively). No significant difference in rates of alcohol or tobacco use was detected among patients with mild and severe psoriasis.

Appendicitis
The prevalence of appendicitis was not significantly different between patients with and without psoriasis or between patients with mild and severe psoriasis, though the incidence rate was slightly higher among patients with psoriasis (0.80 per 1000 patient-years compared to 0.62 per 1000 patient-years among patients without psoriasis)(Table 1). However, there was not a significant difference in risk for appendicitis between healthy patients, patients with severe psoriasis, and patients with mild psoriasis after adjusting for potential confounding factors (Table 2). Interestingly, patients with severe psoriasis who had a diagnosis of appendicitis had a significantly shorter time to diagnosis of appendicitis compared to patients with mild psoriasis (7.4 years vs 8.1 years; P<.0001).

Cholecystitis
Psoriasis patients also did not have an increased prevalence of cholecystitis compared to healthy patients. However, patients with severe psoriasis had a significantly higher prevalence of cholecystitis compared to patients with mild psoriasis (P=.0038). Overall, patients with psoriasis had a slightly higher incidence rate (1.72 per 1000 patient-years) compared to healthy patients (1.46 per 1000 patient-years). Moreover, the time to diagnosis of cholecystitis was significantly shorter for patients with severe psoriasis than for patients with mild psoriasis (7.4 years vs 8.1 years; P<.0001). Mild psoriasis was associated with a significantly increased risk (HR, 1.33; 95% CI, 1.09-1.63; P<.01) for cholecystitis compared to individuals without psoriasis in both the crude and adjusted models (Table 2). There was no difference between mild psoriasis patients and severe psoriasis patients in risk for cholecystitis.

Diverticulitis
Patients with psoriasis had a significantly greater prevalence of diverticulitis compared to the control cohort (5.1% vs 4.2%; P<.0001). There was no difference in prevalence between the severe psoriasis group and the mild psoriasis group (P=.96), but the time to diagnosis of diverticulitis was shorter in the severe psoriasis group than in the mild psoriasis group (7.2 years vs 7.9 years; P<.0001). Psoriasis patients had an incidence rate of diverticulitis of 6.61 per 1000 patient-years compared to 5.38 per 1000 patient-years in the control group. Psoriasis conferred a higher risk for diverticulitis in both the crude and adjusted models (HR, 1.23; 95% CI, 1.11-1.35 [P<.001] and HR, 1.16; 95% CI, 1.05-1.29; [P<.01], respectively)(Table 3); however, when stratified by disease severity, only patients with severe psoriasis were found to be at higher risk (HR, 1.26; 95% CI, 1.15-1.61; P<.001 for the adjusted model).

 

Comment

The objective of this study was to examine the background risks for specific gastrointestinal pathologies in a large cohort of patients with psoriasis compared to the general population. After adjusting for measured confounders, patients with severe psoriasis had a significantly higher risk of diverticulitis compared to the general population. Although more patients with severe psoriasis developed appendicitis or cholecystitis, the difference was not significant.

The pathogenesis of diverticulosis and diverticulitis has been thought to be related to increased intracolonic pressure and decreased dietary fiber intake, leading to formation of diverticula in the colon.¹⁹ Our study did not correct for differences in diet between the 2 groups, making it a possible confounding variable. Studies evaluating dietary habits of psoriatic patients have found that adult males with psoriasis might consume less fiber compared to healthy patients,²⁰ and psoriasis patients also might consume less whole-grain fiber.²¹ Furthermore, fiber deficiency also might affect gut flora, causing low-grade chronic inflammation,¹⁸ which also has been supported by response to anti-inflammatory medications such as mesalazine.²² Given the autoimmune association between psoriasis and IBD, it is possible that psoriasis also might create an environment of chronic inflammation in the gut, predisposing patients with psoriasis to diverticulitis. However, further research is needed to better evaluate this possibility.

Our study also does not address any potential effects on outcomes of specific treatments for psoriasis. Brandl et al²³ found that patients on immunosuppressive therapy for autoimmune diseases had longer hospital and intensive care unit stays, higher rates of emergency operations, and higher mortality while hospitalized. Because our results suggest that patients with severe psoriasis, who are therefore more likely to require treatment with an immunomodulator, are at higher risk for diverticulitis, these patients also might be at risk for poorer outcomes.

There is no literature evaluating the relationship between psoriasis and appendicitis. Our study found a slightly lower incidence rate compared to the national trend (9.38 per 10,000 patient-years in the United States in 2008) in both healthy patients and psoriasis patients.²⁴ Of note, this statistic includes children, whereas our study did not, which might in part account for the lower rate. However, Cheluvappa et al²⁵ hypothesized a relationship between appendicitis and subsequent appendectomy at a young age and protection against IBD. They also found that the mechanism for protection involves downregulation of the helper T cell (T_H17) pathway,²⁵ which also has been found to play a role in psoriasis pathogenesis.^26,27 Although our results suggest that the risk for appendicitis is not increased for patients with psoriasis, further research might be able to determine if appendicitis and subsequent appendectomy also can offer protection against development of psoriasis.

We found that patients with severe psoriasis had a higher incidence rate of cholecystitis compared to patients with mild psoriasis. Egeberg et al²⁸ found an increased risk for cholelithiasis among patients with psoriasis, which may contribute to a higher rate of cholecystitis. Although both acute and chronic cholecystitis were incorporated in this study, a Russian study found that chronic cholecystitis may be a predictor of progression of psoriasis.²⁹ Moreover, patients with severe psoriasis had a shorter duration to diagnosis of cholecystitis than patients with mild psoriasis. It is possible that patients with severe psoriasis are in a state of greater chronic inflammation than those with mild psoriasis, and therefore, when combined with other risk factors for cholecystitis, may progress to disease more quickly. Alternatively, this finding could be treatment related, as there have been reported cases of cholecystitis related to etanercept use in patients treated for psoriasis and juvenile polyarticular rheumatoid arthritis.^30,31 The relationship is not yet well defined, however, and further research is necessary to evaluate this association.

Study Strengths
Key strengths of this study include the large sample size and diversity of the patient population. Kaiser Permanente Southern California membership generally is representative of the broader community, making our results fairly generalizable to populations with health insurance. Use of a matched control cohort allows the results to be more specific to the disease of interest, and the population-based design minimizes bias.

Study Limitations
This study has several limitations. Although the cohorts were categorized based on type of treatment received, exact therapies were not specified. As a retrospective study, it is difficult to control for potential confounding variables that are not included in the electronic medical record. The results of this study also demonstrated significantly shorter durations to diagnosis of all 3 conditions, indicating that surveillance bias may be present.

Conclusion

Patients with psoriasis may be at an increased risk for diverticulitis compared to patients without psoriasis, which could be due to the chronic inflammatory state induced by psoriasis. Therefore, it may be beneficial for clinicians to evaluate psoriasis patients for other risk factors for diverticulitis and subsequently provide counseling to these patients to minimize their risk for diverticulitis. Psoriasis patients do not appear to be at an increased risk for appendicitis or cholecystitis compared to controls; however, further research is needed for confirmation.

Patients at risk for clinical deterioration in the inpatient setting may not be identified efficiently or effectively by health care providers. Early warning systems that link clinical observations to rapid response mechanisms (such as medical emergency teams) have the potential to improve outcomes, but rigorous studies are lacking.¹ The pediatric Rothman Index (pRI) is an automated early warning system sold by the company PeraHealth that is integrated with the electronic health record. The system incorporates vital signs, labs, and nursing assessments from existing electronic health record data to provide a single numeric score that generates alerts based on low absolute scores and acute decreases in score (low scores indicate high mortality risk).² Automated alerts or rules based on the pRI score are meant to bring important changes in clinical status to the attention of clinicians.

