Original Research

ABSTRACT

Objective To reduce unnecessary orthopedic referrals by developing a protocol for managing physiologic bow legs in the primary care environment through the use of a noninvasive technique that simultaneously tracks normal varus progression and screens for potential pathologic bowing requiring an orthopedic referral.

Methods Retrospective study of 155 patients with physiologic genu varum and 10 with infantile Blount’s disease. We used fingerbreadth measurements to document progression or resolution of bow legs. Final diagnoses were made by one orthopedic surgeon using clinical and radiographic evidence. We divided genu varum patients into 3 groups: patients presenting with bow legs before 18 months of age (MOA), patients presenting between 18 and 23 MOA, and patients presenting at 24 MOA or older for analyses relevant to the development of the follow-up protocol.

Results Physiologic genu varum patients walked earlier than average infants (10 months vs 12-15 months; P<.001). Physiologic genu varum patients presenting before 18 MOA demonstrated initial signs of correction between 18 and 24 MOA and resolution by 30 MOA. Physiologic genu varum patients presenting between 18 and 23 MOA demonstrated initial signs of correction between 24 MOA and 30 MOA and resolution by 36 MOA.

Conclusion Primary care physicians can manage most children presenting with bow legs. Management focuses on following the progression or resolution of varus with regular follow-up. For patients presenting with bow legs, we recommend a follow-up protocol using mainly well-child checkups and a simple clinical assessment to monitor varus progression and screen for pathologic bowing.

Bow legs in young children can be a concern for parents.^1,2 By far, the most common reason for bow legs is physiologic genu varum,^3-5 a nonprogressive stage of normal development in young children that generally resolves spontaneously without treatment.^1,6-11 Normally developing children undergo a varus phase between birth and 18 to 24 months of age (MOA), at which time there is usually a transition in alignment from varus to straight to valgus (knock knees), which will correct to straight or mild valgus throughout adolescence.^{1,6,7,9,10,12-17}

The most common form of pathologic bow legs is Blount’s disease, also known as tibia vara, which must be differentiated from physiologic genu varum.^{8-10,15,18-24} The progressive varus deformity of Blount’s disease usually requires orthopedic intervention.^1,10,23-26 Early diagnosis may spare patients complex interventions, improve prognosis, and limit complications that include gait abnormalities,^4,8,10,27 knee joint instability,^4,24,27 osteoarthritis,^9,20,27 meniscal tears,²⁷ and degenerative joint disease.^19,20,27

Although variables such as walking age, race, weight, and gender have been suggested as risk factors for Blount’s disease, they have not been useful in differentiating between Blount’s pathology and physiologic genu varum.^{1,4,5,7,10,20,28} In the primary care setting, distinguishing physiologic from pathologic forms of bow legs is possible with a thorough history and physical exam and with radiographs, as warranted.^1,2,15 More than 40% of genu varum/genu valgum cases referred for orthopedic consultation turn out to be the physiologic form,² suggesting a need for guidelines in the primary care setting to help direct referral and follow-up. The purpose of this study was to provide recommendations to family physicians for evaluating and managing children with bow legs.

Materials and methods

This study, approved by the Internal Review Board of Akron Children’s Hospital, is a retrospective review of children seen by a single pediatric orthopedic surgeon (DSW) from 1970 to 2012. Four-hundred twenty-four children were received for evaluation of bow legs. Excluded from our final analysis were 220 subjects seen only once for this specific referral and 39 subjects diagnosed with a condition other than genu varum or Blount’s disease (ie, rickets, skeletal dysplasia, sequelae of trauma, or infection). Ten subjects with Blount’s disease and 155 subjects with physiologic genu varum were included in the final data analysis.

More than 40% of genu varum cases referred for orthopedic consultation turn out to be the physiologic form.In addition to noting the age at which a patient walked independently, at each visit we documented age and the fingerbreadth (varus) distance between the medial femoral condyles with the child’s ankles held together. Parents reported age of independent walking for just 3 children with Blount’s disease and for 134 children with physiologic genu varum. Study variables for the genu varum data analysis were age of walking, age at presentation, age at varus correction, age at varus resolution, time between presentation and varus correction, and time between presentation and varus resolution. Varus correction is defined as any decrease in varus angulation since presentation. Varus resolution is defined as varus correction to less than or equal to half of the varus angulation at presentation. For inclusion in the age-at-resolution analysis, a child must have been evaluated at regular follow-up visits (all rechecks within 8 months).

To measure varus distance, we used the fingerbreadth method described by Weiner in a study of 600 cases (FIGURE).⁶ This simple technique, which requires no special equipment, accurately detected differences in varus angulation and tracked the normal pattern of lower limb angular development. The patient should be supine on the examination table with legs extended. With one hand, the examiner holds the child’s ankles together, ensuring the medial malleoli are in contact. With the other hand, the examiner measures the fingerbreadth distance between the medial femoral condyles. Alternatively, a ruler may be used to measure the distance. This latter method may be especially useful in practices where the patient is likely to see more than one provider for well child care.

We divided the genu varum subject group into 3 subgroups by age at presentation: 103 subjects were younger than 18 months; 47 were 18 to 23 months; and 5 were 24 months or older. We used the data analysis toolkit in Microsoft Excel 2013 to perform a statistical analysis of study variables. We assumed the genu varum population is a normally distributed population. We used a 95% confidence level (α=0.05) for all calculations of confidence intervals (CIs), student t-tests, and tolerance intervals. Based on the data analysis results, we developed a series of follow-up and referral guidelines for practitioners.

Results

The mean walking age for those diagnosed with physiologic genu varum was 10 months (95% CI, 9.8-10.4), which is significantly younger than the 12 months of age (at the earliest) typical of toddlers in general (P<.001). There was no significant difference between the walking age of male and female children diagnosed with genu varum (P=.37).

Of the children presenting with the primary complaint of bow legs, 6% subsequently developed Blount’s disease. These patients presented at a mean age of 20.9 months and were diagnosed at a mean age of 23.9 months. Following the Blount’s disease diagnosis, we initiated therapy in all cases (3 surgical, 7 bracing).

Physiologic genu varum patients presented at a mean age of 16.4 months, with only 3.23% presenting at older than 23 months. On average, physiologic genu varum patients presenting before 24 months of age showed measurable varus correction 5 months after presentation and achieved varus resolution 7.3 months after presentation (TABLE 1). Assuming the patient population is normally distributed, we can be 95% confident that 95% of physiologic genu varum patients presenting before 18 months of age will show measurable varus correction by 24 months and will resolve without intervention by 30 months (TABLE 2). Patients presenting between 18 and 23 months of age should show measurable varus correction by 30 months and resolution by 36 months (TABLE 3).

Discussion

Primary care physicians have the ability to differentiate physiologic genu varum from pathologic forms of bow legs with a thorough history, physical exam, and radiographic examination, if necessary^1,2,13 (TABLE 4^{1,7,8,10,12,14,18-20,22,24,27}). Several approaches to differentiating Blount’s disease and physiologic genu varum have been described in the literature.^{1,4,7,8,10,14,22,23}

The average age at which children begin to walk independently is between 13 and 15 months.^5,18,29-31 Recently, it has been suggested that the range be expanded to include 12 months of age.³⁰ The association between early walking (at 10-11 months)^12,20,22 and Blount’s disease is generally accepted in the orthopedic literature.^{1,4,7,10,19-22} However, some authors have suggested early walking also contributes to genu varum.^{1,5,8,10,18,28} The mean age of independent walking for children with physiologic genu varum suggested in the literature (10 months) was confirmed in our study and found to be significantly younger than the average for toddlers generally.^1,22 Early walking is clearly associated with both physiologic genu varum and Blount’s disease, but no direct causation has been identified in either case. An alternative means of differentiating these entities is needed.

Primary care physicians can differentiate physiologic genu varum from pathologic forms of bow legs with a thorough history, physical exam, and radiographic examination (if needed).Radiographic examination of the knee is essential to the diagnosis of Blount’s disease as well as other, less common causes of pathologic bow legs (skeletal dysplasia, rickets, traumatic growth plate insults, infections, neoplasms).^1,8,14,19 The common radiologic classification of staging for Blount’s disease is the Langenskiöld staging system, which involves identification of characteristic radiographic changes at the tibial physis.^{5,8,14,15,18,22,24}

The fingerbreadth method, which requires no special equipment, accurately detected differences in varus angulation and tracked the normal pattern of lower limb angular development.Sequential measurement of genu varum is most useful in differentiating between physiologic and pathologic processes. Physiologic genu varum, an exaggeration of the normal developmental pattern, characteristically resolves and evolves into physiologic genu valgum by 3 years of age.^1,6-11 The pathophysiology of Blount’s disease is believed to be related to biomechanical overloading of the posteromedial proximal tibia during gait with the knee in a varus orientation. Excess loading on the proximal medial physis contributes to varus progression.^{4,10,14,20,25,27} Patients with Blount’s disease progress with varus and concomitant internal tibial torsion associated with growth plate irregularities and eventually exhibit premature closure.^{1,10,14,18,20,23,24,26} In the months prior to Blount’s disease diagnosis, increasing varus has been reported.^4,7,10,19 Varus progression that differs from the expected pattern indicates possible pathologic bow legs and should prompt radiologic evaluation and, often, an orthopedic referral.^{3,4,7-9,12,13,21}

In our study, only 3% of children with physiologic genu varum presented at 24 months of age or older, compared with 20% of Blount’s disease patients. We recommend considering orthopedic referral for any patient presenting with bow legs at 24 months of age or older. Additionally, consider orthopedic referral for any patient whose varus has not begun to correct within 8 months or has not resolved within 14 months of presentation, as more than 95% of patients with physiologic genu varum are expected to meet these milestones (TABLE 1). And do not hesitate to refer patients at any stage of follow-up if you suspect pathology or if parents are anxious.

If no sign of pathology is immediately identified, we recommend the following course of action:

• Record a reference fingerbreadth or ruler measurement at the initial presentation.
• Re-examine the knee varus at the next regular well-child visit (TABLE 5).

In our study, only 3% of children with physiologic genu varum presented at 24 months of age or older, compared with 20% of Blount's disease patients.Re-examining the patient prior to the next well-child visit is unnecessary, as some degree of bowing is typical until age 18 to 24 months.^{1,6,7,9,12,13,17} Recommend orthopedic referral for any patient with varus that has progressed since initial presentation. Without signs of pathology, repeat varus assessment at the next well-child visit. This schedule minimizes the need for additional physician appointments by integrating follow-up into the typical well-child visits at 18, 24, 30, and 36 months of age.³² The 6-month follow-up interval was a feature of our study and is recommended in the related literature.¹²

• Consider orthopedic referral for patients whose varus has not corrected by the second follow-up appointment, as more than 95% of patients should have measurable varus correction at this visit. Most patients will have exhibited varus resolution by this time and will not require additional follow-up. For patients with observable correction who do not yet meet the criteria for resolution, we recommend a third, final follow-up appointment in another 6 months.
• Refer any patient whose varus has not resolved by the third follow-up appointment, as more than 95% of genu varum cases should have resolved by this time. This finding is echoed in the literature; any varus beyond 36 months of age is considered abnormal and suggestive of pathology.^5,7,8,13,14 If evidence of Blount’s or skeletal dysplasia is identified, orthopedic management will likely consist of bracing (orthotics) or surgical management.

CORRESPONDENCE
Dennis S. Weiner, MD, Department of Orthopedic Surgery, Akron Children’s Hospital, 300 Locust Street, Suite 250, Akron, OH, 44302; mdicintio@chmca.org.

ACKNOWLEDGEMENTS
The authors thank Meadow Newton, BS, assistant research coordinator, Akron Children’s Hospital, for her editing and technical assistance and Richard Steiner, PhD, The University of Akron, for his statistical review.

ABSTRACT

Objective To reduce unnecessary orthopedic referrals by developing a protocol for managing physiologic bow legs in the primary care environment through the use of a noninvasive technique that simultaneously tracks normal varus progression and screens for potential pathologic bowing requiring an orthopedic referral.

Methods Retrospective study of 155 patients with physiologic genu varum and 10 with infantile Blount’s disease. We used fingerbreadth measurements to document progression or resolution of bow legs. Final diagnoses were made by one orthopedic surgeon using clinical and radiographic evidence. We divided genu varum patients into 3 groups: patients presenting with bow legs before 18 months of age (MOA), patients presenting between 18 and 23 MOA, and patients presenting at 24 MOA or older for analyses relevant to the development of the follow-up protocol.

Results Physiologic genu varum patients walked earlier than average infants (10 months vs 12-15 months; P<.001). Physiologic genu varum patients presenting before 18 MOA demonstrated initial signs of correction between 18 and 24 MOA and resolution by 30 MOA. Physiologic genu varum patients presenting between 18 and 23 MOA demonstrated initial signs of correction between 24 MOA and 30 MOA and resolution by 36 MOA.

Conclusion Primary care physicians can manage most children presenting with bow legs. Management focuses on following the progression or resolution of varus with regular follow-up. For patients presenting with bow legs, we recommend a follow-up protocol using mainly well-child checkups and a simple clinical assessment to monitor varus progression and screen for pathologic bowing.

Bow legs in young children can be a concern for parents.^1,2 By far, the most common reason for bow legs is physiologic genu varum,^3-5 a nonprogressive stage of normal development in young children that generally resolves spontaneously without treatment.^1,6-11 Normally developing children undergo a varus phase between birth and 18 to 24 months of age (MOA), at which time there is usually a transition in alignment from varus to straight to valgus (knock knees), which will correct to straight or mild valgus throughout adolescence.^{1,6,7,9,10,12-17}

The most common form of pathologic bow legs is Blount’s disease, also known as tibia vara, which must be differentiated from physiologic genu varum.^{8-10,15,18-24} The progressive varus deformity of Blount’s disease usually requires orthopedic intervention.^1,10,23-26 Early diagnosis may spare patients complex interventions, improve prognosis, and limit complications that include gait abnormalities,^4,8,10,27 knee joint instability,^4,24,27 osteoarthritis,^9,20,27 meniscal tears,²⁷ and degenerative joint disease.^19,20,27

Although variables such as walking age, race, weight, and gender have been suggested as risk factors for Blount’s disease, they have not been useful in differentiating between Blount’s pathology and physiologic genu varum.^{1,4,5,7,10,20,28} In the primary care setting, distinguishing physiologic from pathologic forms of bow legs is possible with a thorough history and physical exam and with radiographs, as warranted.^1,2,15 More than 40% of genu varum/genu valgum cases referred for orthopedic consultation turn out to be the physiologic form,² suggesting a need for guidelines in the primary care setting to help direct referral and follow-up. The purpose of this study was to provide recommendations to family physicians for evaluating and managing children with bow legs.

Materials and methods

This study, approved by the Internal Review Board of Akron Children’s Hospital, is a retrospective review of children seen by a single pediatric orthopedic surgeon (DSW) from 1970 to 2012. Four-hundred twenty-four children were received for evaluation of bow legs. Excluded from our final analysis were 220 subjects seen only once for this specific referral and 39 subjects diagnosed with a condition other than genu varum or Blount’s disease (ie, rickets, skeletal dysplasia, sequelae of trauma, or infection). Ten subjects with Blount’s disease and 155 subjects with physiologic genu varum were included in the final data analysis.

More than 40% of genu varum cases referred for orthopedic consultation turn out to be the physiologic form.In addition to noting the age at which a patient walked independently, at each visit we documented age and the fingerbreadth (varus) distance between the medial femoral condyles with the child’s ankles held together. Parents reported age of independent walking for just 3 children with Blount’s disease and for 134 children with physiologic genu varum. Study variables for the genu varum data analysis were age of walking, age at presentation, age at varus correction, age at varus resolution, time between presentation and varus correction, and time between presentation and varus resolution. Varus correction is defined as any decrease in varus angulation since presentation. Varus resolution is defined as varus correction to less than or equal to half of the varus angulation at presentation. For inclusion in the age-at-resolution analysis, a child must have been evaluated at regular follow-up visits (all rechecks within 8 months).

To measure varus distance, we used the fingerbreadth method described by Weiner in a study of 600 cases (FIGURE).⁶ This simple technique, which requires no special equipment, accurately detected differences in varus angulation and tracked the normal pattern of lower limb angular development. The patient should be supine on the examination table with legs extended. With one hand, the examiner holds the child’s ankles together, ensuring the medial malleoli are in contact. With the other hand, the examiner measures the fingerbreadth distance between the medial femoral condyles. Alternatively, a ruler may be used to measure the distance. This latter method may be especially useful in practices where the patient is likely to see more than one provider for well child care.

We divided the genu varum subject group into 3 subgroups by age at presentation: 103 subjects were younger than 18 months; 47 were 18 to 23 months; and 5 were 24 months or older. We used the data analysis toolkit in Microsoft Excel 2013 to perform a statistical analysis of study variables. We assumed the genu varum population is a normally distributed population. We used a 95% confidence level (α=0.05) for all calculations of confidence intervals (CIs), student t-tests, and tolerance intervals. Based on the data analysis results, we developed a series of follow-up and referral guidelines for practitioners.

Results

The mean walking age for those diagnosed with physiologic genu varum was 10 months (95% CI, 9.8-10.4), which is significantly younger than the 12 months of age (at the earliest) typical of toddlers in general (P<.001). There was no significant difference between the walking age of male and female children diagnosed with genu varum (P=.37).

Of the children presenting with the primary complaint of bow legs, 6% subsequently developed Blount’s disease. These patients presented at a mean age of 20.9 months and were diagnosed at a mean age of 23.9 months. Following the Blount’s disease diagnosis, we initiated therapy in all cases (3 surgical, 7 bracing).

Physiologic genu varum patients presented at a mean age of 16.4 months, with only 3.23% presenting at older than 23 months. On average, physiologic genu varum patients presenting before 24 months of age showed measurable varus correction 5 months after presentation and achieved varus resolution 7.3 months after presentation (TABLE 1). Assuming the patient population is normally distributed, we can be 95% confident that 95% of physiologic genu varum patients presenting before 18 months of age will show measurable varus correction by 24 months and will resolve without intervention by 30 months (TABLE 2). Patients presenting between 18 and 23 months of age should show measurable varus correction by 30 months and resolution by 36 months (TABLE 3).