Adverse outcomes (eg, unplanned intensive care unit [ICU] transfers and mortality) are associated with low pRI scores, and scores appear to decline prior to such events.² However, the limitation of this and other studies evaluating the sensitivity of early warning systems^3-6 is that the generated alerts are assigned “true positive” status if they precede clinical deterioration, regardless of whether or not they provide meaningful information to the clinicians caring for the patients. There are two potential critiques of this approach. First, the alert may have preceded a deterioration event but may not have been clinically relevant (eg, an alert triggered by a finding unrelated to the patient’s acute health status, such as a scar that was newly documented as an abnormal skin finding and as a result led to a worsening in the pRI). Second, even if the preceding alert demonstrated clinical relevance to a deterioration event, the clinicians at the bedside may have been aware of the patient’s deterioration for hours and have already escalated care. In this situation, the alert would simply confirm what the clinician already knew.

To better understand the relationship between early warning system acuity alerts and clinical practice, we examined a cohort of hospitalized patients who experienced a critical deterioration event (CDE)⁷ and who would have triggered a preceding pRI alert. We evaluated the clinical relationship of the alert to the CDE (ie, whether the alert reflected physiologic changes related to a CDE or was instead an artifact of documentation) and identified whether the alert would have preceded evidence that clinicians recognized deterioration or escalated care.

 

METHODS

Patients and Setting

This retrospective cross-sectional study was performed at Children’s Hospital of Philadelphia (CHOP), a freestanding children’s hospital with 546 beds. Eligible patients were hospitalized on nonintensive care, noncardiology, surgical wards between January 1, 2013, and December 31, 2013. The CHOP Institutional Review Board (IRB) approved the study with waivers of consent and assent. A HIPAA Business Associate Agreement and an IRB Reliance Agreement were in place with PeraHealth to permit data transfer.

Definition of Critical Deterioration Events

Critical deterioration events (CDEs) were defined according to an existing, validated measure⁷ as unplanned transfers to the ICU with continuous or bilevel positive airway pressure, tracheal intubation, and/or vasopressor infusion in the 12 hours after transfer. At CHOP, all unplanned ICU transfers are routed through the hospital’s rapid response or code blue teams, so these patients were identified using an existing database managed by the CHOP Resuscitation Committee. In the database, the elements of CDEs are entered as part of ongoing quality improvement activities. The time of CDE was defined as the time of the rapid response call precipitating unplanned transfer to the ICU.

The Pediatric Rothman Index

The pRI is an automated acuity score that has been validated in hospitalized pediatric patients.² The pRI is calculated using existing variables from the electronic health record, including manually entered vital signs, laboratory values, cardiac rhythm, and nursing assessments of organ systems. The weights assigned to continuous variables are a function of deviation from the norm.^2,8 (See Supplement 1 for a complete list of variables.)

The pRI is integrated with the electronic health record and automatically generates a score each time a new data observation becomes available. Changes in score over time and low absolute scores generate a graduated series of alerts ranging from medium to very high acuity. This analysis used PeraHealth’s standard pRI alerts. Medium acuity alerts occurred when the pRI score decreased by ≥30% in 24 hours. A high acuity alert occurred when the pRI score decreased by ≥40% in 6 hours. A very high acuity alert occurred when the pRI absolute score was ≤ 30.

Development of the Source Dataset

In 2014, CHOP shared one year of clinical data with PeraHealth as part of the process of deciding whether or not to implement the pRI. The pRI algorithm retrospectively generated scores and acuity alerts for all CHOP patients who experienced CDEs between January 1, 2013, and December 31, 2013. The pRI algorithm was not active in the hospital environment during this time period; the scores and acuity alerts were not visible to clinicians. This dataset was provided to the investigators at CHOP to conduct this project.

Data Collection

Pediatric intensive care nurses trained in clinical research data abstraction from the CHOP Critical Care Center for Evidence and Outcomes performed the chart review for this study. Chart abstraction comparisons were completed on the first 15 charts to ensure interrater reliability, and additional quality assurance checks were performed on intermittent charts to ensure consistency and definition adherence. We managed all data using Research Electronic Data Capture.⁹

 

To study the value of alerts labeled as “true positives,” we restricted the dataset to CDEs in which acuity alert(s) within the prior 72 hours would have been triggered if the pRI had been in clinical use at the time.

To identify the clinical relationship between pRI and CDE, we reviewed each chart with the goal of determining whether the preceding acuity alerts were clinically associated with the etiology of the CDE. We determined the etiology of the CDE by reviewing the cause(s) identified in the note written by rapid response or code blue team responders or by the admitting clinical team after transfer to the ICU. We then used a tool provided by PeraHealth to identify the specific score components that led to worsening pRI. If the score components that worsened were (a) consistent with a clinical change as opposed to a documentation artifact and (b) an organ system change that was plausibly related to the CDE etiology, we concluded that the alert was clinically related to the etiology of the CDE.

We defined documentation artifacts as instances in nursing documentation in which a finding unrelated to the patient’s acute health status, such as a scar, was newly documented as abnormal and led to worsening pRI. Any cases in which the clinical relevance was unclear underwent review by additional members of the team, and the determination was made by consensus.

To determine the temporal relationship among pRI, CDE, and clinician awareness or action, we then sought to systematically determine whether the preceding acuity alerts preceded documented evidence of clinicians recognizing deterioration or escalation of care. We made the a priori decision that acuity alerts that occurred more than 24 hours prior to a deterioration event had questionable clinical actionability. Therefore, we restricted this next analysis to CDEs with acuity alerts during the 24 hours prior to a CDE. We reviewed time-stamped progress notes written by clinicians in the 24 hours period prior to the time of the CDE and identified whether the notes reflected an adverse change in patient status or a clinical intervention. We then compared the times of these notes with the times of the alerts and CDEs. Given that documentation of change in clinical status often occurs after clinical intervention, we also reviewed new orders placed in the 24 hours prior to each CDE to determine escalation of care. We identified the following orders as reflective of escalation of care independent of specific disease process: administration of intravenous fluid bolus, blood product, steroid, or antibiotic, increased respiratory support, new imaging studies, and new laboratory studies. We then compared the time of each order with the time of the alert and CDE.

RESULTS

During the study period, 73 events met the CDE criteria and had a pRI alert during admission. Of the 73 events, 50 would have triggered at least one pRI alert in the 72-hour period leading up to the CDE (sensitivity 68%). Of the 50 events, 39 generated pRI alerts in the 24 hours leading up to the event, and 11 others generated pRI alerts between 24 and 72 hours prior to the event but did not generate any alerts during the 24 hours leading up to the event (Figure).

 

Patient Characteristics

The 50 CDEs labeled as true positives occurred in 46 unique patients. Table 1 displays the event characteristics.

Acuity Alerts

A total of 79 pRI alerts preceded the 50 CDEs. Of these acuity alerts, 44 (56%) were medium acuity alerts, 17 (22%) were high acuity alerts, and 18 (23%) were very high acuity alerts. Of the 50 CDEs that would have triggered pRI alerts, 33 (66%) would have triggered a single acuity alert and 17 (34%) would have triggered multiple acuity alerts.

Of the 50 CDEs, 39 (78%) had a preceding acuity alert within 24 hours prior to the CDE. In these cases, the alert preceded the CDE by a median of 3.1 hours (interquartile range of 0.7 to 10.3 hours).