Discussion

Primary care physicians have the ability to differentiate physiologic genu varum from pathologic forms of bow legs with a thorough history, physical exam, and radiographic examination, if necessary^1,2,13 (TABLE 4^{1,7,8,10,12,14,18-20,22,24,27}). Several approaches to differentiating Blount’s disease and physiologic genu varum have been described in the literature.^{1,4,7,8,10,14,22,23}

The average age at which children begin to walk independently is between 13 and 15 months.^5,18,29-31 Recently, it has been suggested that the range be expanded to include 12 months of age.³⁰ The association between early walking (at 10-11 months)^12,20,22 and Blount’s disease is generally accepted in the orthopedic literature.^{1,4,7,10,19-22} However, some authors have suggested early walking also contributes to genu varum.^{1,5,8,10,18,28} The mean age of independent walking for children with physiologic genu varum suggested in the literature (10 months) was confirmed in our study and found to be significantly younger than the average for toddlers generally.^1,22 Early walking is clearly associated with both physiologic genu varum and Blount’s disease, but no direct causation has been identified in either case. An alternative means of differentiating these entities is needed.

Primary care physicians can differentiate physiologic genu varum from pathologic forms of bow legs with a thorough history, physical exam, and radiographic examination (if needed).Radiographic examination of the knee is essential to the diagnosis of Blount’s disease as well as other, less common causes of pathologic bow legs (skeletal dysplasia, rickets, traumatic growth plate insults, infections, neoplasms).^1,8,14,19 The common radiologic classification of staging for Blount’s disease is the Langenskiöld staging system, which involves identification of characteristic radiographic changes at the tibial physis.^{5,8,14,15,18,22,24}

The fingerbreadth method, which requires no special equipment, accurately detected differences in varus angulation and tracked the normal pattern of lower limb angular development.Sequential measurement of genu varum is most useful in differentiating between physiologic and pathologic processes. Physiologic genu varum, an exaggeration of the normal developmental pattern, characteristically resolves and evolves into physiologic genu valgum by 3 years of age.^1,6-11 The pathophysiology of Blount’s disease is believed to be related to biomechanical overloading of the posteromedial proximal tibia during gait with the knee in a varus orientation. Excess loading on the proximal medial physis contributes to varus progression.^{4,10,14,20,25,27} Patients with Blount’s disease progress with varus and concomitant internal tibial torsion associated with growth plate irregularities and eventually exhibit premature closure.^{1,10,14,18,20,23,24,26} In the months prior to Blount’s disease diagnosis, increasing varus has been reported.^4,7,10,19 Varus progression that differs from the expected pattern indicates possible pathologic bow legs and should prompt radiologic evaluation and, often, an orthopedic referral.^{3,4,7-9,12,13,21}

In our study, only 3% of children with physiologic genu varum presented at 24 months of age or older, compared with 20% of Blount’s disease patients. We recommend considering orthopedic referral for any patient presenting with bow legs at 24 months of age or older. Additionally, consider orthopedic referral for any patient whose varus has not begun to correct within 8 months or has not resolved within 14 months of presentation, as more than 95% of patients with physiologic genu varum are expected to meet these milestones (TABLE 1). And do not hesitate to refer patients at any stage of follow-up if you suspect pathology or if parents are anxious.

If no sign of pathology is immediately identified, we recommend the following course of action:

• Record a reference fingerbreadth or ruler measurement at the initial presentation.
• Re-examine the knee varus at the next regular well-child visit (TABLE 5).

In our study, only 3% of children with physiologic genu varum presented at 24 months of age or older, compared with 20% of Blount's disease patients.Re-examining the patient prior to the next well-child visit is unnecessary, as some degree of bowing is typical until age 18 to 24 months.^{1,6,7,9,12,13,17} Recommend orthopedic referral for any patient with varus that has progressed since initial presentation. Without signs of pathology, repeat varus assessment at the next well-child visit. This schedule minimizes the need for additional physician appointments by integrating follow-up into the typical well-child visits at 18, 24, 30, and 36 months of age.³² The 6-month follow-up interval was a feature of our study and is recommended in the related literature.¹²

• Consider orthopedic referral for patients whose varus has not corrected by the second follow-up appointment, as more than 95% of patients should have measurable varus correction at this visit. Most patients will have exhibited varus resolution by this time and will not require additional follow-up. For patients with observable correction who do not yet meet the criteria for resolution, we recommend a third, final follow-up appointment in another 6 months.
• Refer any patient whose varus has not resolved by the third follow-up appointment, as more than 95% of genu varum cases should have resolved by this time. This finding is echoed in the literature; any varus beyond 36 months of age is considered abnormal and suggestive of pathology.^5,7,8,13,14 If evidence of Blount’s or skeletal dysplasia is identified, orthopedic management will likely consist of bracing (orthotics) or surgical management.

CORRESPONDENCE
Dennis S. Weiner, MD, Department of Orthopedic Surgery, Akron Children’s Hospital, 300 Locust Street, Suite 250, Akron, OH, 44302; mdicintio@chmca.org.

ACKNOWLEDGEMENTS
The authors thank Meadow Newton, BS, assistant research coordinator, Akron Children’s Hospital, for her editing and technical assistance and Richard Steiner, PhD, The University of Akron, for his statistical review.

ABSTRACT

Objective To reduce unnecessary orthopedic referrals by developing a protocol for managing physiologic bow legs in the primary care environment through the use of a noninvasive technique that simultaneously tracks normal varus progression and screens for potential pathologic bowing requiring an orthopedic referral.

Methods Retrospective study of 155 patients with physiologic genu varum and 10 with infantile Blount’s disease. We used fingerbreadth measurements to document progression or resolution of bow legs. Final diagnoses were made by one orthopedic surgeon using clinical and radiographic evidence. We divided genu varum patients into 3 groups: patients presenting with bow legs before 18 months of age (MOA), patients presenting between 18 and 23 MOA, and patients presenting at 24 MOA or older for analyses relevant to the development of the follow-up protocol.

Results Physiologic genu varum patients walked earlier than average infants (10 months vs 12-15 months; P<.001). Physiologic genu varum patients presenting before 18 MOA demonstrated initial signs of correction between 18 and 24 MOA and resolution by 30 MOA. Physiologic genu varum patients presenting between 18 and 23 MOA demonstrated initial signs of correction between 24 MOA and 30 MOA and resolution by 36 MOA.

Conclusion Primary care physicians can manage most children presenting with bow legs. Management focuses on following the progression or resolution of varus with regular follow-up. For patients presenting with bow legs, we recommend a follow-up protocol using mainly well-child checkups and a simple clinical assessment to monitor varus progression and screen for pathologic bowing.

Bow legs in young children can be a concern for parents.^1,2 By far, the most common reason for bow legs is physiologic genu varum,^3-5 a nonprogressive stage of normal development in young children that generally resolves spontaneously without treatment.^1,6-11 Normally developing children undergo a varus phase between birth and 18 to 24 months of age (MOA), at which time there is usually a transition in alignment from varus to straight to valgus (knock knees), which will correct to straight or mild valgus throughout adolescence.^{1,6,7,9,10,12-17}

The most common form of pathologic bow legs is Blount’s disease, also known as tibia vara, which must be differentiated from physiologic genu varum.^{8-10,15,18-24} The progressive varus deformity of Blount’s disease usually requires orthopedic intervention.^1,10,23-26 Early diagnosis may spare patients complex interventions, improve prognosis, and limit complications that include gait abnormalities,^4,8,10,27 knee joint instability,^4,24,27 osteoarthritis,^9,20,27 meniscal tears,²⁷ and degenerative joint disease.^19,20,27

Although variables such as walking age, race, weight, and gender have been suggested as risk factors for Blount’s disease, they have not been useful in differentiating between Blount’s pathology and physiologic genu varum.^{1,4,5,7,10,20,28} In the primary care setting, distinguishing physiologic from pathologic forms of bow legs is possible with a thorough history and physical exam and with radiographs, as warranted.^1,2,15 More than 40% of genu varum/genu valgum cases referred for orthopedic consultation turn out to be the physiologic form,² suggesting a need for guidelines in the primary care setting to help direct referral and follow-up. The purpose of this study was to provide recommendations to family physicians for evaluating and managing children with bow legs.

Materials and methods

This study, approved by the Internal Review Board of Akron Children’s Hospital, is a retrospective review of children seen by a single pediatric orthopedic surgeon (DSW) from 1970 to 2012. Four-hundred twenty-four children were received for evaluation of bow legs. Excluded from our final analysis were 220 subjects seen only once for this specific referral and 39 subjects diagnosed with a condition other than genu varum or Blount’s disease (ie, rickets, skeletal dysplasia, sequelae of trauma, or infection). Ten subjects with Blount’s disease and 155 subjects with physiologic genu varum were included in the final data analysis.

More than 40% of genu varum cases referred for orthopedic consultation turn out to be the physiologic form.In addition to noting the age at which a patient walked independently, at each visit we documented age and the fingerbreadth (varus) distance between the medial femoral condyles with the child’s ankles held together. Parents reported age of independent walking for just 3 children with Blount’s disease and for 134 children with physiologic genu varum. Study variables for the genu varum data analysis were age of walking, age at presentation, age at varus correction, age at varus resolution, time between presentation and varus correction, and time between presentation and varus resolution. Varus correction is defined as any decrease in varus angulation since presentation. Varus resolution is defined as varus correction to less than or equal to half of the varus angulation at presentation. For inclusion in the age-at-resolution analysis, a child must have been evaluated at regular follow-up visits (all rechecks within 8 months).

To measure varus distance, we used the fingerbreadth method described by Weiner in a study of 600 cases (FIGURE).⁶ This simple technique, which requires no special equipment, accurately detected differences in varus angulation and tracked the normal pattern of lower limb angular development. The patient should be supine on the examination table with legs extended. With one hand, the examiner holds the child’s ankles together, ensuring the medial malleoli are in contact. With the other hand, the examiner measures the fingerbreadth distance between the medial femoral condyles. Alternatively, a ruler may be used to measure the distance. This latter method may be especially useful in practices where the patient is likely to see more than one provider for well child care.

We divided the genu varum subject group into 3 subgroups by age at presentation: 103 subjects were younger than 18 months; 47 were 18 to 23 months; and 5 were 24 months or older. We used the data analysis toolkit in Microsoft Excel 2013 to perform a statistical analysis of study variables. We assumed the genu varum population is a normally distributed population. We used a 95% confidence level (α=0.05) for all calculations of confidence intervals (CIs), student t-tests, and tolerance intervals. Based on the data analysis results, we developed a series of follow-up and referral guidelines for practitioners.

Results

The mean walking age for those diagnosed with physiologic genu varum was 10 months (95% CI, 9.8-10.4), which is significantly younger than the 12 months of age (at the earliest) typical of toddlers in general (P<.001). There was no significant difference between the walking age of male and female children diagnosed with genu varum (P=.37).

Of the children presenting with the primary complaint of bow legs, 6% subsequently developed Blount’s disease. These patients presented at a mean age of 20.9 months and were diagnosed at a mean age of 23.9 months. Following the Blount’s disease diagnosis, we initiated therapy in all cases (3 surgical, 7 bracing).

Physiologic genu varum patients presented at a mean age of 16.4 months, with only 3.23% presenting at older than 23 months. On average, physiologic genu varum patients presenting before 24 months of age showed measurable varus correction 5 months after presentation and achieved varus resolution 7.3 months after presentation (TABLE 1). Assuming the patient population is normally distributed, we can be 95% confident that 95% of physiologic genu varum patients presenting before 18 months of age will show measurable varus correction by 24 months and will resolve without intervention by 30 months (TABLE 2). Patients presenting between 18 and 23 months of age should show measurable varus correction by 30 months and resolution by 36 months (TABLE 3).

Discussion

Primary care physicians have the ability to differentiate physiologic genu varum from pathologic forms of bow legs with a thorough history, physical exam, and radiographic examination, if necessary^1,2,13 (TABLE 4^{1,7,8,10,12,14,18-20,22,24,27}). Several approaches to differentiating Blount’s disease and physiologic genu varum have been described in the literature.^{1,4,7,8,10,14,22,23}

The average age at which children begin to walk independently is between 13 and 15 months.^5,18,29-31 Recently, it has been suggested that the range be expanded to include 12 months of age.³⁰ The association between early walking (at 10-11 months)^12,20,22 and Blount’s disease is generally accepted in the orthopedic literature.^{1,4,7,10,19-22} However, some authors have suggested early walking also contributes to genu varum.^{1,5,8,10,18,28} The mean age of independent walking for children with physiologic genu varum suggested in the literature (10 months) was confirmed in our study and found to be significantly younger than the average for toddlers generally.^1,22 Early walking is clearly associated with both physiologic genu varum and Blount’s disease, but no direct causation has been identified in either case. An alternative means of differentiating these entities is needed.

Primary care physicians can differentiate physiologic genu varum from pathologic forms of bow legs with a thorough history, physical exam, and radiographic examination (if needed).Radiographic examination of the knee is essential to the diagnosis of Blount’s disease as well as other, less common causes of pathologic bow legs (skeletal dysplasia, rickets, traumatic growth plate insults, infections, neoplasms).^1,8,14,19 The common radiologic classification of staging for Blount’s disease is the Langenskiöld staging system, which involves identification of characteristic radiographic changes at the tibial physis.^{5,8,14,15,18,22,24}

The fingerbreadth method, which requires no special equipment, accurately detected differences in varus angulation and tracked the normal pattern of lower limb angular development.Sequential measurement of genu varum is most useful in differentiating between physiologic and pathologic processes. Physiologic genu varum, an exaggeration of the normal developmental pattern, characteristically resolves and evolves into physiologic genu valgum by 3 years of age.^1,6-11 The pathophysiology of Blount’s disease is believed to be related to biomechanical overloading of the posteromedial proximal tibia during gait with the knee in a varus orientation. Excess loading on the proximal medial physis contributes to varus progression.^{4,10,14,20,25,27} Patients with Blount’s disease progress with varus and concomitant internal tibial torsion associated with growth plate irregularities and eventually exhibit premature closure.^{1,10,14,18,20,23,24,26} In the months prior to Blount’s disease diagnosis, increasing varus has been reported.^4,7,10,19 Varus progression that differs from the expected pattern indicates possible pathologic bow legs and should prompt radiologic evaluation and, often, an orthopedic referral.^{3,4,7-9,12,13,21}

In our study, only 3% of children with physiologic genu varum presented at 24 months of age or older, compared with 20% of Blount’s disease patients. We recommend considering orthopedic referral for any patient presenting with bow legs at 24 months of age or older. Additionally, consider orthopedic referral for any patient whose varus has not begun to correct within 8 months or has not resolved within 14 months of presentation, as more than 95% of patients with physiologic genu varum are expected to meet these milestones (TABLE 1). And do not hesitate to refer patients at any stage of follow-up if you suspect pathology or if parents are anxious.

If no sign of pathology is immediately identified, we recommend the following course of action:

• Record a reference fingerbreadth or ruler measurement at the initial presentation.
• Re-examine the knee varus at the next regular well-child visit (TABLE 5).

In our study, only 3% of children with physiologic genu varum presented at 24 months of age or older, compared with 20% of Blount's disease patients.Re-examining the patient prior to the next well-child visit is unnecessary, as some degree of bowing is typical until age 18 to 24 months.^{1,6,7,9,12,13,17} Recommend orthopedic referral for any patient with varus that has progressed since initial presentation. Without signs of pathology, repeat varus assessment at the next well-child visit. This schedule minimizes the need for additional physician appointments by integrating follow-up into the typical well-child visits at 18, 24, 30, and 36 months of age.³² The 6-month follow-up interval was a feature of our study and is recommended in the related literature.¹²

• Consider orthopedic referral for patients whose varus has not corrected by the second follow-up appointment, as more than 95% of patients should have measurable varus correction at this visit. Most patients will have exhibited varus resolution by this time and will not require additional follow-up. For patients with observable correction who do not yet meet the criteria for resolution, we recommend a third, final follow-up appointment in another 6 months.
• Refer any patient whose varus has not resolved by the third follow-up appointment, as more than 95% of genu varum cases should have resolved by this time. This finding is echoed in the literature; any varus beyond 36 months of age is considered abnormal and suggestive of pathology.^5,7,8,13,14 If evidence of Blount’s or skeletal dysplasia is identified, orthopedic management will likely consist of bracing (orthotics) or surgical management.

CORRESPONDENCE
Dennis S. Weiner, MD, Department of Orthopedic Surgery, Akron Children’s Hospital, 300 Locust Street, Suite 250, Akron, OH, 44302; mdicintio@chmca.org.

ACKNOWLEDGEMENTS
The authors thank Meadow Newton, BS, assistant research coordinator, Akron Children’s Hospital, for her editing and technical assistance and Richard Steiner, PhD, The University of Akron, for his statistical review.

The deleterious impact of inappropriate and/or excessive antimicrobial usage is well recognized. In the United States, the Centers for Disease Control and Prevention (CDC) estimates that at least 2 million people become infected with antimicrobial-resistant bacteria with 23,000 subsequent deaths and at least $1 billion in excess medical costs per year.¹

In response, many healthcare organizations have developed antimicrobial stewardship programs (ASPs). Guidelines co-sponsored by the Infectious Diseases Society of America and the Society for Healthcare Epidemiology of America, as well as recent statements from the CDC and the Transatlantic Taskforce on Antimicrobial Resistance, all recommend core ASP elements.^2-5 The guidelines provide general recommendations on ASP structure, strategies, and activities. The recommended ASP structure is a team of physicians and pharmacists that collaborates with facility governing committees and other stakeholders to optimize antimicrobial use. While personnel with expertise in infectious diseases (ID) often lead ASPs, hospitalists are also recognized as key contributors, especially in quality improvement.^6,7 Recommended strategies include prospective audit of antimicrobial use with intervention and feedback and formulary restriction with preauthorization. Recommended activities include education, creation of guidelines, clinical pathways, and order forms, and programs to promote de-escalation and conversion from parenteral (IV) to oral (PO) antimicrobial therapy. However, limited evidence exists regarding the effectiveness of these ASP core elements.^8,9 While Cochrane reviews found clear evidence that particular stewardship strategies (eg, audit and feedback, formulary restriction, guidelines implemented with or without feedback, protocols, computerized decision support) can be effective in reducing antimicrobial usage and improving clinical outcomes over the long term, little evidence exists favoring 1 strategy over another.⁸ Furthermore, most individual studies of ASPs are single-center, making their conclusions less generalizable.