We assessed the score components that caused each alert to trigger. All of the vital sign and laboratory components were assessed as clinically related to the CDE’s etiology. By contrast, about half of nursing assessment components were assessed as clinically related to the etiology of the CDE (Table 2). Abnormal cardiac, respiratory, and neurologic assessments were most frequently assessed as clinically relevant.

Escalation Orders

To determine whether the pRI alert would have preceded the earliest documented treatment efforts, we restricted evaluation to the 39 CDEs that had at least one alert in the 24-hour window prior to the CDE. When we reviewed escalation orders placed by clinicians, we found that in 26 cases (67%), the first clinician order reflecting escalation of care would have preceded the first pRI alert within the 24-hour period prior to the CDE. In 13 cases (33%), the first pRI alert would have preceded the first escalation order placed by the clinician. The first pRI alert and the first escalation order would have occurred within the same 1-hour period in 6 of these cases.

Provider Notes

When we reviewed clinician notes for the 39 CDEs that had at least one alert in the 24-hour window prior to the CDE, we found that in 36 cases, there were preceding notes documenting adverse changes in patient status consistent with signs of deterioration or clinical intervention. In 30 cases (77%), the first clinician note preceded the first pRI alert within the 24-hour period prior to the CDE. In nine cases (23%), the first pRI alert would have preceded the first note. The first pRI alert and the first note would have occurred within the same 1-hour period in 4 of these cases.

Temporal Relationships

In Supplement 2, we present the proportion of CDEs in which the order or note preceded the pRI alert for each abnormal organ system.

The Figure shows the temporal relationships among escalation orders, clinician notes, and acuity alerts for the 39 CDEs with one or more alerts in the 24 hours leading up to the event. In 21 cases (54%), both an escalation order and a note preceded the first acuity alert. In 14 cases (36%), either an escalation order or a note preceded the first acuity alert. In four cases (10%), the alert preceded any documented evidence that clinicians had recognized deterioration or escalating care.

 

DISCUSSION

The main finding of this study is that 90% of CDE events that generated “true positive” pRI alerts had evidence suggesting that clinicians had already recognized deterioration and/or were already escalating care before most pRI alerts would have been triggered.

The impacts of early warning scores on patient safety outcomes are not well established. In a recent 21-hospital cluster randomized trial of the BedsidePEWS, a pediatric early warning score system, investigators found that implementing the system does not significantly decrease all-cause mortality in hospitalized children, although hospitals using the BedsidePEWS have low rates of significant CDEs.¹⁰ In other studies, early warning scores were often coimplemented with rapid response teams, and separating the incremental benefit of the scoring tool from the availability of a rapid response team is usually not possible.¹¹

Therefore, the benefits of early warning scores are often inferred based on their test characteristics (eg, sensitivity and positive predictive value).¹² Sensitivity, which is the proportion of patients who deteriorated and also triggered the early warning score within a reasonable time window preceding the event, is an important consideration when deciding whether an early warning score is worth implementing. A challenging follow-up question that goes beyond sensitivity is how often an early warning score adds new knowledge by identifying patients on a path toward deterioration who were not yet recognized. This study is the first to address that follow-up question. Our results revealed that the score appeared to precede evidence of clinician recognition of deterioration in 10% of CDEs. In some patients, the alert could have contributed to a detection of deterioration that was not previously evident. In the portion of CDEs in which the alert and escalation order or note occurred within the same one-hour window, the alert could have been used as confirmation of clinical suspicion. Notably, we did not evaluate the 16 cases in which a CDE preceded any pRI alert because we chose to focus on “true positive” cases in which pRI alerts preceded CDEs. These events could have had timely recognition by clinicians that we did not capture, so these results may provide an overestimation of CDEs in which the pRI preceded clinician recognition.

Prior work has described a range of mechanisms by which early warning scores can impact patient safety.¹³ The results of this study suggest limited incremental benefit for the pRI to alert physicians and nurses to new concerning changes at this hospital, although the benefits to low-resourced community hospitals that care for children may be great. The pRI score may also serve as evidence that empowers nurses to overcome barriers to further escalate care, even if the process of escalation has already begun. In addition to empowering nurses, the score may support trainees and clinicians with varying levels of pediatric expertise in the decision to escalate care. Evaluating these potential benefits would require prospective study.

We used the pRI alerts as they were already defined by PeraHealth for CHOP, and different alert thresholds may change score performance. Our study did not identify additional variables to improve score performance, but they can be investigated in future research.

This study had several limitations. First, this work is a single-center study with highly skilled pediatric providers, a mature rapid response system, and low rates of cardiopulmonary arrest outside ICUs. Therefore, the results that we obtained were not immediately generalizable. In a community environment with nurses and physicians who are less experienced in caring for ill children, an early warning score with high sensitivity may be beneficial in ensuring patient safety.

Second, by using escalation orders and notes from the patient chart, we did not capture all the undocumented ways in which clinicians demonstrate awareness of deterioration. For example, a resident may alert the attending on service or a team may informally request consultation with a specialist. We also gave equal weight to escalation orders and clinician notes as evidence of recognition of deterioration. It could be that either orders or notes more closely correlated with clinician awareness.

Finally, the data were from 2013. Although the score components have not changed, efforts to standardize nursing assessments may have altered the performance of the score in the intervening years.

 

CONCLUSIONS

In most patients who had a CDE at a large freestanding children’s hospital, escalation orders or documented changes in patient status would have occurred before a pRI alert. However, in a minority of patients, the alert could have contributed to the detection of deterioration that was not previously evident.

Disclosures

The authors have nothing to disclose

Funding

The study was supported by funds from the Department of Biomedical and Health Informatics at Children’s Hospital of Philadelphia. PeraHealth, the company that sells the Rothman Index software, provided a service to the investigators but no funding. They applied their proprietary scoring algorithm to the data from Children’s Hospital of Philadelphia to generate alerts retrospectively. This service was provided free of charge in 2014 during the time period when Children’s Hospital of Philadelphia was considering purchasing and implementing PeraHealth software, which it subsequently did. We did not receive any funding for the study from PeraHealth. PeraHealth personnel did not influence the study design, the interpretation of data, the writing of the report, or the decision to submit the article for publication.

 

Patients at risk for clinical deterioration in the inpatient setting may not be identified efficiently or effectively by health care providers. Early warning systems that link clinical observations to rapid response mechanisms (such as medical emergency teams) have the potential to improve outcomes, but rigorous studies are lacking.¹ The pediatric Rothman Index (pRI) is an automated early warning system sold by the company PeraHealth that is integrated with the electronic health record. The system incorporates vital signs, labs, and nursing assessments from existing electronic health record data to provide a single numeric score that generates alerts based on low absolute scores and acute decreases in score (low scores indicate high mortality risk).² Automated alerts or rules based on the pRI score are meant to bring important changes in clinical status to the attention of clinicians.

Adverse outcomes (eg, unplanned intensive care unit [ICU] transfers and mortality) are associated with low pRI scores, and scores appear to decline prior to such events.² However, the limitation of this and other studies evaluating the sensitivity of early warning systems^3-6 is that the generated alerts are assigned “true positive” status if they precede clinical deterioration, regardless of whether or not they provide meaningful information to the clinicians caring for the patients. There are two potential critiques of this approach. First, the alert may have preceded a deterioration event but may not have been clinically relevant (eg, an alert triggered by a finding unrelated to the patient’s acute health status, such as a scar that was newly documented as an abnormal skin finding and as a result led to a worsening in the pRI). Second, even if the preceding alert demonstrated clinical relevance to a deterioration event, the clinicians at the bedside may have been aware of the patient’s deterioration for hours and have already escalated care. In this situation, the alert would simply confirm what the clinician already knew.

To better understand the relationship between early warning system acuity alerts and clinical practice, we examined a cohort of hospitalized patients who experienced a critical deterioration event (CDE)⁷ and who would have triggered a preceding pRI alert. We evaluated the clinical relationship of the alert to the CDE (ie, whether the alert reflected physiologic changes related to a CDE or was instead an artifact of documentation) and identified whether the alert would have preceded evidence that clinicians recognized deterioration or escalated care.

 

METHODS

Patients and Setting

This retrospective cross-sectional study was performed at Children’s Hospital of Philadelphia (CHOP), a freestanding children’s hospital with 546 beds. Eligible patients were hospitalized on nonintensive care, noncardiology, surgical wards between January 1, 2013, and December 31, 2013. The CHOP Institutional Review Board (IRB) approved the study with waivers of consent and assent. A HIPAA Business Associate Agreement and an IRB Reliance Agreement were in place with PeraHealth to permit data transfer.

Definition of Critical Deterioration Events

Critical deterioration events (CDEs) were defined according to an existing, validated measure⁷ as unplanned transfers to the ICU with continuous or bilevel positive airway pressure, tracheal intubation, and/or vasopressor infusion in the 12 hours after transfer. At CHOP, all unplanned ICU transfers are routed through the hospital’s rapid response or code blue teams, so these patients were identified using an existing database managed by the CHOP Resuscitation Committee. In the database, the elements of CDEs are entered as part of ongoing quality improvement activities. The time of CDE was defined as the time of the rapid response call precipitating unplanned transfer to the ICU.

The Pediatric Rothman Index

The pRI is an automated acuity score that has been validated in hospitalized pediatric patients.² The pRI is calculated using existing variables from the electronic health record, including manually entered vital signs, laboratory values, cardiac rhythm, and nursing assessments of organ systems. The weights assigned to continuous variables are a function of deviation from the norm.^2,8 (See Supplement 1 for a complete list of variables.)

The pRI is integrated with the electronic health record and automatically generates a score each time a new data observation becomes available. Changes in score over time and low absolute scores generate a graduated series of alerts ranging from medium to very high acuity. This analysis used PeraHealth’s standard pRI alerts. Medium acuity alerts occurred when the pRI score decreased by ≥30% in 24 hours. A high acuity alert occurred when the pRI score decreased by ≥40% in 6 hours. A very high acuity alert occurred when the pRI absolute score was ≤ 30.

Development of the Source Dataset

In 2014, CHOP shared one year of clinical data with PeraHealth as part of the process of deciding whether or not to implement the pRI. The pRI algorithm retrospectively generated scores and acuity alerts for all CHOP patients who experienced CDEs between January 1, 2013, and December 31, 2013. The pRI algorithm was not active in the hospital environment during this time period; the scores and acuity alerts were not visible to clinicians. This dataset was provided to the investigators at CHOP to conduct this project.

Data Collection

Pediatric intensive care nurses trained in clinical research data abstraction from the CHOP Critical Care Center for Evidence and Outcomes performed the chart review for this study. Chart abstraction comparisons were completed on the first 15 charts to ensure interrater reliability, and additional quality assurance checks were performed on intermittent charts to ensure consistency and definition adherence. We managed all data using Research Electronic Data Capture.⁹

 

To study the value of alerts labeled as “true positives,” we restricted the dataset to CDEs in which acuity alert(s) within the prior 72 hours would have been triggered if the pRI had been in clinical use at the time.

To identify the clinical relationship between pRI and CDE, we reviewed each chart with the goal of determining whether the preceding acuity alerts were clinically associated with the etiology of the CDE. We determined the etiology of the CDE by reviewing the cause(s) identified in the note written by rapid response or code blue team responders or by the admitting clinical team after transfer to the ICU. We then used a tool provided by PeraHealth to identify the specific score components that led to worsening pRI. If the score components that worsened were (a) consistent with a clinical change as opposed to a documentation artifact and (b) an organ system change that was plausibly related to the CDE etiology, we concluded that the alert was clinically related to the etiology of the CDE.

We defined documentation artifacts as instances in nursing documentation in which a finding unrelated to the patient’s acute health status, such as a scar, was newly documented as abnormal and led to worsening pRI. Any cases in which the clinical relevance was unclear underwent review by additional members of the team, and the determination was made by consensus.

To determine the temporal relationship among pRI, CDE, and clinician awareness or action, we then sought to systematically determine whether the preceding acuity alerts preceded documented evidence of clinicians recognizing deterioration or escalation of care. We made the a priori decision that acuity alerts that occurred more than 24 hours prior to a deterioration event had questionable clinical actionability. Therefore, we restricted this next analysis to CDEs with acuity alerts during the 24 hours prior to a CDE. We reviewed time-stamped progress notes written by clinicians in the 24 hours period prior to the time of the CDE and identified whether the notes reflected an adverse change in patient status or a clinical intervention. We then compared the times of these notes with the times of the alerts and CDEs. Given that documentation of change in clinical status often occurs after clinical intervention, we also reviewed new orders placed in the 24 hours prior to each CDE to determine escalation of care. We identified the following orders as reflective of escalation of care independent of specific disease process: administration of intravenous fluid bolus, blood product, steroid, or antibiotic, increased respiratory support, new imaging studies, and new laboratory studies. We then compared the time of each order with the time of the alert and CDE.

RESULTS

During the study period, 73 events met the CDE criteria and had a pRI alert during admission. Of the 73 events, 50 would have triggered at least one pRI alert in the 72-hour period leading up to the CDE (sensitivity 68%). Of the 50 events, 39 generated pRI alerts in the 24 hours leading up to the event, and 11 others generated pRI alerts between 24 and 72 hours prior to the event but did not generate any alerts during the 24 hours leading up to the event (Figure).

 

Patient Characteristics

The 50 CDEs labeled as true positives occurred in 46 unique patients. Table 1 displays the event characteristics.

Acuity Alerts

A total of 79 pRI alerts preceded the 50 CDEs. Of these acuity alerts, 44 (56%) were medium acuity alerts, 17 (22%) were high acuity alerts, and 18 (23%) were very high acuity alerts. Of the 50 CDEs that would have triggered pRI alerts, 33 (66%) would have triggered a single acuity alert and 17 (34%) would have triggered multiple acuity alerts.

Of the 50 CDEs, 39 (78%) had a preceding acuity alert within 24 hours prior to the CDE. In these cases, the alert preceded the CDE by a median of 3.1 hours (interquartile range of 0.7 to 10.3 hours).

We assessed the score components that caused each alert to trigger. All of the vital sign and laboratory components were assessed as clinically related to the CDE’s etiology. By contrast, about half of nursing assessment components were assessed as clinically related to the etiology of the CDE (Table 2). Abnormal cardiac, respiratory, and neurologic assessments were most frequently assessed as clinically relevant.

Escalation Orders

To determine whether the pRI alert would have preceded the earliest documented treatment efforts, we restricted evaluation to the 39 CDEs that had at least one alert in the 24-hour window prior to the CDE. When we reviewed escalation orders placed by clinicians, we found that in 26 cases (67%), the first clinician order reflecting escalation of care would have preceded the first pRI alert within the 24-hour period prior to the CDE. In 13 cases (33%), the first pRI alert would have preceded the first escalation order placed by the clinician. The first pRI alert and the first escalation order would have occurred within the same 1-hour period in 6 of these cases.