In 2012, the VA National Antimicrobial Stewardship Task Force (ASTF), in conjunction with the VA Healthcare Analysis and Information Group (HAIG) administered a survey on the characteristics of ASPs at all 130 acute care VA facilities (Appendix A). We used these survey results to build an implementation model and then assess associations between facility-level variables and 4 antimicrobial utilization measures.

Survey and Data

In 2011, the ASTF was chartered to develop, deploy, and monitor a strategic plan for optimizing antimicrobial therapy management. Monthly educational webinars and sample policies were offered to all facilities, including a sample business plan for stewardship and policies to encourage de-escalation from broad-spectrum antimicrobials, promote conversion from parenteral to oral antimicrobial therapy, avoid unnecessary double anaerobic coverage, and mitigate unnecessary antimicrobial usage in the context of Clostridium difficile infection.¹⁰

At the time that ASTF was chartered, the understanding of how ASP structures across VA facilities operated was limited. Hence, to capture baseline institutional characteristics and stewardship activities, ASTF and HAIG developed an inventory assessment of ASPs that was distributed online in November 2012. All 130 VA facilities providing inpatient acute care services responded.

We derived 57 facility characteristics relevant to antimicrobial utilization and conducted a series of factor analyses to simplify the complex dataset, and identify underlying latent constructs. We categorized resulting factors into domains of evidence, context, or facilitation as guided by the Promoting Action on Research Implementation in Health Services framework.¹¹ Briefly, the evidence domain describes how the facility uses codified and noncodified sources of knowledge (eg, research evidence, clinical experience). Organizational context comprises a facility’s characteristics that ensure a more conducive environment to put evidence into practice (eg, supportive leadership, organizational structure, evaluative systems). Facilitation emphasizes a facility personnel’s “state of preparedness” and receptivity to implementation.

Using factor analysis to identify facility factors as correlates of the outcomes, we first examined polychoric correlations among facility characteristics to assess multicollinearity. We performed independent component analysis to create latent constructs of variables that were defined by factor loadings (that indicated the proportion of variance accounted for by the construct) and uniqueness factors (that determined how well the variables were interpreted by the construct). Factors retained included variables that had uniqueness values of less than 0.7 and factor loadings greater than 0.3. Those associated with uniqueness values greater than 0.7 were left as single items, as were characteristics deemed a priori to be particularly important to antimicrobial stewardship. Factor scales that had only 2 items were converted into indices, while factor scores were generated for those factors that contained 3 or more items.^12-15

Data for facility-level antimicrobial utilization measures were obtained from the VA Corporate Data Warehouse from calendar year 2012. The analysis was conducted within the VA Informatics and Computing Infrastructure. All study procedures were approved by the VA Central Institutional Review Board.

Measures

Four utilization measures were defined as dependent measures: overall antimicrobial use; antimicrobial use in patients with non-infectious discharge diagnoses; missed opportunities to convert from parenteral to oral antimicrobial therapy; and missed opportunities to avoid double anaerobic coverage with metronidazole.

Overall antimicrobial use was defined as total acute care (ie, medical/surgical/intensive care) antibacterial use for each facility aggregated as per CDC National Healthcare Safety Network Antimicrobial Use Option guidelines (antimicrobial days per 1000 patient days present). A subanalysis of overall antimicrobial use was restricted to antimicrobial use among patients without an infection-related discharge diagnosis, as we surmised that this measure may capture a greater proportion of potentially unnecessary antimicrobial use. International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM)¹⁶ codes for infectious processes were identified by a combination of those classified previously in the literature,¹⁷ and those identified by finding the descendants of all infections named in the Systematized Nomenclature of Medicine--Clinical Terms.¹⁸ Next, all remaining codes for principal discharge diagnoses for which antimicrobials were administered were reviewed for potential indications for systemic antibacterial use. Discharges were considered noninfectious if no codes were identified when systemic antimicrobials were or could be indicated. For this measure, antimicrobial days were not counted if administered on or 1 day after the calendar day of surgery warranting antimicrobial prophylaxis.

Missed opportunities for conversion from parenteral to oral (IV to PO) formulations of highly bioavailable oral antimicrobials (ciprofloxacin, levofloxacin, moxifloxacin, azithromycin, clindamycin, linezolid, metronidazole, and fluconazole) were defined as the percentage of days of unnecessary IV therapy that were given when PO therapy could have been used among patients who were not in intensive care units at the time of antimicrobial administration who were receiving other oral medications, using previously described methodology.¹⁹ Missed opportunities for avoiding redundant anaerobic coverage with metronidazole were defined as the percentage of days in which patients receiving metronidazole also received antibiotics with activity against anaerobic bacteria, specifically beta-lactam/beta-lactamase inhibitors, carbapenems, cefotetan/cefoxitin, clindamycin, moxifloxacin, or tigecycline), using previously described methodology.²⁰ Patients for whom C. difficile testing was either ordered or positive within the prior 28 days (indicating potential clinical concern for C. difficile infection) were excluded from this endpoint.

Analysis

The variables derived above were entered into a multivariable model for each of the 4 antimicrobial utilization measures. The least absolute shrinkage and selection operator (LASSO) regression was used to determine significant associations between variables and individual utilization measures.²¹ LASSO was chosen because it offers advantages over traditional subset selection approaches in large multivariable analyses by assessing covariates simultaneously rather than sequentially, supporting prediction rather than estimation of effect.²²P values were not reported as they are not useful in determining statistical significance in this methodology. A tuning parameter of 0.025 was determined for the model based on a cross-validation approach. Significant variables remaining in the model were reported with the percent change in each utilization measure per unit change in the variable of interest. For binary factors, percent change was reported according to whether the variable was present or not. For ordinal variables, percent change was reported according to incremental increase in ordinal score. For continuous variables or variables represented by factor or index scores, percent change was reported per each 25% increase in the range of the score.

Inpatient Facility Antimicrobial Stewardship Characteristics and Antimicrobial Utilization

Frequencies of key facility characteristics that contributed to variable development are included in Table 1. Full survey results across all facilities are included in Appendix B. Factor analysis reduced the total number of variables to 32; however, we also included hospital size and VA complexity score. Thus, 34 variables were evaluated for association with antimicrobial utilization measures: 4 in the evidence domain, 23 in the context domain, and 7 in the facilitation domain (Table 2).

Median facility antimicrobial use was 619 antimicrobial days per 1000 days present (interquartile range [IQR], 554-700; overall range, 346-974). Median facility noninfectious antimicrobial use was 236 per 1000 days present (IQR, 200-286). Missed opportunities for conversion from IV to PO antimicrobial therapy were common, with a median facility value of 40.4% (391/969) of potentially eligible days of therapy (IQR, 32.2-47.8%). Missed opportunities to avoid double anaerobic coverage were less common (median 15.3% (186/1214) of potentially eligible days of therapy (IQR, 11.8%-20.2%; Figure).

Overall Antimicrobial Use

Four variables were associated with decreased overall antimicrobial use, although with small magnitude of change: presence of postgraduate physician/pharmacy training programs (0.03% decrease per quarter increase in factor score; on the order of 0.2 antimicrobial days per 1000 patient days present), presence of pharmacists and/or ID attendings on general medicine ward teams (0.02% decrease per quarter increase in index score), frequency of systematic de-escalation review (0.01% decrease per ordinal increase in score), and degree of involvement of ID physicians and/or fellows in antimicrobial approvals (0.007% decrease per quarter increase in index score). No variables were associated with increased overall antimicrobial use.

Antimicrobial Use among Discharges without Infectious Diagnoses

Six variables were associated with decreased antimicrobial use in patients without infectious discharge diagnoses, while 4 variables were associated with increased use. Variables associated with the greatest magnitude of decreased use included facility educational programs for prudent antimicrobial use (1.8% on the order of 4 antimicrobial days per 1000 patient days present), frequency of systematic de-escalation review (1.5% per incremental increase in score), and whether a facility’s lead antimicrobial stewardship pharmacist had ID training (1.3%). Also significantly associated with decreased use was a factor summarizing the presence of 4 condition-specific stewardship processes (de-escalation policies, policies for addressing antimicrobial use in the context of C. difficile infection, blood culture review, and automatic ID consults for certain conditions) (0.6% per quarter increase in factor score range), the extent to which postgraduate physician/pharmacy training programs were present (0.6% per quarter increase in factor score range), and the number of electronic antimicrobial-specific order sets present (0.4% per order set). The variables associated with increased use of antimicrobials included the presence of antimicrobial stop orders (4.6%), the degree to which non-ID physicians were involved in antimicrobial approvals (0.7% per increase in ordinal score), the level engagement with ASTF online resources (0.6% per quarter increase in factor score range), and hospital size (0.6% per 50-bed increase).

Missed Opportunities for Parenteral to Oral Antimicrobial Conversion

Missed opportunities for IV to PO antimicrobial conversion had the largest number of significant associations with organizational variables: 14 variables were associated with fewer missed opportunities, while 5 were associated with greater missed opportunities. Variables associated with the largest reductions in missed opportunities for IV to PO conversion included having guidelines for antimicrobial duration (12.8%), participating in regional stewardship collaboratives (8.1%), number of antimicrobial-specific order sets (6.0% per order set), ID training of the ASP pharmacist (4.9%), and VA facility complexity designation (4.2% per quarter increase in score indicating greater complexity).²³ Variables associated with more missed opportunities included stop orders (11.7%), overall perceived receptiveness to antimicrobial stewardship among clinical services (9.4%), the degree of engagement with ASTF online resources (6.9% per quarter increase in factor score range), educational programs for prudent antimicrobial use (4.1%), and hospital size (1.0% per 50-bed increase).

Missed Opportunities for Avoidance of Double Anaerobic Coverage

Four variables were associated with more avoidance of double anaerobic coverage: ID training of the lead ASP pharmacist (8.8%), presence of pharmacists and/or ID attendings on acute care ward teams (6.2% per quarter increase in index score), degree of ID pharmacist involvement in antimicrobial approvals, ranging from not at all (score=0) to both weekdays and nights/weekends (score=2; 4.3% per ordinal increase), and the number of antimicrobial-specific order sets (1.5% per order set). No variables were associated with less avoidance of double anaerobic coverage.

Variables Associated with Multiple Favorable or Unfavorable Antimicrobial Utilization Measures

To better assess the consistency of the relationship between organizational variables and measures of antimicrobial use, we tabulated variables that were associated with at least 3 potentially favorable (ie, reduced overall or noninfectious antimicrobial use or fewer missed opportunities) measures. Altogether, 5 variables satisfied this criterion: the presence of postgraduate physician/pharmacy training programs, the number of antimicrobial-specific order sets, frequency of systematic de-escalation review, the presence of pharmacists and/or ID attendings on acute care ward teams, and formal ID training of the lead ASP pharmacist (Table 3). Three other variables were associated with at least 2 unfavorable measures: hospital size, the degree to which the facility engaged with ASTF online resources, and presence of antimicrobial stop orders.

Variability in ASP implementation across VA allowed us to assess the relationship between ASP and facility elements and baseline patterns of antimicrobial utilization. Hospitalists and hospital policy-makers are becoming more and more engaged in inpatient antimicrobial stewardship. While our results suggest that having pharmacists and/or physicians with formal ID training participate in everyday inpatient activities can favorably improve antimicrobial utilization, considerable input into stewardship can be made by hospitalists and policy makers. In particular, based on this work, the highest yield from an organizational standpoint may be in working to develop order sets within the electronic medical record and systematic efforts to promote de-escalation of broad-spectrum therapy, as well as encouraging hospital administration to devote specific physician and pharmacy salary support to stewardship efforts.

While we noted that finding the ASTF online resources helpful was associated with potentially unfavorable antimicrobial utilization, we speculate that this may represent reverse causality due to facilities recognizing that their antimicrobial usage is suboptimal and thus seeking out sample ASTF policies to implement. The association between the presence of automatic stop orders and potentially unfavorable antimicrobial utilization is less clear since the timeframe was not specified in the survey; it may be that setting stop orders too far in advance may promote an environment in which critical thinking about antimicrobial de-escalation is not encouraged or timely. The larger magnitude of association between ASP characteristics and antimicrobial usage among patients without infectious discharge diagnoses versus overall antimicrobial usage also suggests that clinical situations where infection was of low enough suspicion to not even have the providers eventually list an infectious diagnosis on their discharge summaries may be particularly malleable to ASP interventions, though further exploration is needed in determining how useful this utilization measure may be as a marker for inappropriate antimicrobial use.

Our results complement those of Pakyz et al.²⁴ who surveyed 44 academic medical facilities in March 2013 to develop an ASP intensity score and correlate this score and its specific components to overall and targeted antimicrobial use. This study found that the overall ASP intensity score was not significantly associated with total or targeted antimicrobial use. However, ASP strategies were more associated with decreased total and targeted antimicrobial use than were specific ASP resources. In particular, the presence of a preauthorization strategy was associated with decreased targeted antimicrobial use. Our particular findings that order set establishment and de-escalation efforts are associated with multiple antibiotic outcomes also line up with the findings of Schuts et al, who performed a meta-analysis of the effects of meeting antimicrobial stewardship objectives and found that achieving guideline concordance (such as through establishment of order sets) and successfully de-escalating antimicrobial therapy was associated with reduced mortality.^25,26 This meta-analysis, however, was limited by low rigor of its studies and potential for reverse causality. While our study has the advantages of capturing an entire national network of 130 acute care facilities with a 100% response rate, it, too, is limited by a number of issues, most notably by the fact that the survey was not specifically designed for the analysis of antimicrobial utilization measures, patient-level risk stratification was not available, the VA population does not reflect the U.S. population at-large, recall bias, and that antimicrobial prescribing and stewardship practices have evolved in VA since 2012. Furthermore, all of the antimicrobial utilization measures studied are imperfect at capturing inappropriate antibiotic use; in particular, our reliance on principal ICD-9 codes for noninfectious outcomes requires prospective validation. Many survey questions were subjective and subject to misinterpretation; other unmeasured confounders may also be present. Causality cannot be inferred from association. Nevertheless, our findings support many core indicators for hospital ASP recommended by the CDC and the Transatlantic Taskforce on Antimicrobial Resistance,^3,4 most notably, having personnel with ID training involved in stewardship and establishing a formal procedure for ASP review for the appropriateness of an antimicrobial at or after 48 hours from the initial order.

In summary, the VA has made efforts to advance the practice of antimicrobial stewardship system-wide, including a 2014 directive that all VA facilities have an ASP,²⁷ since the 2012 HAIG assessment reported considerable variability in antimicrobial utilization and antimicrobial stewardship activities. Our study identifies areas of stewardship that may correlate with, positively or negatively, antimicrobial utilization measures that will require further investigation. A repeat and more detailed antimicrobial stewardship survey was recently completed and will help VA gauge ongoing effects of ASTF activities. We hope to re-evaluate our model with newer data when available.

Acknowledgments

The authors wish to thank Michael Fletcher, Jaime Lopez, and Catherine Loc-Carrillo for their administrative and organizational support of the project and Allison Kelly, MD, for her pivotal role in survey development and distribution. This work was supported by the VA Health Services Research and Development Service Collaborative Research to Enhance and Advance Transformation and Excellence (CREATE) Initiative; Cognitive Support Informatics for Antimicrobial Stewardship project (CRE 12-313).

Disclosure

 The authors report no financial conflicts of interest.

The deleterious impact of inappropriate and/or excessive antimicrobial usage is well recognized. In the United States, the Centers for Disease Control and Prevention (CDC) estimates that at least 2 million people become infected with antimicrobial-resistant bacteria with 23,000 subsequent deaths and at least $1 billion in excess medical costs per year.¹

In response, many healthcare organizations have developed antimicrobial stewardship programs (ASPs). Guidelines co-sponsored by the Infectious Diseases Society of America and the Society for Healthcare Epidemiology of America, as well as recent statements from the CDC and the Transatlantic Taskforce on Antimicrobial Resistance, all recommend core ASP elements.^2-5 The guidelines provide general recommendations on ASP structure, strategies, and activities. The recommended ASP structure is a team of physicians and pharmacists that collaborates with facility governing committees and other stakeholders to optimize antimicrobial use. While personnel with expertise in infectious diseases (ID) often lead ASPs, hospitalists are also recognized as key contributors, especially in quality improvement.^6,7 Recommended strategies include prospective audit of antimicrobial use with intervention and feedback and formulary restriction with preauthorization. Recommended activities include education, creation of guidelines, clinical pathways, and order forms, and programs to promote de-escalation and conversion from parenteral (IV) to oral (PO) antimicrobial therapy. However, limited evidence exists regarding the effectiveness of these ASP core elements.^8,9 While Cochrane reviews found clear evidence that particular stewardship strategies (eg, audit and feedback, formulary restriction, guidelines implemented with or without feedback, protocols, computerized decision support) can be effective in reducing antimicrobial usage and improving clinical outcomes over the long term, little evidence exists favoring 1 strategy over another.⁸ Furthermore, most individual studies of ASPs are single-center, making their conclusions less generalizable.

In 2012, the VA National Antimicrobial Stewardship Task Force (ASTF), in conjunction with the VA Healthcare Analysis and Information Group (HAIG) administered a survey on the characteristics of ASPs at all 130 acute care VA facilities (Appendix A). We used these survey results to build an implementation model and then assess associations between facility-level variables and 4 antimicrobial utilization measures.

Survey and Data

In 2011, the ASTF was chartered to develop, deploy, and monitor a strategic plan for optimizing antimicrobial therapy management. Monthly educational webinars and sample policies were offered to all facilities, including a sample business plan for stewardship and policies to encourage de-escalation from broad-spectrum antimicrobials, promote conversion from parenteral to oral antimicrobial therapy, avoid unnecessary double anaerobic coverage, and mitigate unnecessary antimicrobial usage in the context of Clostridium difficile infection.¹⁰

At the time that ASTF was chartered, the understanding of how ASP structures across VA facilities operated was limited. Hence, to capture baseline institutional characteristics and stewardship activities, ASTF and HAIG developed an inventory assessment of ASPs that was distributed online in November 2012. All 130 VA facilities providing inpatient acute care services responded.

We derived 57 facility characteristics relevant to antimicrobial utilization and conducted a series of factor analyses to simplify the complex dataset, and identify underlying latent constructs. We categorized resulting factors into domains of evidence, context, or facilitation as guided by the Promoting Action on Research Implementation in Health Services framework.¹¹ Briefly, the evidence domain describes how the facility uses codified and noncodified sources of knowledge (eg, research evidence, clinical experience). Organizational context comprises a facility’s characteristics that ensure a more conducive environment to put evidence into practice (eg, supportive leadership, organizational structure, evaluative systems). Facilitation emphasizes a facility personnel’s “state of preparedness” and receptivity to implementation.