Provider Notes

When we reviewed clinician notes for the 39 CDEs that had at least one alert in the 24-hour window prior to the CDE, we found that in 36 cases, there were preceding notes documenting adverse changes in patient status consistent with signs of deterioration or clinical intervention. In 30 cases (77%), the first clinician note preceded the first pRI alert within the 24-hour period prior to the CDE. In nine cases (23%), the first pRI alert would have preceded the first note. The first pRI alert and the first note would have occurred within the same 1-hour period in 4 of these cases.

Temporal Relationships

In Supplement 2, we present the proportion of CDEs in which the order or note preceded the pRI alert for each abnormal organ system.

The Figure shows the temporal relationships among escalation orders, clinician notes, and acuity alerts for the 39 CDEs with one or more alerts in the 24 hours leading up to the event. In 21 cases (54%), both an escalation order and a note preceded the first acuity alert. In 14 cases (36%), either an escalation order or a note preceded the first acuity alert. In four cases (10%), the alert preceded any documented evidence that clinicians had recognized deterioration or escalating care.

 

DISCUSSION

The main finding of this study is that 90% of CDE events that generated “true positive” pRI alerts had evidence suggesting that clinicians had already recognized deterioration and/or were already escalating care before most pRI alerts would have been triggered.

The impacts of early warning scores on patient safety outcomes are not well established. In a recent 21-hospital cluster randomized trial of the BedsidePEWS, a pediatric early warning score system, investigators found that implementing the system does not significantly decrease all-cause mortality in hospitalized children, although hospitals using the BedsidePEWS have low rates of significant CDEs.¹⁰ In other studies, early warning scores were often coimplemented with rapid response teams, and separating the incremental benefit of the scoring tool from the availability of a rapid response team is usually not possible.¹¹

Therefore, the benefits of early warning scores are often inferred based on their test characteristics (eg, sensitivity and positive predictive value).¹² Sensitivity, which is the proportion of patients who deteriorated and also triggered the early warning score within a reasonable time window preceding the event, is an important consideration when deciding whether an early warning score is worth implementing. A challenging follow-up question that goes beyond sensitivity is how often an early warning score adds new knowledge by identifying patients on a path toward deterioration who were not yet recognized. This study is the first to address that follow-up question. Our results revealed that the score appeared to precede evidence of clinician recognition of deterioration in 10% of CDEs. In some patients, the alert could have contributed to a detection of deterioration that was not previously evident. In the portion of CDEs in which the alert and escalation order or note occurred within the same one-hour window, the alert could have been used as confirmation of clinical suspicion. Notably, we did not evaluate the 16 cases in which a CDE preceded any pRI alert because we chose to focus on “true positive” cases in which pRI alerts preceded CDEs. These events could have had timely recognition by clinicians that we did not capture, so these results may provide an overestimation of CDEs in which the pRI preceded clinician recognition.

Prior work has described a range of mechanisms by which early warning scores can impact patient safety.¹³ The results of this study suggest limited incremental benefit for the pRI to alert physicians and nurses to new concerning changes at this hospital, although the benefits to low-resourced community hospitals that care for children may be great. The pRI score may also serve as evidence that empowers nurses to overcome barriers to further escalate care, even if the process of escalation has already begun. In addition to empowering nurses, the score may support trainees and clinicians with varying levels of pediatric expertise in the decision to escalate care. Evaluating these potential benefits would require prospective study.

We used the pRI alerts as they were already defined by PeraHealth for CHOP, and different alert thresholds may change score performance. Our study did not identify additional variables to improve score performance, but they can be investigated in future research.

This study had several limitations. First, this work is a single-center study with highly skilled pediatric providers, a mature rapid response system, and low rates of cardiopulmonary arrest outside ICUs. Therefore, the results that we obtained were not immediately generalizable. In a community environment with nurses and physicians who are less experienced in caring for ill children, an early warning score with high sensitivity may be beneficial in ensuring patient safety.

Second, by using escalation orders and notes from the patient chart, we did not capture all the undocumented ways in which clinicians demonstrate awareness of deterioration. For example, a resident may alert the attending on service or a team may informally request consultation with a specialist. We also gave equal weight to escalation orders and clinician notes as evidence of recognition of deterioration. It could be that either orders or notes more closely correlated with clinician awareness.

Finally, the data were from 2013. Although the score components have not changed, efforts to standardize nursing assessments may have altered the performance of the score in the intervening years.

 

CONCLUSIONS

In most patients who had a CDE at a large freestanding children’s hospital, escalation orders or documented changes in patient status would have occurred before a pRI alert. However, in a minority of patients, the alert could have contributed to the detection of deterioration that was not previously evident.

Disclosures

The authors have nothing to disclose

Funding

The study was supported by funds from the Department of Biomedical and Health Informatics at Children’s Hospital of Philadelphia. PeraHealth, the company that sells the Rothman Index software, provided a service to the investigators but no funding. They applied their proprietary scoring algorithm to the data from Children’s Hospital of Philadelphia to generate alerts retrospectively. This service was provided free of charge in 2014 during the time period when Children’s Hospital of Philadelphia was considering purchasing and implementing PeraHealth software, which it subsequently did. We did not receive any funding for the study from PeraHealth. PeraHealth personnel did not influence the study design, the interpretation of data, the writing of the report, or the decision to submit the article for publication.

 

Patients at risk for clinical deterioration in the inpatient setting may not be identified efficiently or effectively by health care providers. Early warning systems that link clinical observations to rapid response mechanisms (such as medical emergency teams) have the potential to improve outcomes, but rigorous studies are lacking.¹ The pediatric Rothman Index (pRI) is an automated early warning system sold by the company PeraHealth that is integrated with the electronic health record. The system incorporates vital signs, labs, and nursing assessments from existing electronic health record data to provide a single numeric score that generates alerts based on low absolute scores and acute decreases in score (low scores indicate high mortality risk).² Automated alerts or rules based on the pRI score are meant to bring important changes in clinical status to the attention of clinicians.

Adverse outcomes (eg, unplanned intensive care unit [ICU] transfers and mortality) are associated with low pRI scores, and scores appear to decline prior to such events.² However, the limitation of this and other studies evaluating the sensitivity of early warning systems^3-6 is that the generated alerts are assigned “true positive” status if they precede clinical deterioration, regardless of whether or not they provide meaningful information to the clinicians caring for the patients. There are two potential critiques of this approach. First, the alert may have preceded a deterioration event but may not have been clinically relevant (eg, an alert triggered by a finding unrelated to the patient’s acute health status, such as a scar that was newly documented as an abnormal skin finding and as a result led to a worsening in the pRI). Second, even if the preceding alert demonstrated clinical relevance to a deterioration event, the clinicians at the bedside may have been aware of the patient’s deterioration for hours and have already escalated care. In this situation, the alert would simply confirm what the clinician already knew.

To better understand the relationship between early warning system acuity alerts and clinical practice, we examined a cohort of hospitalized patients who experienced a critical deterioration event (CDE)⁷ and who would have triggered a preceding pRI alert. We evaluated the clinical relationship of the alert to the CDE (ie, whether the alert reflected physiologic changes related to a CDE or was instead an artifact of documentation) and identified whether the alert would have preceded evidence that clinicians recognized deterioration or escalated care.

 

METHODS

Patients and Setting

This retrospective cross-sectional study was performed at Children’s Hospital of Philadelphia (CHOP), a freestanding children’s hospital with 546 beds. Eligible patients were hospitalized on nonintensive care, noncardiology, surgical wards between January 1, 2013, and December 31, 2013. The CHOP Institutional Review Board (IRB) approved the study with waivers of consent and assent. A HIPAA Business Associate Agreement and an IRB Reliance Agreement were in place with PeraHealth to permit data transfer.