Using factor analysis to identify facility factors as correlates of the outcomes, we first examined polychoric correlations among facility characteristics to assess multicollinearity. We performed independent component analysis to create latent constructs of variables that were defined by factor loadings (that indicated the proportion of variance accounted for by the construct) and uniqueness factors (that determined how well the variables were interpreted by the construct). Factors retained included variables that had uniqueness values of less than 0.7 and factor loadings greater than 0.3. Those associated with uniqueness values greater than 0.7 were left as single items, as were characteristics deemed a priori to be particularly important to antimicrobial stewardship. Factor scales that had only 2 items were converted into indices, while factor scores were generated for those factors that contained 3 or more items.^12-15

Data for facility-level antimicrobial utilization measures were obtained from the VA Corporate Data Warehouse from calendar year 2012. The analysis was conducted within the VA Informatics and Computing Infrastructure. All study procedures were approved by the VA Central Institutional Review Board.

Measures

Four utilization measures were defined as dependent measures: overall antimicrobial use; antimicrobial use in patients with non-infectious discharge diagnoses; missed opportunities to convert from parenteral to oral antimicrobial therapy; and missed opportunities to avoid double anaerobic coverage with metronidazole.

Overall antimicrobial use was defined as total acute care (ie, medical/surgical/intensive care) antibacterial use for each facility aggregated as per CDC National Healthcare Safety Network Antimicrobial Use Option guidelines (antimicrobial days per 1000 patient days present). A subanalysis of overall antimicrobial use was restricted to antimicrobial use among patients without an infection-related discharge diagnosis, as we surmised that this measure may capture a greater proportion of potentially unnecessary antimicrobial use. International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM)¹⁶ codes for infectious processes were identified by a combination of those classified previously in the literature,¹⁷ and those identified by finding the descendants of all infections named in the Systematized Nomenclature of Medicine--Clinical Terms.¹⁸ Next, all remaining codes for principal discharge diagnoses for which antimicrobials were administered were reviewed for potential indications for systemic antibacterial use. Discharges were considered noninfectious if no codes were identified when systemic antimicrobials were or could be indicated. For this measure, antimicrobial days were not counted if administered on or 1 day after the calendar day of surgery warranting antimicrobial prophylaxis.

Missed opportunities for conversion from parenteral to oral (IV to PO) formulations of highly bioavailable oral antimicrobials (ciprofloxacin, levofloxacin, moxifloxacin, azithromycin, clindamycin, linezolid, metronidazole, and fluconazole) were defined as the percentage of days of unnecessary IV therapy that were given when PO therapy could have been used among patients who were not in intensive care units at the time of antimicrobial administration who were receiving other oral medications, using previously described methodology.¹⁹ Missed opportunities for avoiding redundant anaerobic coverage with metronidazole were defined as the percentage of days in which patients receiving metronidazole also received antibiotics with activity against anaerobic bacteria, specifically beta-lactam/beta-lactamase inhibitors, carbapenems, cefotetan/cefoxitin, clindamycin, moxifloxacin, or tigecycline), using previously described methodology.²⁰ Patients for whom C. difficile testing was either ordered or positive within the prior 28 days (indicating potential clinical concern for C. difficile infection) were excluded from this endpoint.

Analysis

The variables derived above were entered into a multivariable model for each of the 4 antimicrobial utilization measures. The least absolute shrinkage and selection operator (LASSO) regression was used to determine significant associations between variables and individual utilization measures.²¹ LASSO was chosen because it offers advantages over traditional subset selection approaches in large multivariable analyses by assessing covariates simultaneously rather than sequentially, supporting prediction rather than estimation of effect.²²P values were not reported as they are not useful in determining statistical significance in this methodology. A tuning parameter of 0.025 was determined for the model based on a cross-validation approach. Significant variables remaining in the model were reported with the percent change in each utilization measure per unit change in the variable of interest. For binary factors, percent change was reported according to whether the variable was present or not. For ordinal variables, percent change was reported according to incremental increase in ordinal score. For continuous variables or variables represented by factor or index scores, percent change was reported per each 25% increase in the range of the score.

Inpatient Facility Antimicrobial Stewardship Characteristics and Antimicrobial Utilization

Frequencies of key facility characteristics that contributed to variable development are included in Table 1. Full survey results across all facilities are included in Appendix B. Factor analysis reduced the total number of variables to 32; however, we also included hospital size and VA complexity score. Thus, 34 variables were evaluated for association with antimicrobial utilization measures: 4 in the evidence domain, 23 in the context domain, and 7 in the facilitation domain (Table 2).

Median facility antimicrobial use was 619 antimicrobial days per 1000 days present (interquartile range [IQR], 554-700; overall range, 346-974). Median facility noninfectious antimicrobial use was 236 per 1000 days present (IQR, 200-286). Missed opportunities for conversion from IV to PO antimicrobial therapy were common, with a median facility value of 40.4% (391/969) of potentially eligible days of therapy (IQR, 32.2-47.8%). Missed opportunities to avoid double anaerobic coverage were less common (median 15.3% (186/1214) of potentially eligible days of therapy (IQR, 11.8%-20.2%; Figure).

Overall Antimicrobial Use

Four variables were associated with decreased overall antimicrobial use, although with small magnitude of change: presence of postgraduate physician/pharmacy training programs (0.03% decrease per quarter increase in factor score; on the order of 0.2 antimicrobial days per 1000 patient days present), presence of pharmacists and/or ID attendings on general medicine ward teams (0.02% decrease per quarter increase in index score), frequency of systematic de-escalation review (0.01% decrease per ordinal increase in score), and degree of involvement of ID physicians and/or fellows in antimicrobial approvals (0.007% decrease per quarter increase in index score). No variables were associated with increased overall antimicrobial use.

Antimicrobial Use among Discharges without Infectious Diagnoses

Six variables were associated with decreased antimicrobial use in patients without infectious discharge diagnoses, while 4 variables were associated with increased use. Variables associated with the greatest magnitude of decreased use included facility educational programs for prudent antimicrobial use (1.8% on the order of 4 antimicrobial days per 1000 patient days present), frequency of systematic de-escalation review (1.5% per incremental increase in score), and whether a facility’s lead antimicrobial stewardship pharmacist had ID training (1.3%). Also significantly associated with decreased use was a factor summarizing the presence of 4 condition-specific stewardship processes (de-escalation policies, policies for addressing antimicrobial use in the context of C. difficile infection, blood culture review, and automatic ID consults for certain conditions) (0.6% per quarter increase in factor score range), the extent to which postgraduate physician/pharmacy training programs were present (0.6% per quarter increase in factor score range), and the number of electronic antimicrobial-specific order sets present (0.4% per order set). The variables associated with increased use of antimicrobials included the presence of antimicrobial stop orders (4.6%), the degree to which non-ID physicians were involved in antimicrobial approvals (0.7% per increase in ordinal score), the level engagement with ASTF online resources (0.6% per quarter increase in factor score range), and hospital size (0.6% per 50-bed increase).

Missed Opportunities for Parenteral to Oral Antimicrobial Conversion

Missed opportunities for IV to PO antimicrobial conversion had the largest number of significant associations with organizational variables: 14 variables were associated with fewer missed opportunities, while 5 were associated with greater missed opportunities. Variables associated with the largest reductions in missed opportunities for IV to PO conversion included having guidelines for antimicrobial duration (12.8%), participating in regional stewardship collaboratives (8.1%), number of antimicrobial-specific order sets (6.0% per order set), ID training of the ASP pharmacist (4.9%), and VA facility complexity designation (4.2% per quarter increase in score indicating greater complexity).²³ Variables associated with more missed opportunities included stop orders (11.7%), overall perceived receptiveness to antimicrobial stewardship among clinical services (9.4%), the degree of engagement with ASTF online resources (6.9% per quarter increase in factor score range), educational programs for prudent antimicrobial use (4.1%), and hospital size (1.0% per 50-bed increase).

Missed Opportunities for Avoidance of Double Anaerobic Coverage

Four variables were associated with more avoidance of double anaerobic coverage: ID training of the lead ASP pharmacist (8.8%), presence of pharmacists and/or ID attendings on acute care ward teams (6.2% per quarter increase in index score), degree of ID pharmacist involvement in antimicrobial approvals, ranging from not at all (score=0) to both weekdays and nights/weekends (score=2; 4.3% per ordinal increase), and the number of antimicrobial-specific order sets (1.5% per order set). No variables were associated with less avoidance of double anaerobic coverage.

Variables Associated with Multiple Favorable or Unfavorable Antimicrobial Utilization Measures

To better assess the consistency of the relationship between organizational variables and measures of antimicrobial use, we tabulated variables that were associated with at least 3 potentially favorable (ie, reduced overall or noninfectious antimicrobial use or fewer missed opportunities) measures. Altogether, 5 variables satisfied this criterion: the presence of postgraduate physician/pharmacy training programs, the number of antimicrobial-specific order sets, frequency of systematic de-escalation review, the presence of pharmacists and/or ID attendings on acute care ward teams, and formal ID training of the lead ASP pharmacist (Table 3). Three other variables were associated with at least 2 unfavorable measures: hospital size, the degree to which the facility engaged with ASTF online resources, and presence of antimicrobial stop orders.

Variability in ASP implementation across VA allowed us to assess the relationship between ASP and facility elements and baseline patterns of antimicrobial utilization. Hospitalists and hospital policy-makers are becoming more and more engaged in inpatient antimicrobial stewardship. While our results suggest that having pharmacists and/or physicians with formal ID training participate in everyday inpatient activities can favorably improve antimicrobial utilization, considerable input into stewardship can be made by hospitalists and policy makers. In particular, based on this work, the highest yield from an organizational standpoint may be in working to develop order sets within the electronic medical record and systematic efforts to promote de-escalation of broad-spectrum therapy, as well as encouraging hospital administration to devote specific physician and pharmacy salary support to stewardship efforts.

While we noted that finding the ASTF online resources helpful was associated with potentially unfavorable antimicrobial utilization, we speculate that this may represent reverse causality due to facilities recognizing that their antimicrobial usage is suboptimal and thus seeking out sample ASTF policies to implement. The association between the presence of automatic stop orders and potentially unfavorable antimicrobial utilization is less clear since the timeframe was not specified in the survey; it may be that setting stop orders too far in advance may promote an environment in which critical thinking about antimicrobial de-escalation is not encouraged or timely. The larger magnitude of association between ASP characteristics and antimicrobial usage among patients without infectious discharge diagnoses versus overall antimicrobial usage also suggests that clinical situations where infection was of low enough suspicion to not even have the providers eventually list an infectious diagnosis on their discharge summaries may be particularly malleable to ASP interventions, though further exploration is needed in determining how useful this utilization measure may be as a marker for inappropriate antimicrobial use.

Our results complement those of Pakyz et al.²⁴ who surveyed 44 academic medical facilities in March 2013 to develop an ASP intensity score and correlate this score and its specific components to overall and targeted antimicrobial use. This study found that the overall ASP intensity score was not significantly associated with total or targeted antimicrobial use. However, ASP strategies were more associated with decreased total and targeted antimicrobial use than were specific ASP resources. In particular, the presence of a preauthorization strategy was associated with decreased targeted antimicrobial use. Our particular findings that order set establishment and de-escalation efforts are associated with multiple antibiotic outcomes also line up with the findings of Schuts et al, who performed a meta-analysis of the effects of meeting antimicrobial stewardship objectives and found that achieving guideline concordance (such as through establishment of order sets) and successfully de-escalating antimicrobial therapy was associated with reduced mortality.^25,26 This meta-analysis, however, was limited by low rigor of its studies and potential for reverse causality. While our study has the advantages of capturing an entire national network of 130 acute care facilities with a 100% response rate, it, too, is limited by a number of issues, most notably by the fact that the survey was not specifically designed for the analysis of antimicrobial utilization measures, patient-level risk stratification was not available, the VA population does not reflect the U.S. population at-large, recall bias, and that antimicrobial prescribing and stewardship practices have evolved in VA since 2012. Furthermore, all of the antimicrobial utilization measures studied are imperfect at capturing inappropriate antibiotic use; in particular, our reliance on principal ICD-9 codes for noninfectious outcomes requires prospective validation. Many survey questions were subjective and subject to misinterpretation; other unmeasured confounders may also be present. Causality cannot be inferred from association. Nevertheless, our findings support many core indicators for hospital ASP recommended by the CDC and the Transatlantic Taskforce on Antimicrobial Resistance,^3,4 most notably, having personnel with ID training involved in stewardship and establishing a formal procedure for ASP review for the appropriateness of an antimicrobial at or after 48 hours from the initial order.

In summary, the VA has made efforts to advance the practice of antimicrobial stewardship system-wide, including a 2014 directive that all VA facilities have an ASP,²⁷ since the 2012 HAIG assessment reported considerable variability in antimicrobial utilization and antimicrobial stewardship activities. Our study identifies areas of stewardship that may correlate with, positively or negatively, antimicrobial utilization measures that will require further investigation. A repeat and more detailed antimicrobial stewardship survey was recently completed and will help VA gauge ongoing effects of ASTF activities. We hope to re-evaluate our model with newer data when available.

Acknowledgments

The authors wish to thank Michael Fletcher, Jaime Lopez, and Catherine Loc-Carrillo for their administrative and organizational support of the project and Allison Kelly, MD, for her pivotal role in survey development and distribution. This work was supported by the VA Health Services Research and Development Service Collaborative Research to Enhance and Advance Transformation and Excellence (CREATE) Initiative; Cognitive Support Informatics for Antimicrobial Stewardship project (CRE 12-313).

Disclosure

 The authors report no financial conflicts of interest.

The deleterious impact of inappropriate and/or excessive antimicrobial usage is well recognized. In the United States, the Centers for Disease Control and Prevention (CDC) estimates that at least 2 million people become infected with antimicrobial-resistant bacteria with 23,000 subsequent deaths and at least $1 billion in excess medical costs per year.¹

In response, many healthcare organizations have developed antimicrobial stewardship programs (ASPs). Guidelines co-sponsored by the Infectious Diseases Society of America and the Society for Healthcare Epidemiology of America, as well as recent statements from the CDC and the Transatlantic Taskforce on Antimicrobial Resistance, all recommend core ASP elements.^2-5 The guidelines provide general recommendations on ASP structure, strategies, and activities. The recommended ASP structure is a team of physicians and pharmacists that collaborates with facility governing committees and other stakeholders to optimize antimicrobial use. While personnel with expertise in infectious diseases (ID) often lead ASPs, hospitalists are also recognized as key contributors, especially in quality improvement.^6,7 Recommended strategies include prospective audit of antimicrobial use with intervention and feedback and formulary restriction with preauthorization. Recommended activities include education, creation of guidelines, clinical pathways, and order forms, and programs to promote de-escalation and conversion from parenteral (IV) to oral (PO) antimicrobial therapy. However, limited evidence exists regarding the effectiveness of these ASP core elements.^8,9 While Cochrane reviews found clear evidence that particular stewardship strategies (eg, audit and feedback, formulary restriction, guidelines implemented with or without feedback, protocols, computerized decision support) can be effective in reducing antimicrobial usage and improving clinical outcomes over the long term, little evidence exists favoring 1 strategy over another.⁸ Furthermore, most individual studies of ASPs are single-center, making their conclusions less generalizable.

In 2012, the VA National Antimicrobial Stewardship Task Force (ASTF), in conjunction with the VA Healthcare Analysis and Information Group (HAIG) administered a survey on the characteristics of ASPs at all 130 acute care VA facilities (Appendix A). We used these survey results to build an implementation model and then assess associations between facility-level variables and 4 antimicrobial utilization measures.

Survey and Data

In 2011, the ASTF was chartered to develop, deploy, and monitor a strategic plan for optimizing antimicrobial therapy management. Monthly educational webinars and sample policies were offered to all facilities, including a sample business plan for stewardship and policies to encourage de-escalation from broad-spectrum antimicrobials, promote conversion from parenteral to oral antimicrobial therapy, avoid unnecessary double anaerobic coverage, and mitigate unnecessary antimicrobial usage in the context of Clostridium difficile infection.¹⁰

At the time that ASTF was chartered, the understanding of how ASP structures across VA facilities operated was limited. Hence, to capture baseline institutional characteristics and stewardship activities, ASTF and HAIG developed an inventory assessment of ASPs that was distributed online in November 2012. All 130 VA facilities providing inpatient acute care services responded.

We derived 57 facility characteristics relevant to antimicrobial utilization and conducted a series of factor analyses to simplify the complex dataset, and identify underlying latent constructs. We categorized resulting factors into domains of evidence, context, or facilitation as guided by the Promoting Action on Research Implementation in Health Services framework.¹¹ Briefly, the evidence domain describes how the facility uses codified and noncodified sources of knowledge (eg, research evidence, clinical experience). Organizational context comprises a facility’s characteristics that ensure a more conducive environment to put evidence into practice (eg, supportive leadership, organizational structure, evaluative systems). Facilitation emphasizes a facility personnel’s “state of preparedness” and receptivity to implementation.

Using factor analysis to identify facility factors as correlates of the outcomes, we first examined polychoric correlations among facility characteristics to assess multicollinearity. We performed independent component analysis to create latent constructs of variables that were defined by factor loadings (that indicated the proportion of variance accounted for by the construct) and uniqueness factors (that determined how well the variables were interpreted by the construct). Factors retained included variables that had uniqueness values of less than 0.7 and factor loadings greater than 0.3. Those associated with uniqueness values greater than 0.7 were left as single items, as were characteristics deemed a priori to be particularly important to antimicrobial stewardship. Factor scales that had only 2 items were converted into indices, while factor scores were generated for those factors that contained 3 or more items.^12-15

Data for facility-level antimicrobial utilization measures were obtained from the VA Corporate Data Warehouse from calendar year 2012. The analysis was conducted within the VA Informatics and Computing Infrastructure. All study procedures were approved by the VA Central Institutional Review Board.

Measures

Four utilization measures were defined as dependent measures: overall antimicrobial use; antimicrobial use in patients with non-infectious discharge diagnoses; missed opportunities to convert from parenteral to oral antimicrobial therapy; and missed opportunities to avoid double anaerobic coverage with metronidazole.

Overall antimicrobial use was defined as total acute care (ie, medical/surgical/intensive care) antibacterial use for each facility aggregated as per CDC National Healthcare Safety Network Antimicrobial Use Option guidelines (antimicrobial days per 1000 patient days present). A subanalysis of overall antimicrobial use was restricted to antimicrobial use among patients without an infection-related discharge diagnosis, as we surmised that this measure may capture a greater proportion of potentially unnecessary antimicrobial use. International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM)¹⁶ codes for infectious processes were identified by a combination of those classified previously in the literature,¹⁷ and those identified by finding the descendants of all infections named in the Systematized Nomenclature of Medicine--Clinical Terms.¹⁸ Next, all remaining codes for principal discharge diagnoses for which antimicrobials were administered were reviewed for potential indications for systemic antibacterial use. Discharges were considered noninfectious if no codes were identified when systemic antimicrobials were or could be indicated. For this measure, antimicrobial days were not counted if administered on or 1 day after the calendar day of surgery warranting antimicrobial prophylaxis.

Missed opportunities for conversion from parenteral to oral (IV to PO) formulations of highly bioavailable oral antimicrobials (ciprofloxacin, levofloxacin, moxifloxacin, azithromycin, clindamycin, linezolid, metronidazole, and fluconazole) were defined as the percentage of days of unnecessary IV therapy that were given when PO therapy could have been used among patients who were not in intensive care units at the time of antimicrobial administration who were receiving other oral medications, using previously described methodology.¹⁹ Missed opportunities for avoiding redundant anaerobic coverage with metronidazole were defined as the percentage of days in which patients receiving metronidazole also received antibiotics with activity against anaerobic bacteria, specifically beta-lactam/beta-lactamase inhibitors, carbapenems, cefotetan/cefoxitin, clindamycin, moxifloxacin, or tigecycline), using previously described methodology.²⁰ Patients for whom C. difficile testing was either ordered or positive within the prior 28 days (indicating potential clinical concern for C. difficile infection) were excluded from this endpoint.

Analysis

The variables derived above were entered into a multivariable model for each of the 4 antimicrobial utilization measures. The least absolute shrinkage and selection operator (LASSO) regression was used to determine significant associations between variables and individual utilization measures.²¹ LASSO was chosen because it offers advantages over traditional subset selection approaches in large multivariable analyses by assessing covariates simultaneously rather than sequentially, supporting prediction rather than estimation of effect.²²P values were not reported as they are not useful in determining statistical significance in this methodology. A tuning parameter of 0.025 was determined for the model based on a cross-validation approach. Significant variables remaining in the model were reported with the percent change in each utilization measure per unit change in the variable of interest. For binary factors, percent change was reported according to whether the variable was present or not. For ordinal variables, percent change was reported according to incremental increase in ordinal score. For continuous variables or variables represented by factor or index scores, percent change was reported per each 25% increase in the range of the score.

Inpatient Facility Antimicrobial Stewardship Characteristics and Antimicrobial Utilization

Frequencies of key facility characteristics that contributed to variable development are included in Table 1. Full survey results across all facilities are included in Appendix B. Factor analysis reduced the total number of variables to 32; however, we also included hospital size and VA complexity score. Thus, 34 variables were evaluated for association with antimicrobial utilization measures: 4 in the evidence domain, 23 in the context domain, and 7 in the facilitation domain (Table 2).

Median facility antimicrobial use was 619 antimicrobial days per 1000 days present (interquartile range [IQR], 554-700; overall range, 346-974). Median facility noninfectious antimicrobial use was 236 per 1000 days present (IQR, 200-286). Missed opportunities for conversion from IV to PO antimicrobial therapy were common, with a median facility value of 40.4% (391/969) of potentially eligible days of therapy (IQR, 32.2-47.8%). Missed opportunities to avoid double anaerobic coverage were less common (median 15.3% (186/1214) of potentially eligible days of therapy (IQR, 11.8%-20.2%; Figure).

Overall Antimicrobial Use

Four variables were associated with decreased overall antimicrobial use, although with small magnitude of change: presence of postgraduate physician/pharmacy training programs (0.03% decrease per quarter increase in factor score; on the order of 0.2 antimicrobial days per 1000 patient days present), presence of pharmacists and/or ID attendings on general medicine ward teams (0.02% decrease per quarter increase in index score), frequency of systematic de-escalation review (0.01% decrease per ordinal increase in score), and degree of involvement of ID physicians and/or fellows in antimicrobial approvals (0.007% decrease per quarter increase in index score). No variables were associated with increased overall antimicrobial use.

Antimicrobial Use among Discharges without Infectious Diagnoses

Six variables were associated with decreased antimicrobial use in patients without infectious discharge diagnoses, while 4 variables were associated with increased use. Variables associated with the greatest magnitude of decreased use included facility educational programs for prudent antimicrobial use (1.8% on the order of 4 antimicrobial days per 1000 patient days present), frequency of systematic de-escalation review (1.5% per incremental increase in score), and whether a facility’s lead antimicrobial stewardship pharmacist had ID training (1.3%). Also significantly associated with decreased use was a factor summarizing the presence of 4 condition-specific stewardship processes (de-escalation policies, policies for addressing antimicrobial use in the context of C. difficile infection, blood culture review, and automatic ID consults for certain conditions) (0.6% per quarter increase in factor score range), the extent to which postgraduate physician/pharmacy training programs were present (0.6% per quarter increase in factor score range), and the number of electronic antimicrobial-specific order sets present (0.4% per order set). The variables associated with increased use of antimicrobials included the presence of antimicrobial stop orders (4.6%), the degree to which non-ID physicians were involved in antimicrobial approvals (0.7% per increase in ordinal score), the level engagement with ASTF online resources (0.6% per quarter increase in factor score range), and hospital size (0.6% per 50-bed increase).

Missed Opportunities for Parenteral to Oral Antimicrobial Conversion

Missed opportunities for IV to PO antimicrobial conversion had the largest number of significant associations with organizational variables: 14 variables were associated with fewer missed opportunities, while 5 were associated with greater missed opportunities. Variables associated with the largest reductions in missed opportunities for IV to PO conversion included having guidelines for antimicrobial duration (12.8%), participating in regional stewardship collaboratives (8.1%), number of antimicrobial-specific order sets (6.0% per order set), ID training of the ASP pharmacist (4.9%), and VA facility complexity designation (4.2% per quarter increase in score indicating greater complexity).²³ Variables associated with more missed opportunities included stop orders (11.7%), overall perceived receptiveness to antimicrobial stewardship among clinical services (9.4%), the degree of engagement with ASTF online resources (6.9% per quarter increase in factor score range), educational programs for prudent antimicrobial use (4.1%), and hospital size (1.0% per 50-bed increase).

Missed Opportunities for Avoidance of Double Anaerobic Coverage

Four variables were associated with more avoidance of double anaerobic coverage: ID training of the lead ASP pharmacist (8.8%), presence of pharmacists and/or ID attendings on acute care ward teams (6.2% per quarter increase in index score), degree of ID pharmacist involvement in antimicrobial approvals, ranging from not at all (score=0) to both weekdays and nights/weekends (score=2; 4.3% per ordinal increase), and the number of antimicrobial-specific order sets (1.5% per order set). No variables were associated with less avoidance of double anaerobic coverage.

Variables Associated with Multiple Favorable or Unfavorable Antimicrobial Utilization Measures

To better assess the consistency of the relationship between organizational variables and measures of antimicrobial use, we tabulated variables that were associated with at least 3 potentially favorable (ie, reduced overall or noninfectious antimicrobial use or fewer missed opportunities) measures. Altogether, 5 variables satisfied this criterion: the presence of postgraduate physician/pharmacy training programs, the number of antimicrobial-specific order sets, frequency of systematic de-escalation review, the presence of pharmacists and/or ID attendings on acute care ward teams, and formal ID training of the lead ASP pharmacist (Table 3). Three other variables were associated with at least 2 unfavorable measures: hospital size, the degree to which the facility engaged with ASTF online resources, and presence of antimicrobial stop orders.

Variability in ASP implementation across VA allowed us to assess the relationship between ASP and facility elements and baseline patterns of antimicrobial utilization. Hospitalists and hospital policy-makers are becoming more and more engaged in inpatient antimicrobial stewardship. While our results suggest that having pharmacists and/or physicians with formal ID training participate in everyday inpatient activities can favorably improve antimicrobial utilization, considerable input into stewardship can be made by hospitalists and policy makers. In particular, based on this work, the highest yield from an organizational standpoint may be in working to develop order sets within the electronic medical record and systematic efforts to promote de-escalation of broad-spectrum therapy, as well as encouraging hospital administration to devote specific physician and pharmacy salary support to stewardship efforts.

While we noted that finding the ASTF online resources helpful was associated with potentially unfavorable antimicrobial utilization, we speculate that this may represent reverse causality due to facilities recognizing that their antimicrobial usage is suboptimal and thus seeking out sample ASTF policies to implement. The association between the presence of automatic stop orders and potentially unfavorable antimicrobial utilization is less clear since the timeframe was not specified in the survey; it may be that setting stop orders too far in advance may promote an environment in which critical thinking about antimicrobial de-escalation is not encouraged or timely. The larger magnitude of association between ASP characteristics and antimicrobial usage among patients without infectious discharge diagnoses versus overall antimicrobial usage also suggests that clinical situations where infection was of low enough suspicion to not even have the providers eventually list an infectious diagnosis on their discharge summaries may be particularly malleable to ASP interventions, though further exploration is needed in determining how useful this utilization measure may be as a marker for inappropriate antimicrobial use.

Our results complement those of Pakyz et al.²⁴ who surveyed 44 academic medical facilities in March 2013 to develop an ASP intensity score and correlate this score and its specific components to overall and targeted antimicrobial use. This study found that the overall ASP intensity score was not significantly associated with total or targeted antimicrobial use. However, ASP strategies were more associated with decreased total and targeted antimicrobial use than were specific ASP resources. In particular, the presence of a preauthorization strategy was associated with decreased targeted antimicrobial use. Our particular findings that order set establishment and de-escalation efforts are associated with multiple antibiotic outcomes also line up with the findings of Schuts et al, who performed a meta-analysis of the effects of meeting antimicrobial stewardship objectives and found that achieving guideline concordance (such as through establishment of order sets) and successfully de-escalating antimicrobial therapy was associated with reduced mortality.^25,26 This meta-analysis, however, was limited by low rigor of its studies and potential for reverse causality. While our study has the advantages of capturing an entire national network of 130 acute care facilities with a 100% response rate, it, too, is limited by a number of issues, most notably by the fact that the survey was not specifically designed for the analysis of antimicrobial utilization measures, patient-level risk stratification was not available, the VA population does not reflect the U.S. population at-large, recall bias, and that antimicrobial prescribing and stewardship practices have evolved in VA since 2012. Furthermore, all of the antimicrobial utilization measures studied are imperfect at capturing inappropriate antibiotic use; in particular, our reliance on principal ICD-9 codes for noninfectious outcomes requires prospective validation. Many survey questions were subjective and subject to misinterpretation; other unmeasured confounders may also be present. Causality cannot be inferred from association. Nevertheless, our findings support many core indicators for hospital ASP recommended by the CDC and the Transatlantic Taskforce on Antimicrobial Resistance,^3,4 most notably, having personnel with ID training involved in stewardship and establishing a formal procedure for ASP review for the appropriateness of an antimicrobial at or after 48 hours from the initial order.

In summary, the VA has made efforts to advance the practice of antimicrobial stewardship system-wide, including a 2014 directive that all VA facilities have an ASP,²⁷ since the 2012 HAIG assessment reported considerable variability in antimicrobial utilization and antimicrobial stewardship activities. Our study identifies areas of stewardship that may correlate with, positively or negatively, antimicrobial utilization measures that will require further investigation. A repeat and more detailed antimicrobial stewardship survey was recently completed and will help VA gauge ongoing effects of ASTF activities. We hope to re-evaluate our model with newer data when available.

Acknowledgments

The authors wish to thank Michael Fletcher, Jaime Lopez, and Catherine Loc-Carrillo for their administrative and organizational support of the project and Allison Kelly, MD, for her pivotal role in survey development and distribution. This work was supported by the VA Health Services Research and Development Service Collaborative Research to Enhance and Advance Transformation and Excellence (CREATE) Initiative; Cognitive Support Informatics for Antimicrobial Stewardship project (CRE 12-313).

Disclosure

 The authors report no financial conflicts of interest.

Older adults commonly experience insomnia and agitation during hospitalization. Unfortunately, the use of benzodiazepines and sedative hypnotics (BSH) to treat these conditions can be ineffective and expose patients to significant adverse effects.^1,2 Choosing Wisely^® is a campaign that promotes dialogue to reduce unnecessary medical tests, procedures, or treatments. This international campaign has highlighted BSHs as potentially harmful and has recommended against their use as first-line treatment of insomnia and agitation.^3-5 Examples of harm with benzodiazepine use include cognitive impairment, impaired postural stability, and an increased incidence of falls and hip fractures in both community and acute care settings.^6-8 In addition, prescriptions initiated in hospital appear to be associated with a higher risk of falls and unplanned readmission.^9,10 The newer nonbenzodiazepine sedative hypnotics, commonly referred to as “z-drugs”, were initially marketed as a safer alternative in older adults due to their more favorable pharmacokinetics. Evidence has emerged that they carry similar risks.^6,11,12 A study comparing benzodiazepines and zolpidem found relatively greater risk of fractures requiring hospitalization with the use of zolpidem compared to lorazepam.¹³

The use of benzodiazepines in the acute care setting has been evaluated in a number of studies and ranges from 20% to 45%.^14-16 Few studies focus on the initiation of these medications in BSH-naïve hospitalized patients; however, reports range from 18% to 29%.^17,18 Factors found to be associated with potentially inappropriate prescriptions (PIP) include Hispanic ethnicity, residing in an assisted care setting, and a greater number of BSH prescriptions prior to admission.^16,19 Additionally, Cumbler et al.¹⁵ found that the presence of dementia was associated with fewer prescriptions for sleep aids in hospital. To our knowledge, there are no published studies that have investigated prescriber factors associated with the use of BSH.

The purpose of our study was to determine the frequency of PIPs of BSH in our academic hospital. Additionally, we aimed to identify patient and prescriber factors that were associated with increased likelihood of prescriptions to help guide future quality improvement initiatives.

Study Design and Setting

This was a retrospective observational study conducted at Mount Sinai Hospital (MSH) in Toronto over a 4-month period from January 2013 to April 2013. The hospital is a 442-bed acute care academic health science center affiliated with the University of Toronto. The MSH electronic health record contains demographic data, medications and allergies, nursing documentation, and medical histories from prior encounters. It also includes computerized physician order entry (CPOE) and a detailed medication administration record. This system is integrated with an electronic pharmacy database used to monitor and dispense medications for each patient.

Patient and Medication Selection

We included inpatients over the age of 65 who were prescribed a BSH during the study period from the following services: general internal medicine, cardiology, general surgery, orthopedic surgery, and otolaryngology. To identify new exposure to BSHs, we excluded patients who were regularly prescribed a BSH prior to admission to hospital. The medications of interest included all benzodiazepines and the nonbenzodiazepine sedative hypnotic, zopiclone. Zopiclone is the most commonly used nonbenzodiazepine sedative hypnotic in Canada and the only 1 available on our hospital formulary. These were selected based on the strength of evidence to recommend against their use as first-line agents in older adults and in consultation with our geriatric medicine consultation team pharmacist.²⁰

Data Collection

The hospital administrative database provided patient demographic information, admission service, admitting diagnosis, length of stay, and the total number of patients discharged from the study units over the study period. We then searched the pharmacy electronic database for all benzodiazepines and zopiclone prescribed during the study period for patients who met the inclusion criteria. Manual review of paper and electronic health records for this cohort of patients was conducted to extract additional variables. We used a standardized form to record data elements. Dr. Pek collected all data elements. Dr. Remfry reviewed a random sample of patient records (10%) to ensure accuracy. The agreement between reviewers was 100%.

In compliance with hospital accreditation standards, a clinical pharmacist documents a best possible medication history (BPMH) on every inpatient on admission. We used the BPMH to identify and exclude patients who were prescribed a BSH prior to hospitalization. Because all medications were ordered through the CPOE system, as-needed medication prescriptions required the selection of a specified indication. Available options included ‘agitation/anxiety’ and necessitated combining these 2 indications into 1 category. Indications were primarily extracted through electronic order entry reviews. Paper charts were reviewed when further clarification was needed.

We identified ordering physicians’ training level and familiarity with the service from administrative records obtained from medical education offices, hospital records, and relevant call schedules. Fellows were defined as trainees with a minimum of 6 years of postgraduate training.

Our primary outcome of interest was the proportion of eligible patients age 65 and older who received a PIP for a BSH. Patient variables of interest included age, sex, comorbid conditions, and a pre-admission diagnosis of dementia. Comorbid conditions and age were used to calculate the Carlson Comorbidity Index for each patient.²¹ Prescription variables included the medication prescribed, time of first prescription (“overnight hours” refer to prescriptions ordered after 7:00 PM and before 7:00 AM), and whether the medication was ordered as part of an admission or postoperative order set. To determine whether patients were discharged home with a prescription for a BSH, we reviewed electronic discharge prescriptions of BSH-naïve patients who received a sedative in hospital. Only medical and cardiology inpatients receive electronic discharge prescriptions, and these were available for 189 patients in our cohort. Provider variables included training level, service, and familiarity with patients. We used the provider’s training program or department of appointment to define the ‘physician on-service’ variable. As an example, a resident registered in internal medicine is defined as ‘on-service’ when prescribing sedatives for a medical inpatient. In contrast, a psychiatry resident would be considered “off-service” if he prescribed a sedative for a surgical inpatient. The familiarity of a provider was categorized as ‘regular’ if they were responsible for a patient’s care on a day-to-day basis and ‘covering’ if they were only covering on call. Other variables included admitting service and hospital length of stay.

Appropriateness Criteria

Criteria for potentially inappropriate use were modified from the American and Canadian Geriatrics Societies’ Choosing Wisely recommendations,^4,5 and included insomnia and agitation. These recommendations are in line with other evidence based guidelines for safe prescribing in older adults.²⁰ For the purposes of our study, prescriptions for “agitation/anxiety”, “agitation”, or “insomnia/sleep” were considered potentially inappropriate. Appropriate indications included alcohol withdrawal, end-of-life symptom control, preprocedural sedation, and seizure.⁵ Patients who were already using a BSH prior to admission for any indication, including a psychiatric diagnosis, were excluded.