Definition of Critical Deterioration Events

Critical deterioration events (CDEs) were defined according to an existing, validated measure⁷ as unplanned transfers to the ICU with continuous or bilevel positive airway pressure, tracheal intubation, and/or vasopressor infusion in the 12 hours after transfer. At CHOP, all unplanned ICU transfers are routed through the hospital’s rapid response or code blue teams, so these patients were identified using an existing database managed by the CHOP Resuscitation Committee. In the database, the elements of CDEs are entered as part of ongoing quality improvement activities. The time of CDE was defined as the time of the rapid response call precipitating unplanned transfer to the ICU.

The Pediatric Rothman Index

The pRI is an automated acuity score that has been validated in hospitalized pediatric patients.² The pRI is calculated using existing variables from the electronic health record, including manually entered vital signs, laboratory values, cardiac rhythm, and nursing assessments of organ systems. The weights assigned to continuous variables are a function of deviation from the norm.^2,8 (See Supplement 1 for a complete list of variables.)

The pRI is integrated with the electronic health record and automatically generates a score each time a new data observation becomes available. Changes in score over time and low absolute scores generate a graduated series of alerts ranging from medium to very high acuity. This analysis used PeraHealth’s standard pRI alerts. Medium acuity alerts occurred when the pRI score decreased by ≥30% in 24 hours. A high acuity alert occurred when the pRI score decreased by ≥40% in 6 hours. A very high acuity alert occurred when the pRI absolute score was ≤ 30.

Development of the Source Dataset

In 2014, CHOP shared one year of clinical data with PeraHealth as part of the process of deciding whether or not to implement the pRI. The pRI algorithm retrospectively generated scores and acuity alerts for all CHOP patients who experienced CDEs between January 1, 2013, and December 31, 2013. The pRI algorithm was not active in the hospital environment during this time period; the scores and acuity alerts were not visible to clinicians. This dataset was provided to the investigators at CHOP to conduct this project.

Data Collection

Pediatric intensive care nurses trained in clinical research data abstraction from the CHOP Critical Care Center for Evidence and Outcomes performed the chart review for this study. Chart abstraction comparisons were completed on the first 15 charts to ensure interrater reliability, and additional quality assurance checks were performed on intermittent charts to ensure consistency and definition adherence. We managed all data using Research Electronic Data Capture.⁹

 

To study the value of alerts labeled as “true positives,” we restricted the dataset to CDEs in which acuity alert(s) within the prior 72 hours would have been triggered if the pRI had been in clinical use at the time.

To identify the clinical relationship between pRI and CDE, we reviewed each chart with the goal of determining whether the preceding acuity alerts were clinically associated with the etiology of the CDE. We determined the etiology of the CDE by reviewing the cause(s) identified in the note written by rapid response or code blue team responders or by the admitting clinical team after transfer to the ICU. We then used a tool provided by PeraHealth to identify the specific score components that led to worsening pRI. If the score components that worsened were (a) consistent with a clinical change as opposed to a documentation artifact and (b) an organ system change that was plausibly related to the CDE etiology, we concluded that the alert was clinically related to the etiology of the CDE.

We defined documentation artifacts as instances in nursing documentation in which a finding unrelated to the patient’s acute health status, such as a scar, was newly documented as abnormal and led to worsening pRI. Any cases in which the clinical relevance was unclear underwent review by additional members of the team, and the determination was made by consensus.

To determine the temporal relationship among pRI, CDE, and clinician awareness or action, we then sought to systematically determine whether the preceding acuity alerts preceded documented evidence of clinicians recognizing deterioration or escalation of care. We made the a priori decision that acuity alerts that occurred more than 24 hours prior to a deterioration event had questionable clinical actionability. Therefore, we restricted this next analysis to CDEs with acuity alerts during the 24 hours prior to a CDE. We reviewed time-stamped progress notes written by clinicians in the 24 hours period prior to the time of the CDE and identified whether the notes reflected an adverse change in patient status or a clinical intervention. We then compared the times of these notes with the times of the alerts and CDEs. Given that documentation of change in clinical status often occurs after clinical intervention, we also reviewed new orders placed in the 24 hours prior to each CDE to determine escalation of care. We identified the following orders as reflective of escalation of care independent of specific disease process: administration of intravenous fluid bolus, blood product, steroid, or antibiotic, increased respiratory support, new imaging studies, and new laboratory studies. We then compared the time of each order with the time of the alert and CDE.

RESULTS

During the study period, 73 events met the CDE criteria and had a pRI alert during admission. Of the 73 events, 50 would have triggered at least one pRI alert in the 72-hour period leading up to the CDE (sensitivity 68%). Of the 50 events, 39 generated pRI alerts in the 24 hours leading up to the event, and 11 others generated pRI alerts between 24 and 72 hours prior to the event but did not generate any alerts during the 24 hours leading up to the event (Figure).

 

Patient Characteristics

The 50 CDEs labeled as true positives occurred in 46 unique patients. Table 1 displays the event characteristics.

Acuity Alerts

A total of 79 pRI alerts preceded the 50 CDEs. Of these acuity alerts, 44 (56%) were medium acuity alerts, 17 (22%) were high acuity alerts, and 18 (23%) were very high acuity alerts. Of the 50 CDEs that would have triggered pRI alerts, 33 (66%) would have triggered a single acuity alert and 17 (34%) would have triggered multiple acuity alerts.

Of the 50 CDEs, 39 (78%) had a preceding acuity alert within 24 hours prior to the CDE. In these cases, the alert preceded the CDE by a median of 3.1 hours (interquartile range of 0.7 to 10.3 hours).

We assessed the score components that caused each alert to trigger. All of the vital sign and laboratory components were assessed as clinically related to the CDE’s etiology. By contrast, about half of nursing assessment components were assessed as clinically related to the etiology of the CDE (Table 2). Abnormal cardiac, respiratory, and neurologic assessments were most frequently assessed as clinically relevant.

Escalation Orders

To determine whether the pRI alert would have preceded the earliest documented treatment efforts, we restricted evaluation to the 39 CDEs that had at least one alert in the 24-hour window prior to the CDE. When we reviewed escalation orders placed by clinicians, we found that in 26 cases (67%), the first clinician order reflecting escalation of care would have preceded the first pRI alert within the 24-hour period prior to the CDE. In 13 cases (33%), the first pRI alert would have preceded the first escalation order placed by the clinician. The first pRI alert and the first escalation order would have occurred within the same 1-hour period in 6 of these cases.

Provider Notes

When we reviewed clinician notes for the 39 CDEs that had at least one alert in the 24-hour window prior to the CDE, we found that in 36 cases, there were preceding notes documenting adverse changes in patient status consistent with signs of deterioration or clinical intervention. In 30 cases (77%), the first clinician note preceded the first pRI alert within the 24-hour period prior to the CDE. In nine cases (23%), the first pRI alert would have preceded the first note. The first pRI alert and the first note would have occurred within the same 1-hour period in 4 of these cases.

Temporal Relationships

In Supplement 2, we present the proportion of CDEs in which the order or note preceded the pRI alert for each abnormal organ system.

The Figure shows the temporal relationships among escalation orders, clinician notes, and acuity alerts for the 39 CDEs with one or more alerts in the 24 hours leading up to the event. In 21 cases (54%), both an escalation order and a note preceded the first acuity alert. In 14 cases (36%), either an escalation order or a note preceded the first acuity alert. In four cases (10%), the alert preceded any documented evidence that clinicians had recognized deterioration or escalating care.