Statistical Analyses

We determined the proportion of patients with at least one PIP, as well as the proportion of all prescribing events that were potentially inappropriate. We used the Chi-square statistic and 2-sample t tests to compare the unadjusted associations between patient-level characteristics and receipt of at least 1 inappropriate prescription and prescribing event-level factors with inappropriate prescriptions. Given that first-year residents are more likely to be working overnight when most PIPs are prescribed, we performed a simple logistic regression of potentially inappropriate prescribing by level of training stratified by time of prescription. A multivariable random-intercept logistic regression model was used to assess the adjusted association between patient- and prescribing event-level characteristics with inappropriate prescribing, adjusting for clustering of prescribing events within patients. Characteristics of interest were identified a priori and those with significant bivariate associations with potentially inappropriate were selected for inclusion in the model. Additionally, we included time of prescription in our model to control for potential confounding. All statistical analyses were performed using SAS 9.4 (SAS Institute Inc, Cary, North Carolina). The MSH Research Ethics Board approved the study.

Description of Patients Prescribed a Benzodiazepine Sedative Hypnotic

There were 1540 patients over the age of 65 discharged during the 4-month study period. We excluded the 232 patients who had been prescribed a BSH prior to admission. Of the remaining eligible 1308 BSH-naïve patients, 251 (19.2%) were prescribed a new BSH in hospital and were included in the study. Of this cohort of 251 patients, 193 (76.9%) patients were prescribed a single BSH during their admission while 58 (23.1%) received 2 or more. Of all eligible patients, 208 (15.9%) were prescribed at least 1 PIP. Approximately half of the cohort was admitted to the general internal medicine service, and the most common reason for admission was cardiovascular disease (Table 1).

Description of Prescriptions of Benzodiazepine Sedative Hypnotic

We reviewed 328 prescriptions for BSH during the study period. The majority of these, 254 (77.4%) were potentially inappropriate (Table 2). The most common PIPs were zopiclone (167; 65.7%) and lorazepam (82; 32.3%). The PIPs were most frequently ordered on an as-needed basis (219; 86%), followed by one-time orders (30; 12%), and standing orders (5; 2%). The majority of PIPs (222; 87.4%) was prescribed for insomnia with a minority (32; 12.6%) prescribed for agitation and/or anxiety.

Most PIP were prescribed during overnight hours (159; 62.6%) and when an in-house pharmacist was unavailable (211; 83.1%). These variables were highly correlated with prescription of sleep aid, which was defined in our criteria as potentially inappropriate. Copies of discharge prescriptions were available for 189 patients. Of these 189 patients, 19 (10.1%) were sent home with a prescription for a new sedative.

Association Between Patient/Provider Variables and Prescriptions

Patient factors associated with fewer PIPs in our bivariate analyses included older age and dementia (Table 1). A greater proportion of nighttime prescriptions were PIPs; however, this finding was not statistically significant (P = 0.067). The majority of all prescriptions was prescribed by residents in their first year of training (64.9%; Table 2), and there was a significant difference in rates of PIP across level of training (P = 0.0007). When stratified by time of prescription, there was no significant difference by level of training for nighttime prescriptions. Among daytime prescriptions, second-year residents and staff (attending physicians and fellows) were less likely to prescribe a PIP than first-year residents (odds ratio [OR], 0.24; 95% confidence interval [CI], 0.09-0.66 and OR, 0.39; 95% CI, 0.14-1.13, respectively; Table 3); however, the association between staff and first-years only approached statistical significance (P = 0.08). Interestingly, 20.4% of all PIPs were ordered routinely as part of an admission or postoperative order set.

In our regression model, admission to a specialty or surgical service, compared to the general internal medicine service, was associated with a significantly higher likelihood of a PIP (OR, 6.61; 95% CI, 2.70-16.17; Table 4). Additionally, compared to cardiovascular admission diagnoses, neoplastic admitting diagnoses were associated with a higher likelihood of a PIP (OR, 4.43; 95% CI, 1.23-15.95). Time of prescription was a significant predictor in our multivariable regression model with nighttime prescriptions having increased odds of a PIP (OR, 4.48; 95% CI, 2.21-9.06,). When comparing prescribers at the extremes of training, attending physicians and fellows were much less likely to prescribe a PIP compared to first-year residents (OR, 0.23; 95% CI, 0.08-0.69; Table 4). However, there were no other significant differences across training levels after adjusting for patient and prescribing event characteristics.

We found that the majority of newly prescribed BSH in hospital was for the potentially inappropriate indications of insomnia and agitation/anxiety. Medications for insomnia were primarily initiated during overnight hours. Training level of prescribers and admitting service were found to be associated with appropriateness of prescriptions.

Our study showed that 15.9% of hospitalized older adults were newly prescribed a PIP during their admission. Of all new in hospital prescriptions, 77% were deemed potentially inappropriate. These numbers are similar to those reported by other centers; however, wide ranges exist.^16,19 This is likely the result of differences in appropriate use and inclusion criteria. Gillis et al.¹⁷ focused their investigation on sleep aids and showed that 26% of all admitted patients and 18% of BSH naïve patients received a prescription for insomnia. While this is similar to our findings, more than half of these patients were under the age of 65, and additional medications, such as trazodone, antihistamines, and antipsychotics were included.¹⁷ Other studies did not exclude patients who used a BSH regularly prior to admission. For example, 21% of veterans admitted to an acute care facility received a prescription for potentially inappropriate indications, but this included continuation of prior home medications.¹⁹ In contrast, we chose to focus on older adults in whom BSH pose a greater risk of harm. Exclusion of patients who regularly used a BSH prior to admission allowed us to better understand the circumstances surrounding the initiation of these medications in hospital. Furthermore, abrupt cessation of benzodiazepines can cause withdrawal and worsen confusion.²²

We found that 10% of patients newly prescribed a BSH in hospital were discharged with a prescription for a BSH. The accuracy of this is limited by the lack of availability of electronic discharge prescriptions on our surgical wards; however, it is likely an underrepresentation of the true effect given the high rates of PIPs on these wards. Our study highlights the concerning practice of continuing newly prescribed BSH following discharge from hospital.

Sleep disruption and poor quality sleep in hospital is a common issue that leads to significant use of BSH.¹⁵ Nonpharmacologic interventions in older adults can be effective in improving sleep quality and reducing the need for BSH; however, they can be time-consuming to implement.²³ With the exception of preventative strategies used on our Acute Care for Elders unit, formal nonpharmacologic interventions for sleep are not practiced in our hospital. We found that the majority of PIPs were prescribed as sleep aids in the overnight hours. This suggests that disruptions in sleep are leading patients and nursing staff to request pharmacologic treatments and highlights an area with significant room for improvement. Work is underway to implement and evaluate safe sleep protocols for older adults.

To our knowledge, we are the first to report an association between training level and PIP of BSH in older adults. The highest rates of PIPs were found among the first-year residents and, after controlling for patient and prescribing event characteristics, such as time of prescription, first-year residents were significantly more likely to prescribe a PIP. First-year residents are more likely to respond first to issues on the wards. There may be pressure on first-year trainees to prescribe sleep aids, as many patients and nurses may seek pharmacologic solutions for symptom management. Knowledge gaps may also be a contributing factor early in their training. A survey of physicians found that residents were more likely than attending physicians to list lack of formal education as a barrier to appropriate prescribing.²⁴

Similarities are seen in a study of antibiotic appropriateness, where residents demonstrated gaps in knowledge of treatment of asymptomatic bacteriuria that seemed to vary by specialty.²⁵ Interestingly, we found that patients admitted to general internal medicine were prescribed fewer PIPs. This service includes our Acute Care for Elders unit, which is staffed by trained geriatric nurses and other allied health professionals. Residents who rotated on internal medicine are also likely to have received informal teaching about medication safety in older adults. Educational interventions highlighting adverse effects of BSH and promoting nonpharmacologic solutions should be targeted at first-year residents. However, an interprofessional team approach to sleep disturbance in hospital, in combination with decision support for appropriate BSH use will achieve greater impact than education alone.

Several limitations of this study merit discussion. First, findings from a single academic center may lack generalizability. However, the demographics of our patient population and our rates of BSH use were similar to those reported in previous studies. Second, our study may be subject to observer bias, as the data collectors were not blinded. To minimize this, a strict template and clear appropriateness criteria were developed. Additionally, a second reviewer independently conducted data validation with 100% agreement among reviewers. Third, we studied prescribing patterns rather than medication administration and lacked data on filling of new BSH prescriptions in the postdischarge period. However, our primary goal is to determine risk of exposure to a BSH to minimize it. Fourth, although BSH are discouraged as “first choice for insomnia, anxiety or delirium,”⁴ they may be appropriate in limited situations where all nonpharmacologic strategies have failed and patient or staff safety is at risk. In our chart reviews, we were unable to determine whether all nonpharmacologic strategies were exhausted prior to prescription initiation. However, more than 20% of all PIP were routinely prescribed as part of an admission or postoperative order set, suggesting a reflexive rather than reflective approach to sedative use. Furthermore, the indications of anxiety and agitation were combined as they appear in the CPOE as a combination indication, thus leaving us unable to determine the true proportion for each indication. However, more than 87% of all PIPs were for insomnia, reflecting a clear opportunity to improve sleep management in hospital. Last, the lack of a power calculation may have resulted in the study being underpowered and thus affected the ability to detect a significant effect of covariates that have real differences on the likelihood of sedative prescriptions. For example, the low number of prescribing events by second-year residents and staff may have resulted in a type II error when comparing PIP rates with first-year residents.

We found that the majority of newly prescribed BSH among older adults in hospital were potentially inappropriate. They were most frequently prescribed by first-year residents overnight in response to insomnia. Our findings demonstrate BSH overuse remains prevalent and is associated with poor sleep in hospital. Future work will focus on implementing and evaluating safe sleep protocols and educational interventions aimed at first-year residents.

Acknowledgments

Elisabeth Pek had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Ciara Pendrith conducted and is responsible for the statistical analysis.

Disclosure

The authors report no financial conflicts of interest.

Older adults commonly experience insomnia and agitation during hospitalization. Unfortunately, the use of benzodiazepines and sedative hypnotics (BSH) to treat these conditions can be ineffective and expose patients to significant adverse effects.^1,2 Choosing Wisely^® is a campaign that promotes dialogue to reduce unnecessary medical tests, procedures, or treatments. This international campaign has highlighted BSHs as potentially harmful and has recommended against their use as first-line treatment of insomnia and agitation.^3-5 Examples of harm with benzodiazepine use include cognitive impairment, impaired postural stability, and an increased incidence of falls and hip fractures in both community and acute care settings.^6-8 In addition, prescriptions initiated in hospital appear to be associated with a higher risk of falls and unplanned readmission.^9,10 The newer nonbenzodiazepine sedative hypnotics, commonly referred to as “z-drugs”, were initially marketed as a safer alternative in older adults due to their more favorable pharmacokinetics. Evidence has emerged that they carry similar risks.^6,11,12 A study comparing benzodiazepines and zolpidem found relatively greater risk of fractures requiring hospitalization with the use of zolpidem compared to lorazepam.¹³

The use of benzodiazepines in the acute care setting has been evaluated in a number of studies and ranges from 20% to 45%.^14-16 Few studies focus on the initiation of these medications in BSH-naïve hospitalized patients; however, reports range from 18% to 29%.^17,18 Factors found to be associated with potentially inappropriate prescriptions (PIP) include Hispanic ethnicity, residing in an assisted care setting, and a greater number of BSH prescriptions prior to admission.^16,19 Additionally, Cumbler et al.¹⁵ found that the presence of dementia was associated with fewer prescriptions for sleep aids in hospital. To our knowledge, there are no published studies that have investigated prescriber factors associated with the use of BSH.

The purpose of our study was to determine the frequency of PIPs of BSH in our academic hospital. Additionally, we aimed to identify patient and prescriber factors that were associated with increased likelihood of prescriptions to help guide future quality improvement initiatives.

Study Design and Setting

This was a retrospective observational study conducted at Mount Sinai Hospital (MSH) in Toronto over a 4-month period from January 2013 to April 2013. The hospital is a 442-bed acute care academic health science center affiliated with the University of Toronto. The MSH electronic health record contains demographic data, medications and allergies, nursing documentation, and medical histories from prior encounters. It also includes computerized physician order entry (CPOE) and a detailed medication administration record. This system is integrated with an electronic pharmacy database used to monitor and dispense medications for each patient.

Patient and Medication Selection

We included inpatients over the age of 65 who were prescribed a BSH during the study period from the following services: general internal medicine, cardiology, general surgery, orthopedic surgery, and otolaryngology. To identify new exposure to BSHs, we excluded patients who were regularly prescribed a BSH prior to admission to hospital. The medications of interest included all benzodiazepines and the nonbenzodiazepine sedative hypnotic, zopiclone. Zopiclone is the most commonly used nonbenzodiazepine sedative hypnotic in Canada and the only 1 available on our hospital formulary. These were selected based on the strength of evidence to recommend against their use as first-line agents in older adults and in consultation with our geriatric medicine consultation team pharmacist.²⁰

Data Collection

The hospital administrative database provided patient demographic information, admission service, admitting diagnosis, length of stay, and the total number of patients discharged from the study units over the study period. We then searched the pharmacy electronic database for all benzodiazepines and zopiclone prescribed during the study period for patients who met the inclusion criteria. Manual review of paper and electronic health records for this cohort of patients was conducted to extract additional variables. We used a standardized form to record data elements. Dr. Pek collected all data elements. Dr. Remfry reviewed a random sample of patient records (10%) to ensure accuracy. The agreement between reviewers was 100%.

In compliance with hospital accreditation standards, a clinical pharmacist documents a best possible medication history (BPMH) on every inpatient on admission. We used the BPMH to identify and exclude patients who were prescribed a BSH prior to hospitalization. Because all medications were ordered through the CPOE system, as-needed medication prescriptions required the selection of a specified indication. Available options included ‘agitation/anxiety’ and necessitated combining these 2 indications into 1 category. Indications were primarily extracted through electronic order entry reviews. Paper charts were reviewed when further clarification was needed.

We identified ordering physicians’ training level and familiarity with the service from administrative records obtained from medical education offices, hospital records, and relevant call schedules. Fellows were defined as trainees with a minimum of 6 years of postgraduate training.

Our primary outcome of interest was the proportion of eligible patients age 65 and older who received a PIP for a BSH. Patient variables of interest included age, sex, comorbid conditions, and a pre-admission diagnosis of dementia. Comorbid conditions and age were used to calculate the Carlson Comorbidity Index for each patient.²¹ Prescription variables included the medication prescribed, time of first prescription (“overnight hours” refer to prescriptions ordered after 7:00 PM and before 7:00 AM), and whether the medication was ordered as part of an admission or postoperative order set. To determine whether patients were discharged home with a prescription for a BSH, we reviewed electronic discharge prescriptions of BSH-naïve patients who received a sedative in hospital. Only medical and cardiology inpatients receive electronic discharge prescriptions, and these were available for 189 patients in our cohort. Provider variables included training level, service, and familiarity with patients. We used the provider’s training program or department of appointment to define the ‘physician on-service’ variable. As an example, a resident registered in internal medicine is defined as ‘on-service’ when prescribing sedatives for a medical inpatient. In contrast, a psychiatry resident would be considered “off-service” if he prescribed a sedative for a surgical inpatient. The familiarity of a provider was categorized as ‘regular’ if they were responsible for a patient’s care on a day-to-day basis and ‘covering’ if they were only covering on call. Other variables included admitting service and hospital length of stay.

Appropriateness Criteria

Criteria for potentially inappropriate use were modified from the American and Canadian Geriatrics Societies’ Choosing Wisely recommendations,^4,5 and included insomnia and agitation. These recommendations are in line with other evidence based guidelines for safe prescribing in older adults.²⁰ For the purposes of our study, prescriptions for “agitation/anxiety”, “agitation”, or “insomnia/sleep” were considered potentially inappropriate. Appropriate indications included alcohol withdrawal, end-of-life symptom control, preprocedural sedation, and seizure.⁵ Patients who were already using a BSH prior to admission for any indication, including a psychiatric diagnosis, were excluded.

Statistical Analyses

We determined the proportion of patients with at least one PIP, as well as the proportion of all prescribing events that were potentially inappropriate. We used the Chi-square statistic and 2-sample t tests to compare the unadjusted associations between patient-level characteristics and receipt of at least 1 inappropriate prescription and prescribing event-level factors with inappropriate prescriptions. Given that first-year residents are more likely to be working overnight when most PIPs are prescribed, we performed a simple logistic regression of potentially inappropriate prescribing by level of training stratified by time of prescription. A multivariable random-intercept logistic regression model was used to assess the adjusted association between patient- and prescribing event-level characteristics with inappropriate prescribing, adjusting for clustering of prescribing events within patients. Characteristics of interest were identified a priori and those with significant bivariate associations with potentially inappropriate were selected for inclusion in the model. Additionally, we included time of prescription in our model to control for potential confounding. All statistical analyses were performed using SAS 9.4 (SAS Institute Inc, Cary, North Carolina). The MSH Research Ethics Board approved the study.

Description of Patients Prescribed a Benzodiazepine Sedative Hypnotic

There were 1540 patients over the age of 65 discharged during the 4-month study period. We excluded the 232 patients who had been prescribed a BSH prior to admission. Of the remaining eligible 1308 BSH-naïve patients, 251 (19.2%) were prescribed a new BSH in hospital and were included in the study. Of this cohort of 251 patients, 193 (76.9%) patients were prescribed a single BSH during their admission while 58 (23.1%) received 2 or more. Of all eligible patients, 208 (15.9%) were prescribed at least 1 PIP. Approximately half of the cohort was admitted to the general internal medicine service, and the most common reason for admission was cardiovascular disease (Table 1).

Description of Prescriptions of Benzodiazepine Sedative Hypnotic

We reviewed 328 prescriptions for BSH during the study period. The majority of these, 254 (77.4%) were potentially inappropriate (Table 2). The most common PIPs were zopiclone (167; 65.7%) and lorazepam (82; 32.3%). The PIPs were most frequently ordered on an as-needed basis (219; 86%), followed by one-time orders (30; 12%), and standing orders (5; 2%). The majority of PIPs (222; 87.4%) was prescribed for insomnia with a minority (32; 12.6%) prescribed for agitation and/or anxiety.