 

DISCUSSION

The main finding of this study is that 90% of CDE events that generated “true positive” pRI alerts had evidence suggesting that clinicians had already recognized deterioration and/or were already escalating care before most pRI alerts would have been triggered.

The impacts of early warning scores on patient safety outcomes are not well established. In a recent 21-hospital cluster randomized trial of the BedsidePEWS, a pediatric early warning score system, investigators found that implementing the system does not significantly decrease all-cause mortality in hospitalized children, although hospitals using the BedsidePEWS have low rates of significant CDEs.¹⁰ In other studies, early warning scores were often coimplemented with rapid response teams, and separating the incremental benefit of the scoring tool from the availability of a rapid response team is usually not possible.¹¹

Therefore, the benefits of early warning scores are often inferred based on their test characteristics (eg, sensitivity and positive predictive value).¹² Sensitivity, which is the proportion of patients who deteriorated and also triggered the early warning score within a reasonable time window preceding the event, is an important consideration when deciding whether an early warning score is worth implementing. A challenging follow-up question that goes beyond sensitivity is how often an early warning score adds new knowledge by identifying patients on a path toward deterioration who were not yet recognized. This study is the first to address that follow-up question. Our results revealed that the score appeared to precede evidence of clinician recognition of deterioration in 10% of CDEs. In some patients, the alert could have contributed to a detection of deterioration that was not previously evident. In the portion of CDEs in which the alert and escalation order or note occurred within the same one-hour window, the alert could have been used as confirmation of clinical suspicion. Notably, we did not evaluate the 16 cases in which a CDE preceded any pRI alert because we chose to focus on “true positive” cases in which pRI alerts preceded CDEs. These events could have had timely recognition by clinicians that we did not capture, so these results may provide an overestimation of CDEs in which the pRI preceded clinician recognition.

Prior work has described a range of mechanisms by which early warning scores can impact patient safety.¹³ The results of this study suggest limited incremental benefit for the pRI to alert physicians and nurses to new concerning changes at this hospital, although the benefits to low-resourced community hospitals that care for children may be great. The pRI score may also serve as evidence that empowers nurses to overcome barriers to further escalate care, even if the process of escalation has already begun. In addition to empowering nurses, the score may support trainees and clinicians with varying levels of pediatric expertise in the decision to escalate care. Evaluating these potential benefits would require prospective study.

We used the pRI alerts as they were already defined by PeraHealth for CHOP, and different alert thresholds may change score performance. Our study did not identify additional variables to improve score performance, but they can be investigated in future research.

This study had several limitations. First, this work is a single-center study with highly skilled pediatric providers, a mature rapid response system, and low rates of cardiopulmonary arrest outside ICUs. Therefore, the results that we obtained were not immediately generalizable. In a community environment with nurses and physicians who are less experienced in caring for ill children, an early warning score with high sensitivity may be beneficial in ensuring patient safety.

Second, by using escalation orders and notes from the patient chart, we did not capture all the undocumented ways in which clinicians demonstrate awareness of deterioration. For example, a resident may alert the attending on service or a team may informally request consultation with a specialist. We also gave equal weight to escalation orders and clinician notes as evidence of recognition of deterioration. It could be that either orders or notes more closely correlated with clinician awareness.

Finally, the data were from 2013. Although the score components have not changed, efforts to standardize nursing assessments may have altered the performance of the score in the intervening years.

 

CONCLUSIONS

In most patients who had a CDE at a large freestanding children’s hospital, escalation orders or documented changes in patient status would have occurred before a pRI alert. However, in a minority of patients, the alert could have contributed to the detection of deterioration that was not previously evident.

Disclosures

The authors have nothing to disclose

Funding

The study was supported by funds from the Department of Biomedical and Health Informatics at Children’s Hospital of Philadelphia. PeraHealth, the company that sells the Rothman Index software, provided a service to the investigators but no funding. They applied their proprietary scoring algorithm to the data from Children’s Hospital of Philadelphia to generate alerts retrospectively. This service was provided free of charge in 2014 during the time period when Children’s Hospital of Philadelphia was considering purchasing and implementing PeraHealth software, which it subsequently did. We did not receive any funding for the study from PeraHealth. PeraHealth personnel did not influence the study design, the interpretation of data, the writing of the report, or the decision to submit the article for publication.

 

With the presence of hypertension in 25% of patients admitted to the hospital,¹ its proper management is imperative. A hypertensive crisis is a severe elevation of blood pressure, defined as systolic ≥180 mm Hg and/or diastolic ≥120 mm Hg. It is further classified as either a hypertensive emergency which includes the presence of end-organ damage,² or hypertensive urgency, defined as asymptomatic blood pressure elevation.³ Although hypertensive emergencies account for only 1%-2% of patients with hypertension,⁴ they are associated with a high one-year mortality rate (>79%).⁵ Hypertensive emergency requires immediate reduction of blood pressure with IV antihypertensive drugs to limit organ damage. In contrast, as per national guidelines, inpatient management of hypertensive urgency requires gradual reductions of blood pressure over hours to days using oral antihypertensives.² It is also recommended that alternative etiologies, such as anxiety or pain, be considered before treatment is initiated.¹

Clinicians often inappropriately treat asymptomatic hypertension in the inpatient setting,^6,7 using intravenous (IV) antihypertensive medications despite evidence showing potential harm.^5,8 This can lead to unpredictable reductions in blood pressure.^7,9 A recent retrospective analysis demonstrated that 32.6% of patients had a blood pressure reduction greater than 25% after the use of an IV antihypertensive.⁷ Reductions greater than 25% lead to shifts in autoregulation, which may result in patient harm, such as hypotension, decreased renal perfusion, and stroke.⁹ IV medications are also more expensive than oral agents, due to the additional cost of administration.

Although overtreatment of asymptomatic hypertension with IV antihypertensive medications is common,⁷ initiatives to address this in inpatient settings are lacking in the literature. The aim of this quality improvement initiative was to reduce unnecessary IV antihypertensive treatment for hypertensive urgency in the inpatient setting.

METHODS

Setting

An interdisciplinary quality improvement intervention was initiated on two inpatient medicine units at an urban, 1,134-bed tertiary medical center affiliated with the Icahn School of Medicine at Mount Sinai. Members of the Mount Sinai High Value Care Committee and the Student High Value Care Initiative¹⁰ developed this project. The intervention was implemented in stages from March 2017 to February 2018. It targeted nurses, housestaff, nurse practitioners, and attendings on general medical teaching and nonteaching services. The components of the intervention included education, a treatment algorithm, audit and feedback, and electronic medical record (EMR) change. This project was submitted to the Quality Committee in the Department of Medicine and determined to be a quality improvement project rather than research and thus, an IRB submission was not required.

 

Treatment Algorithm and Education

A clinical algorithm was designed with nursing and cardiology representatives to provide guidance for nurses regarding the best practice for evaluation of inpatient hypertension, focusing on assessing patients before recommending treatment (“Assess Before Rx”; Figure 1). Educational sessions reinforcing the clinical algorithm were held monthly at nursing huddles. These involved an introduction session providing the background and purpose of the project, with follow-up sessions including interactive mock cases on the assessment of hypertensive urgency.

A second treatment algorithm was designed, with housestaff and cardiology input, to provide guidance for the internal medicine housestaff and nurse practitioners. It utilized a similar approach regarding identification, evaluation, and assessment of alternate etiologies but included more detailed treatment recommendations with a table outlining the oral medications used for hypertensive urgency (Figure 2). The flowchart and table were uploaded to an existing mobile application used by housestaff and nurse practitioners for quick access. The mobile application is frequently used by housestaff and contains many clinical resources. Additionally, e-mails including the purpose of the project and the treatment algorithm were sent to rotating housestaff at the start of each new medicine rotation.