Most PIP were prescribed during overnight hours (159; 62.6%) and when an in-house pharmacist was unavailable (211; 83.1%). These variables were highly correlated with prescription of sleep aid, which was defined in our criteria as potentially inappropriate. Copies of discharge prescriptions were available for 189 patients. Of these 189 patients, 19 (10.1%) were sent home with a prescription for a new sedative.

Association Between Patient/Provider Variables and Prescriptions

Patient factors associated with fewer PIPs in our bivariate analyses included older age and dementia (Table 1). A greater proportion of nighttime prescriptions were PIPs; however, this finding was not statistically significant (P = 0.067). The majority of all prescriptions was prescribed by residents in their first year of training (64.9%; Table 2), and there was a significant difference in rates of PIP across level of training (P = 0.0007). When stratified by time of prescription, there was no significant difference by level of training for nighttime prescriptions. Among daytime prescriptions, second-year residents and staff (attending physicians and fellows) were less likely to prescribe a PIP than first-year residents (odds ratio [OR], 0.24; 95% confidence interval [CI], 0.09-0.66 and OR, 0.39; 95% CI, 0.14-1.13, respectively; Table 3); however, the association between staff and first-years only approached statistical significance (P = 0.08). Interestingly, 20.4% of all PIPs were ordered routinely as part of an admission or postoperative order set.

In our regression model, admission to a specialty or surgical service, compared to the general internal medicine service, was associated with a significantly higher likelihood of a PIP (OR, 6.61; 95% CI, 2.70-16.17; Table 4). Additionally, compared to cardiovascular admission diagnoses, neoplastic admitting diagnoses were associated with a higher likelihood of a PIP (OR, 4.43; 95% CI, 1.23-15.95). Time of prescription was a significant predictor in our multivariable regression model with nighttime prescriptions having increased odds of a PIP (OR, 4.48; 95% CI, 2.21-9.06,). When comparing prescribers at the extremes of training, attending physicians and fellows were much less likely to prescribe a PIP compared to first-year residents (OR, 0.23; 95% CI, 0.08-0.69; Table 4). However, there were no other significant differences across training levels after adjusting for patient and prescribing event characteristics.

We found that the majority of newly prescribed BSH in hospital was for the potentially inappropriate indications of insomnia and agitation/anxiety. Medications for insomnia were primarily initiated during overnight hours. Training level of prescribers and admitting service were found to be associated with appropriateness of prescriptions.

Our study showed that 15.9% of hospitalized older adults were newly prescribed a PIP during their admission. Of all new in hospital prescriptions, 77% were deemed potentially inappropriate. These numbers are similar to those reported by other centers; however, wide ranges exist.^16,19 This is likely the result of differences in appropriate use and inclusion criteria. Gillis et al.¹⁷ focused their investigation on sleep aids and showed that 26% of all admitted patients and 18% of BSH naïve patients received a prescription for insomnia. While this is similar to our findings, more than half of these patients were under the age of 65, and additional medications, such as trazodone, antihistamines, and antipsychotics were included.¹⁷ Other studies did not exclude patients who used a BSH regularly prior to admission. For example, 21% of veterans admitted to an acute care facility received a prescription for potentially inappropriate indications, but this included continuation of prior home medications.¹⁹ In contrast, we chose to focus on older adults in whom BSH pose a greater risk of harm. Exclusion of patients who regularly used a BSH prior to admission allowed us to better understand the circumstances surrounding the initiation of these medications in hospital. Furthermore, abrupt cessation of benzodiazepines can cause withdrawal and worsen confusion.²²

We found that 10% of patients newly prescribed a BSH in hospital were discharged with a prescription for a BSH. The accuracy of this is limited by the lack of availability of electronic discharge prescriptions on our surgical wards; however, it is likely an underrepresentation of the true effect given the high rates of PIPs on these wards. Our study highlights the concerning practice of continuing newly prescribed BSH following discharge from hospital.

Sleep disruption and poor quality sleep in hospital is a common issue that leads to significant use of BSH.¹⁵ Nonpharmacologic interventions in older adults can be effective in improving sleep quality and reducing the need for BSH; however, they can be time-consuming to implement.²³ With the exception of preventative strategies used on our Acute Care for Elders unit, formal nonpharmacologic interventions for sleep are not practiced in our hospital. We found that the majority of PIPs were prescribed as sleep aids in the overnight hours. This suggests that disruptions in sleep are leading patients and nursing staff to request pharmacologic treatments and highlights an area with significant room for improvement. Work is underway to implement and evaluate safe sleep protocols for older adults.

To our knowledge, we are the first to report an association between training level and PIP of BSH in older adults. The highest rates of PIPs were found among the first-year residents and, after controlling for patient and prescribing event characteristics, such as time of prescription, first-year residents were significantly more likely to prescribe a PIP. First-year residents are more likely to respond first to issues on the wards. There may be pressure on first-year trainees to prescribe sleep aids, as many patients and nurses may seek pharmacologic solutions for symptom management. Knowledge gaps may also be a contributing factor early in their training. A survey of physicians found that residents were more likely than attending physicians to list lack of formal education as a barrier to appropriate prescribing.²⁴

Similarities are seen in a study of antibiotic appropriateness, where residents demonstrated gaps in knowledge of treatment of asymptomatic bacteriuria that seemed to vary by specialty.²⁵ Interestingly, we found that patients admitted to general internal medicine were prescribed fewer PIPs. This service includes our Acute Care for Elders unit, which is staffed by trained geriatric nurses and other allied health professionals. Residents who rotated on internal medicine are also likely to have received informal teaching about medication safety in older adults. Educational interventions highlighting adverse effects of BSH and promoting nonpharmacologic solutions should be targeted at first-year residents. However, an interprofessional team approach to sleep disturbance in hospital, in combination with decision support for appropriate BSH use will achieve greater impact than education alone.

Several limitations of this study merit discussion. First, findings from a single academic center may lack generalizability. However, the demographics of our patient population and our rates of BSH use were similar to those reported in previous studies. Second, our study may be subject to observer bias, as the data collectors were not blinded. To minimize this, a strict template and clear appropriateness criteria were developed. Additionally, a second reviewer independently conducted data validation with 100% agreement among reviewers. Third, we studied prescribing patterns rather than medication administration and lacked data on filling of new BSH prescriptions in the postdischarge period. However, our primary goal is to determine risk of exposure to a BSH to minimize it. Fourth, although BSH are discouraged as “first choice for insomnia, anxiety or delirium,”⁴ they may be appropriate in limited situations where all nonpharmacologic strategies have failed and patient or staff safety is at risk. In our chart reviews, we were unable to determine whether all nonpharmacologic strategies were exhausted prior to prescription initiation. However, more than 20% of all PIP were routinely prescribed as part of an admission or postoperative order set, suggesting a reflexive rather than reflective approach to sedative use. Furthermore, the indications of anxiety and agitation were combined as they appear in the CPOE as a combination indication, thus leaving us unable to determine the true proportion for each indication. However, more than 87% of all PIPs were for insomnia, reflecting a clear opportunity to improve sleep management in hospital. Last, the lack of a power calculation may have resulted in the study being underpowered and thus affected the ability to detect a significant effect of covariates that have real differences on the likelihood of sedative prescriptions. For example, the low number of prescribing events by second-year residents and staff may have resulted in a type II error when comparing PIP rates with first-year residents.

We found that the majority of newly prescribed BSH among older adults in hospital were potentially inappropriate. They were most frequently prescribed by first-year residents overnight in response to insomnia. Our findings demonstrate BSH overuse remains prevalent and is associated with poor sleep in hospital. Future work will focus on implementing and evaluating safe sleep protocols and educational interventions aimed at first-year residents.

Acknowledgments

Elisabeth Pek had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Ciara Pendrith conducted and is responsible for the statistical analysis.

Disclosure

The authors report no financial conflicts of interest.

Older adults commonly experience insomnia and agitation during hospitalization. Unfortunately, the use of benzodiazepines and sedative hypnotics (BSH) to treat these conditions can be ineffective and expose patients to significant adverse effects.^1,2 Choosing Wisely^® is a campaign that promotes dialogue to reduce unnecessary medical tests, procedures, or treatments. This international campaign has highlighted BSHs as potentially harmful and has recommended against their use as first-line treatment of insomnia and agitation.^3-5 Examples of harm with benzodiazepine use include cognitive impairment, impaired postural stability, and an increased incidence of falls and hip fractures in both community and acute care settings.^6-8 In addition, prescriptions initiated in hospital appear to be associated with a higher risk of falls and unplanned readmission.^9,10 The newer nonbenzodiazepine sedative hypnotics, commonly referred to as “z-drugs”, were initially marketed as a safer alternative in older adults due to their more favorable pharmacokinetics. Evidence has emerged that they carry similar risks.^6,11,12 A study comparing benzodiazepines and zolpidem found relatively greater risk of fractures requiring hospitalization with the use of zolpidem compared to lorazepam.¹³

The use of benzodiazepines in the acute care setting has been evaluated in a number of studies and ranges from 20% to 45%.^14-16 Few studies focus on the initiation of these medications in BSH-naïve hospitalized patients; however, reports range from 18% to 29%.^17,18 Factors found to be associated with potentially inappropriate prescriptions (PIP) include Hispanic ethnicity, residing in an assisted care setting, and a greater number of BSH prescriptions prior to admission.^16,19 Additionally, Cumbler et al.¹⁵ found that the presence of dementia was associated with fewer prescriptions for sleep aids in hospital. To our knowledge, there are no published studies that have investigated prescriber factors associated with the use of BSH.

The purpose of our study was to determine the frequency of PIPs of BSH in our academic hospital. Additionally, we aimed to identify patient and prescriber factors that were associated with increased likelihood of prescriptions to help guide future quality improvement initiatives.

Study Design and Setting

This was a retrospective observational study conducted at Mount Sinai Hospital (MSH) in Toronto over a 4-month period from January 2013 to April 2013. The hospital is a 442-bed acute care academic health science center affiliated with the University of Toronto. The MSH electronic health record contains demographic data, medications and allergies, nursing documentation, and medical histories from prior encounters. It also includes computerized physician order entry (CPOE) and a detailed medication administration record. This system is integrated with an electronic pharmacy database used to monitor and dispense medications for each patient.

Patient and Medication Selection

We included inpatients over the age of 65 who were prescribed a BSH during the study period from the following services: general internal medicine, cardiology, general surgery, orthopedic surgery, and otolaryngology. To identify new exposure to BSHs, we excluded patients who were regularly prescribed a BSH prior to admission to hospital. The medications of interest included all benzodiazepines and the nonbenzodiazepine sedative hypnotic, zopiclone. Zopiclone is the most commonly used nonbenzodiazepine sedative hypnotic in Canada and the only 1 available on our hospital formulary. These were selected based on the strength of evidence to recommend against their use as first-line agents in older adults and in consultation with our geriatric medicine consultation team pharmacist.²⁰

Data Collection

The hospital administrative database provided patient demographic information, admission service, admitting diagnosis, length of stay, and the total number of patients discharged from the study units over the study period. We then searched the pharmacy electronic database for all benzodiazepines and zopiclone prescribed during the study period for patients who met the inclusion criteria. Manual review of paper and electronic health records for this cohort of patients was conducted to extract additional variables. We used a standardized form to record data elements. Dr. Pek collected all data elements. Dr. Remfry reviewed a random sample of patient records (10%) to ensure accuracy. The agreement between reviewers was 100%.

In compliance with hospital accreditation standards, a clinical pharmacist documents a best possible medication history (BPMH) on every inpatient on admission. We used the BPMH to identify and exclude patients who were prescribed a BSH prior to hospitalization. Because all medications were ordered through the CPOE system, as-needed medication prescriptions required the selection of a specified indication. Available options included ‘agitation/anxiety’ and necessitated combining these 2 indications into 1 category. Indications were primarily extracted through electronic order entry reviews. Paper charts were reviewed when further clarification was needed.

We identified ordering physicians’ training level and familiarity with the service from administrative records obtained from medical education offices, hospital records, and relevant call schedules. Fellows were defined as trainees with a minimum of 6 years of postgraduate training.

Our primary outcome of interest was the proportion of eligible patients age 65 and older who received a PIP for a BSH. Patient variables of interest included age, sex, comorbid conditions, and a pre-admission diagnosis of dementia. Comorbid conditions and age were used to calculate the Carlson Comorbidity Index for each patient.²¹ Prescription variables included the medication prescribed, time of first prescription (“overnight hours” refer to prescriptions ordered after 7:00 PM and before 7:00 AM), and whether the medication was ordered as part of an admission or postoperative order set. To determine whether patients were discharged home with a prescription for a BSH, we reviewed electronic discharge prescriptions of BSH-naïve patients who received a sedative in hospital. Only medical and cardiology inpatients receive electronic discharge prescriptions, and these were available for 189 patients in our cohort. Provider variables included training level, service, and familiarity with patients. We used the provider’s training program or department of appointment to define the ‘physician on-service’ variable. As an example, a resident registered in internal medicine is defined as ‘on-service’ when prescribing sedatives for a medical inpatient. In contrast, a psychiatry resident would be considered “off-service” if he prescribed a sedative for a surgical inpatient. The familiarity of a provider was categorized as ‘regular’ if they were responsible for a patient’s care on a day-to-day basis and ‘covering’ if they were only covering on call. Other variables included admitting service and hospital length of stay.

Appropriateness Criteria

Criteria for potentially inappropriate use were modified from the American and Canadian Geriatrics Societies’ Choosing Wisely recommendations,^4,5 and included insomnia and agitation. These recommendations are in line with other evidence based guidelines for safe prescribing in older adults.²⁰ For the purposes of our study, prescriptions for “agitation/anxiety”, “agitation”, or “insomnia/sleep” were considered potentially inappropriate. Appropriate indications included alcohol withdrawal, end-of-life symptom control, preprocedural sedation, and seizure.⁵ Patients who were already using a BSH prior to admission for any indication, including a psychiatric diagnosis, were excluded.

Statistical Analyses

We determined the proportion of patients with at least one PIP, as well as the proportion of all prescribing events that were potentially inappropriate. We used the Chi-square statistic and 2-sample t tests to compare the unadjusted associations between patient-level characteristics and receipt of at least 1 inappropriate prescription and prescribing event-level factors with inappropriate prescriptions. Given that first-year residents are more likely to be working overnight when most PIPs are prescribed, we performed a simple logistic regression of potentially inappropriate prescribing by level of training stratified by time of prescription. A multivariable random-intercept logistic regression model was used to assess the adjusted association between patient- and prescribing event-level characteristics with inappropriate prescribing, adjusting for clustering of prescribing events within patients. Characteristics of interest were identified a priori and those with significant bivariate associations with potentially inappropriate were selected for inclusion in the model. Additionally, we included time of prescription in our model to control for potential confounding. All statistical analyses were performed using SAS 9.4 (SAS Institute Inc, Cary, North Carolina). The MSH Research Ethics Board approved the study.

Description of Patients Prescribed a Benzodiazepine Sedative Hypnotic

There were 1540 patients over the age of 65 discharged during the 4-month study period. We excluded the 232 patients who had been prescribed a BSH prior to admission. Of the remaining eligible 1308 BSH-naïve patients, 251 (19.2%) were prescribed a new BSH in hospital and were included in the study. Of this cohort of 251 patients, 193 (76.9%) patients were prescribed a single BSH during their admission while 58 (23.1%) received 2 or more. Of all eligible patients, 208 (15.9%) were prescribed at least 1 PIP. Approximately half of the cohort was admitted to the general internal medicine service, and the most common reason for admission was cardiovascular disease (Table 1).

Description of Prescriptions of Benzodiazepine Sedative Hypnotic

We reviewed 328 prescriptions for BSH during the study period. The majority of these, 254 (77.4%) were potentially inappropriate (Table 2). The most common PIPs were zopiclone (167; 65.7%) and lorazepam (82; 32.3%). The PIPs were most frequently ordered on an as-needed basis (219; 86%), followed by one-time orders (30; 12%), and standing orders (5; 2%). The majority of PIPs (222; 87.4%) was prescribed for insomnia with a minority (32; 12.6%) prescribed for agitation and/or anxiety.

Most PIP were prescribed during overnight hours (159; 62.6%) and when an in-house pharmacist was unavailable (211; 83.1%). These variables were highly correlated with prescription of sleep aid, which was defined in our criteria as potentially inappropriate. Copies of discharge prescriptions were available for 189 patients. Of these 189 patients, 19 (10.1%) were sent home with a prescription for a new sedative.

Association Between Patient/Provider Variables and Prescriptions

Patient factors associated with fewer PIPs in our bivariate analyses included older age and dementia (Table 1). A greater proportion of nighttime prescriptions were PIPs; however, this finding was not statistically significant (P = 0.067). The majority of all prescriptions was prescribed by residents in their first year of training (64.9%; Table 2), and there was a significant difference in rates of PIP across level of training (P = 0.0007). When stratified by time of prescription, there was no significant difference by level of training for nighttime prescriptions. Among daytime prescriptions, second-year residents and staff (attending physicians and fellows) were less likely to prescribe a PIP than first-year residents (odds ratio [OR], 0.24; 95% confidence interval [CI], 0.09-0.66 and OR, 0.39; 95% CI, 0.14-1.13, respectively; Table 3); however, the association between staff and first-years only approached statistical significance (P = 0.08). Interestingly, 20.4% of all PIPs were ordered routinely as part of an admission or postoperative order set.

In our regression model, admission to a specialty or surgical service, compared to the general internal medicine service, was associated with a significantly higher likelihood of a PIP (OR, 6.61; 95% CI, 2.70-16.17; Table 4). Additionally, compared to cardiovascular admission diagnoses, neoplastic admitting diagnoses were associated with a higher likelihood of a PIP (OR, 4.43; 95% CI, 1.23-15.95). Time of prescription was a significant predictor in our multivariable regression model with nighttime prescriptions having increased odds of a PIP (OR, 4.48; 95% CI, 2.21-9.06,). When comparing prescribers at the extremes of training, attending physicians and fellows were much less likely to prescribe a PIP compared to first-year residents (OR, 0.23; 95% CI, 0.08-0.69; Table 4). However, there were no other significant differences across training levels after adjusting for patient and prescribing event characteristics.

We found that the majority of newly prescribed BSH in hospital was for the potentially inappropriate indications of insomnia and agitation/anxiety. Medications for insomnia were primarily initiated during overnight hours. Training level of prescribers and admitting service were found to be associated with appropriateness of prescriptions.