Audit and Feedback

Monthly feedback was e-mailed to the nurses, which reinforced the goals and provided positive feedback on outcomes with an announcement of the “Nurse of the Month.” The winners were selected based on the most accurate and appropriate documentation of their assessments determined through retrospective chart review.

Targeted e-mail feedback was also sent to providers who ordered IV antihypertensives without the appropriate indication. The e-mails included the medical record number, date and time of the order, any alternate etiologies that were documented, and any adverse events that occurred as a result of the medication.

Systems Change: Electronic Medical Record Orders

EMR advisory warnings were placed on IV antihypertensive orders of labetalol and hydralazine. The alerts served to nonintrusively remind providers to assess for symptoms before placing the order to ensure that the order was appropriate.

Data Collection and Assessment

Seven-month preintervention (January-July 2016) and 12-month postintervention (March 2017-February 2018) data were compared. The months prior to intervention were excluded to account for project development and educational lag. Data were obtained from EMR utilization reports of one-time orders of IV labetalol and hydralazine, and retrospective chart review. Patients who were pregnant, less than 18 years of age, or postoperative were excluded. Orders were designated as inappropriate if there was no evidence of hypertensive emergency through documentation in progress notes, or if the patient was able to take oral medication (not NPO). Adverse events were defined as a blood pressure drop of more than 25%, a change in the heart rate by more than 20 beats per minute, or the need for IV fluids, based on previous studies.⁷ Although decreased blood pressure is not necessarily dangerous in and of itself, adverse events arising from blood pressure decreasing too rapidly from IV antihypertensives are well documented.^9,11 The presence of alternate etiologies of high blood pressure that were documented in progress notes, including pain, anxiety, agitation, and holding of home blood pressure medications, were recorded. The numbers of inappropriate orders pre- and postintervention were compared. Confounding factors of patient age and length of stay (LOS) were compared pre- and postintervention in order to rule out other factors to which the intervention’s effect could be attributed. Additionally, as a balancing measure, a random sample of patients with elevated blood pressure were monitored on a biweekly basis for adverse events that occurred as a result of not receiving IV treatment, including stroke, myocardial infarction, and pulmonary edema.

 

For this study, orders were reported on the standardized form of orders per 1,000 patient days. This was calculated as the number of orders divided by the total number of patient days from the two medicine units. For the univariate analysis, pre- and postintervention orders were compared for the different order categories using a t-test. Results were considered statistically significant at P < .05. Data analysis was conducted using SAS v. 9.4 (SAS Institute, Cary, North Carolina).

Additionally, a cost analysis was performed to estimate the hospital-wide annual cost of inappropriate orders. The analysis used the cost per dose¹² and included nurse-time derived from the median salary of those on our units. The hospital-wide cost was extrapolated to estimate the potential annual savings for the institution.

RESULTS

A total of 260 one-time orders of IV antihypertensives were analyzed in this study, 127 in the seven-month preintervention period and 133 in the 12-month postintervention period. The majority, 67.3% (n = 175), were labetalol orders. Inappropriate orders (ie, neither NPO nor hypertensive emergency) decreased from 8.3 to 3.3 orders per 1,000 patient days (P = .0099; Figure 3).

In total, there were 86 adverse events (33.1%), the majority of which (94.2%, n = 81) were a >25% decrease in blood pressure (Table 1). The number of adverse events per 1,000 patient days decreased from 4.4 in the preintervention period to 1.9 postintervention, P = .0112. Of the inappropriate orders, adverse events decreased from 3.7 to 0.8 per 1,000 patient days, P = .0072. Overall, there were 76 orders (29.2%) with documented alternate etiologies. The number of orders per 1,000 patient days with an alternate etiology decreased from 4.7 in the preintervention period to 1.2 postintervention, P =.0044 (Table 2). Descriptive analysis of patient characteristics pre- and postintervention were not statistically significant; for age 68.4 vs 70.7, P = .0823 and for LOS 14.8 vs 15.4, P = .0769. As a balancing measure, 111 patients with elevated blood pressure were monitored for adverse events during the postintervention period. Among patients who did not receive IV medication based on our algorithm, there were no adverse events.

Cost analysis estimated a $17,890 annual hospital-wide cost for unnecessary IV antihypertensive medications before the intervention. The estimate was calculated using the number of orders on the two medical units observed during the seven-month preintervention period, extrapolated to a 12-month period and to the total number of 15 medical units in the hospital. The intervention on the two studied medical units themselves led to an estimated $1,421 cost reduction (59.6%). Had the intervention been implemented hospital-wide with similar results, the resulting cost reduction would have amounted to $10,662.

DISCUSSION

Our initiative successfully demonstrated a significant reduction of 60% in inappropriate one-time orders of IV antihypertensives per 1,000 patient days. Accordingly, the number of adverse events per 1,000 patient days decreased by 57%. There was also a decrease in the number and percentage of IV orders with documented alternate etiologies. We hypothesize that this was due to nurses and physicians assessing and treating these conditions prior to treating hypertension in the intervention period, consequently avoiding an IV order.

 

The goal of the intervention was to have nurses assess for end-organ damage and alternate etiologies and include this information on their assessment provided to the physician, which would result in appropriate treatment of elevated blood pressure. By performing an interdisciplinary intervention, we addressed the knowledge deficit of both nurses and physicians, improved the triage of elevated blood pressure, and likely decreased the number of pages to providers.

To our knowledge, this is the first intervention addressing the inpatient overuse of IV antihypertensive medications for the treatment of asymptomatic hypertension. Additionally, this study bolsters prior evidence that the use of IV antihypertensives in asymptomatic patients leads to a large number of adverse events.⁷ A third of patients in the preintervention period had documented alternate etiologies of their blood pressure elevation, highlighting the need to assess and potentially treat these causes prior to treating blood pressure itself.

Reducing unnecessary treatment of asymptomatic blood pressure elevation is challenging. Evidence shows that both clinicians and patients overestimate the benefits and underestimate the harms of medical interventions.^13,14 This unfortunately leads to unjustified enthusiasm for medical treatments, which can worsen outcomes.¹⁵ Additionally, there may be a lack of knowledge of the guidelines, as well as the amount of time required in the full assessment of hypertensive urgency, that creates a culture of “treating the number.”

Changing physician behavior is difficult.¹⁶ However, active forms of continuing education and multifaceted interventions, such as ours, are most effective.¹⁷ Our message focused on patient safety and harm reduction, addressed clinicians’ safety concerns, and included stories of real cases where this overuse led to adverse events—all of which are encouraged in order to facilitate clinician engagement.¹⁸

There were limitations to this study. Only blood pressure elevations associated with an IV antihypertensive order and not all blood pressure elevations meeting the criteria for hypertensive urgency in general were examined. Additionally, our documentation of symptoms of hypertensive emergency and alternate etiologies was based only on documentation in the medical record. Ideally, we would have liked to conduct an interrupted time series analysis to assess the effect of the intervention over time; however, there were not enough orders of IV antihypertensives to perform such an analysis.

CONCLUSION

Treatment of asymptomatic blood pressure with IV antihypertensive medications can lead to patient harm. To reduce inappropriate treatment, our Student High Value Care team set out to challenge this common practice. Our interdisciplinary intervention successfully reduced unnecessary IV antihypertensive treatment. This may serve as a model for other institutions.

Disclosures

There are no relevant conflicts of interest to disclose for any authors.