Our study showed that 15.9% of hospitalized older adults were newly prescribed a PIP during their admission. Of all new in hospital prescriptions, 77% were deemed potentially inappropriate. These numbers are similar to those reported by other centers; however, wide ranges exist.^16,19 This is likely the result of differences in appropriate use and inclusion criteria. Gillis et al.¹⁷ focused their investigation on sleep aids and showed that 26% of all admitted patients and 18% of BSH naïve patients received a prescription for insomnia. While this is similar to our findings, more than half of these patients were under the age of 65, and additional medications, such as trazodone, antihistamines, and antipsychotics were included.¹⁷ Other studies did not exclude patients who used a BSH regularly prior to admission. For example, 21% of veterans admitted to an acute care facility received a prescription for potentially inappropriate indications, but this included continuation of prior home medications.¹⁹ In contrast, we chose to focus on older adults in whom BSH pose a greater risk of harm. Exclusion of patients who regularly used a BSH prior to admission allowed us to better understand the circumstances surrounding the initiation of these medications in hospital. Furthermore, abrupt cessation of benzodiazepines can cause withdrawal and worsen confusion.²²

We found that 10% of patients newly prescribed a BSH in hospital were discharged with a prescription for a BSH. The accuracy of this is limited by the lack of availability of electronic discharge prescriptions on our surgical wards; however, it is likely an underrepresentation of the true effect given the high rates of PIPs on these wards. Our study highlights the concerning practice of continuing newly prescribed BSH following discharge from hospital.

Sleep disruption and poor quality sleep in hospital is a common issue that leads to significant use of BSH.¹⁵ Nonpharmacologic interventions in older adults can be effective in improving sleep quality and reducing the need for BSH; however, they can be time-consuming to implement.²³ With the exception of preventative strategies used on our Acute Care for Elders unit, formal nonpharmacologic interventions for sleep are not practiced in our hospital. We found that the majority of PIPs were prescribed as sleep aids in the overnight hours. This suggests that disruptions in sleep are leading patients and nursing staff to request pharmacologic treatments and highlights an area with significant room for improvement. Work is underway to implement and evaluate safe sleep protocols for older adults.

To our knowledge, we are the first to report an association between training level and PIP of BSH in older adults. The highest rates of PIPs were found among the first-year residents and, after controlling for patient and prescribing event characteristics, such as time of prescription, first-year residents were significantly more likely to prescribe a PIP. First-year residents are more likely to respond first to issues on the wards. There may be pressure on first-year trainees to prescribe sleep aids, as many patients and nurses may seek pharmacologic solutions for symptom management. Knowledge gaps may also be a contributing factor early in their training. A survey of physicians found that residents were more likely than attending physicians to list lack of formal education as a barrier to appropriate prescribing.²⁴

Similarities are seen in a study of antibiotic appropriateness, where residents demonstrated gaps in knowledge of treatment of asymptomatic bacteriuria that seemed to vary by specialty.²⁵ Interestingly, we found that patients admitted to general internal medicine were prescribed fewer PIPs. This service includes our Acute Care for Elders unit, which is staffed by trained geriatric nurses and other allied health professionals. Residents who rotated on internal medicine are also likely to have received informal teaching about medication safety in older adults. Educational interventions highlighting adverse effects of BSH and promoting nonpharmacologic solutions should be targeted at first-year residents. However, an interprofessional team approach to sleep disturbance in hospital, in combination with decision support for appropriate BSH use will achieve greater impact than education alone.

Several limitations of this study merit discussion. First, findings from a single academic center may lack generalizability. However, the demographics of our patient population and our rates of BSH use were similar to those reported in previous studies. Second, our study may be subject to observer bias, as the data collectors were not blinded. To minimize this, a strict template and clear appropriateness criteria were developed. Additionally, a second reviewer independently conducted data validation with 100% agreement among reviewers. Third, we studied prescribing patterns rather than medication administration and lacked data on filling of new BSH prescriptions in the postdischarge period. However, our primary goal is to determine risk of exposure to a BSH to minimize it. Fourth, although BSH are discouraged as “first choice for insomnia, anxiety or delirium,”⁴ they may be appropriate in limited situations where all nonpharmacologic strategies have failed and patient or staff safety is at risk. In our chart reviews, we were unable to determine whether all nonpharmacologic strategies were exhausted prior to prescription initiation. However, more than 20% of all PIP were routinely prescribed as part of an admission or postoperative order set, suggesting a reflexive rather than reflective approach to sedative use. Furthermore, the indications of anxiety and agitation were combined as they appear in the CPOE as a combination indication, thus leaving us unable to determine the true proportion for each indication. However, more than 87% of all PIPs were for insomnia, reflecting a clear opportunity to improve sleep management in hospital. Last, the lack of a power calculation may have resulted in the study being underpowered and thus affected the ability to detect a significant effect of covariates that have real differences on the likelihood of sedative prescriptions. For example, the low number of prescribing events by second-year residents and staff may have resulted in a type II error when comparing PIP rates with first-year residents.

We found that the majority of newly prescribed BSH among older adults in hospital were potentially inappropriate. They were most frequently prescribed by first-year residents overnight in response to insomnia. Our findings demonstrate BSH overuse remains prevalent and is associated with poor sleep in hospital. Future work will focus on implementing and evaluating safe sleep protocols and educational interventions aimed at first-year residents.

Acknowledgments

Elisabeth Pek had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Ciara Pendrith conducted and is responsible for the statistical analysis.

Disclosure

The authors report no financial conflicts of interest.

Hospital-acquired anemia (HAA) is defined as having a normal hemoglobin value upon admission but developing anemia during the course of hospitalization. The condition is common, with an incidence ranging from approximately 25% when defined by using the hemoglobin value prior to discharge to 74% when using the nadir hemoglobin value during hospitalization.^1-5 While there are many potential etiologies for HAA, given that iatrogenic blood loss from phlebotomy may lead to its development,^6,7 HAA has been postulated to be a hazard of hospitalization that is potentially preventable.⁸ However, it is unclear whether the development of HAA portends worse outcomes after hospital discharge.

The limited number of studies on the association between HAA and postdischarge outcomes has been restricted to patients hospitalized for acute myocardial infarction (AMI).^3,9,10 Among this subpopulation, HAA is independently associated with greater morbidity and mortality following hospital discharge.^3,9,10 In a more broadly representative population of hospitalized adults, Koch et al.² found that the development of HAA is associated with greater length of stay (LOS), hospital charges, and inpatient mortality. However, given that HAA was defined by the lowest hemoglobin level during hospitalization (and not necessarily the last value prior to discharge), it is unclear if the worse outcomes observed were the cause of the HAA, rather than its effect, since hospital LOS is a robust predictor for the development of HAA, as well as a major driver of hospital costs and a prognostic marker for inpatient mortality.^3,9 Furthermore, this study evaluated outcomes only during the index hospitalization, so it is unclear if patients who develop HAA have worse clinical outcomes after discharge.

Therefore, in this study, we used clinically granular electronic health record (EHR) data from a diverse cohort of consecutive medicine inpatients hospitalized for any reason at 1 of 6 hospitals to: 1) describe the epidemiology of HAA; 2) identify predictors of its development; and 3) examine its association with 30-day postdischarge adverse outcomes. We hypothesized that the development of HAA would be independently associated with 30-day readmission and mortality in a dose-dependent fashion, with increasing severity of HAA associated with worse outcomes.

Study Design, Population, and Data Sources

We conducted a retrospective observational cohort study using EHR data collected from November 1, 2009 to October 30, 2010 from 6 hospitals in the north Texas region. One site was a university-affiliated safety-net hospital; the remaining 5 community hospitals were a mix of teaching and nonteaching sites. All hospitals used the Epic EHR system (Epic Systems Corporation, Verona, Wisconsin). Details of this cohort have been published.^11,12This study included consecutive hospitalizations among adults age 18 years or older who were discharged from a medicine inpatient service with any diagnosis. We excluded hospitalizations by individuals who were anemic within the first 24 hours of admission (hematocrit less than 36% for women and less than 40% for men), were missing a hematocrit value within the first 24 hours of hospitalization or a repeat hematocrit value prior to discharge, had a hospitalization in the preceding 30 days (ie, index hospitalization was considered a readmission), died in the hospital, were transferred to another hospital, or left against medical advice. For individuals with multiple eligible hospitalizations during the study period, we included only the first hospitalization. We also excluded those discharged to hospice, given that this population of individuals may have intentionally desired less aggressive care.

Definition of Hospital-Acquired Anemia

HAA was defined as having a normal hematocrit value (36% or greater for women and 40% or greater for men) within the first 24 hours of admission and a hematocrit value at the time of hospital discharge lower than the World Health Organization’s sex-specific cut points.¹³ If there was more than 1 hematocrit value on the same day, we chose the lowest value. Based on prior studies, HAA was further categorized by severity as mild (hematocrit greater than 33% and less than 36% in women; and greater than 33% and less than 40% in men), moderate (hematocrit greater than 27% and 33% or less for all), or severe (hematocrit 27% or less for all).^2,14

Characteristics

We extracted information on sociodemographic characteristics, comorbidities, LOS, procedures, blood transfusions, and laboratory values from the EHR. Hospitalizations in the 12 months preceding the index hospitalization were ascertained from the EHR and from an all-payer regional hospitalization database that captures hospitalizations from 75 acute care hospitals within a 100-mile radius of Dallas-Fort Worth. International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) discharge diagnosis codes were categorized according to the Agency for Healthcare Research and Quality (AHRQ) Clinical Classifications Software (CCS).¹⁵ We defined a diagnosis for hemorrhage and coagulation, and hemorrhagic disorder as the presence of any ICD-9-CM code (primary or secondary) that mapped to the AHRQ CCS diagnoses 60 and 153, and 62, respectively. Procedures were categorized as minor diagnostic, minor therapeutic, major diagnostic, and major therapeutic using the AHRQ Healthcare Cost and Utilization Procedure Classes tool.¹⁶

Outcomes

The primary outcome was a composite of death or readmission within 30 days of hospital discharge. Hospital readmissions were ascertained at the index hospital and at any of 75 acute care hospitals in the region as described earlier. Death was ascertained from each of the hospitals’ EHR and administrative data and the Social Security Death Index. Individuals who had both outcomes (eg, a 30-day readmission and death) were considered to have only 1 occurrence of the composite primary outcome measure. Our secondary outcomes were death and readmission within 30 days of discharge, considered as separate outcomes.

Statistical Analysis

We used logistic regression models to evaluate predictors of HAA and to estimate the association of HAA on subsequent 30-day adverse outcomes after hospital discharge. All models accounted for clustering of patients by hospital. For the outcomes analyses, models were adjusted for potential confounders based on prior literature and our group’s expertise, which included age, sex, race/ethnicity, Charlson comorbidity index, prior hospitalizations, nonelective admission status, creatinine level on admission, blood urea nitrogen (BUN) to creatinine ratio of more than 20:1 on admission, LOS, receipt of a major diagnostic or therapeutic procedure during the index hospitalization, a discharge diagnosis for hemorrhage, and a discharge diagnosis for a coagulation or hemorrhagic disorder. For the mortality analyses, given the limited number of 30-day deaths after hospital discharge in our cohort, we collapsed moderate and severe HAA into a single category. In sensitivity analyses, we repeated the adjusted model, but excluded patients in our cohort who had received at least 1 blood transfusion during the index hospitalization (2.6%) given its potential for harm, and patients with a primary discharge diagnosis for AMI (3.1%).¹⁷

The functional forms of continuous variables were assessed using restricted cubic splines and locally weighted scatterplot smoothing techniques. All analyses were performed using STATA statistical software version 12.0 (StataCorp, College Station, Texas). The University of Texas Southwestern Medical Center institutional review board approved this study.

Of 53,995 consecutive medicine hospitalizations among adults age 18 years or older during our study period, 11,309 index hospitalizations were included in our study cohort (Supplemental Figure 1). The majority of patients excluded were because of having documented anemia within the first 24 hours of admission (n=24,950). With increasing severity of HAA, patients were older, more likely to be female, non-Hispanic white, electively admitted, have fewer comorbidities, less likely to be hospitalized in the past year, more likely to have had a major procedure, receive a blood transfusion, have a longer LOS, and have a primary or secondary discharge diagnosis for a hemorrhage or a coagulation or hemorrhagic disorder (Table 1).

Epidemiology of HAA

Among this cohort of patients without anemia on admission, the median hematocrit value on admission was 40.6 g/dL and on discharge was 38.9 g/dL. One-third of patients with normal hematocrit value at admission developed HAA, with 21.6% developing mild HAA, 10.1% developing moderate HAA, and 1.4% developing severe HAA. The median discharge hematocrit value was 36 g/dL (interquartile range [IQR]), 35-38 g/dL) for the group of patients who developed mild HAA, 31 g/dL (IQR, 30-32 g/dL) for moderate HAA, and 26 g/dL (IQR, 25-27 g/dL) for severe HAA (Supplemental Figure 2). Among the severe HAA group, 135 of the 159 patients (85%) had a major procedure (n=123, accounting for 219 unique major procedures), a diagnosis for hemorrhage (n=30), and/or a diagnosis for a coagulation or hemorrhagic disorder (n=23) during the index hospitalization. Of the 219 major procedures among patients with severe HAA, most were musculoskeletal (92 procedures), cardiovascular (61 procedures), or digestive system-related (41 procedures). The most common types of procedures were coronary artery bypass graft (36 procedures), hip replacement (25 procedures), knee replacement (17 procedures), and femur fracture reduction (15 procedures). The 10 most common principal discharge diagnoses of the index hospitalization by HAA group are shown in Supplemental Table 1. For the severe HAA group, the most common diagnosis was hip fracture (20.8%).

Predictors of HAA

Compared to no or mild HAA, female sex, elective admission status, serum creatinine on admission, BUN to creatinine ratio greater than 20 to 1, hospital LOS, and undergoing a major diagnostic or therapeutic procedure were predictors for the development of moderate or severe HAA (Table 2). The model explained 23% of the variance (McFadden’s pseudo R²).

Incidence of Postdischarge Outcomes by Severity of HAA

The severity of HAA was associated with a dose-dependent increase in the incidence of 30-day adverse outcomes, such that patients with increasing severity of HAA had greater 30-day composite, mortality, and readmission outcomes (P < 0.001; Figure). The 30-day postdischarge composite outcome was primarily driven by hospital readmissions given the low mortality rate in our cohort. Patients who did not develop HAA had an incidence of 9.7% for the composite outcome, whereas patients with severe HAA had an incidence of 16.4%. Among the 24 patients with severe HAA but who had not undergone a major procedure or had a discharge diagnosis for hemorrhage or for a coagulation or hemorrhagic disorder, only 3 (12.5%) had a composite postdischarge adverse outcome (2 readmissions and 1 death). The median time to readmission was similar between groups, but more patients with severe HAA had an early readmission within 7 days of hospital discharge than patients who did not develop HAA (6.9% vs. 2.9%, P = 0.001; Supplemental Table 2).

Association of HAA and Postdischarge Outcomes

In unadjusted analyses, compared to not developing HAA, mild, moderate, and severe HAA were associated with a 29%, 61%, and 81% increase in the odds for a composite outcome, respectively (Table 3). After adjustment for confounders, the effect size for HAA attenuated and was no longer statistically significant for mild and moderate HAA. However, severe HAA was significantly associated with a 39% increase in the odds for the composite outcome and a 41% increase in the odds for 30-day readmission (P = 0.008 and P = 0.02, respectively).

In sensitivity analyses, the exclusion of individuals who received at least 1 blood transfusion during the index hospitalization (n=298) and individuals who had a primary discharge diagnosis for AMI (n=353) did not substantively change the estimates of the association between severe HAA and postdischarge outcomes (Supplemental Tables 3 and 4). However, because of the fewer number of adverse events for each analysis, the confidence intervals were wider and the association of severe HAA and the composite outcome and readmission were no longer statistically significant in these subcohorts.

In this large and diverse sample of medical inpatients, we found that HAA occurs in one-third of adults with normal hematocrit value at admission, where 10.1% of the cohort developed moderately severe HAA and 1.4% developed severe HAA by the time of discharge. Length of stay and undergoing a major diagnostic or therapeutic procedure were the 2 strongest potentially modifiable predictors of developing moderate or severe HAA. Severe HAA was independently associated with a 39% increase in the odds of being readmitted or dying within 30 days after hospital discharge compared to not developing HAA. However, the associations between mild or moderate HAA with adverse outcomes were attenuated after adjusting for confounders and were no longer statistically significant.

To our knowledge, this is the first study on the postdischarge adverse outcomes of HAA among a diverse cohort of medical inpatients hospitalized for any reason. In a more restricted population, Salisbury et al.³ found that patients hospitalized for AMI who developed moderate to severe HAA (hemoglobin value at discharge of 11 g/dL or less) had greater 1-year mortality than those without HAA (8.4% vs. 2.6%, P < 0.001), and had an 82% increase in the hazard for mortality (95% confidence interval, hazard ratio 1.11-2.98). Others have similarly shown that HAA is common among patients hospitalized with AMI and is associated with greater mortality.^5,9,18 Our study extends upon this prior research by showing that severe HAA increases the risk for adverse outcomes for all adult inpatients, not only those hospitalized for AMI or among those receiving blood transfusions.

Despite the increased harm associated with severe HAA, it is unclear whether HAA is a preventable hazard of hospitalization, as suggested by others.^6,8 Most patients in our cohort who developed severe HAA underwent a major procedure, had a discharge diagnosis for hemorrhage, and/or had a discharge diagnosis for a coagulation or hemorrhagic disorder. Thus, blood loss due to phlebotomy, 1 of the more modifiable etiologies of HAA, was unlikely to have been the primary driver for most patients who developed severe HAA. Since it has been estimated to take 15 days of daily phlebotomy of 53 mL of whole blood in females of average body weight (and 20 days for average weight males) with no bone marrow synthesis for severe anemia to develop, it is even less likely that phlebotomy was the principal etiology given an 8-day median LOS among patients with severe HAA.^19,20 However, since the etiology of HAA can be multifactorial, limiting blood loss due to phlebotomy by using smaller volume tubes, blood conservation devices, or reducing unnecessary testing may mitigate the development of severe HAA.^21,22 Additionally, since more than three-quarters of patients who developed severe HAA underwent a major procedure, more care and attention to minimizing operative blood loss could lessen the severity of HAA and facilitate better recovery. If minimizing blood loss is not feasible, in the absence of symptoms related to anemia or ongoing blood loss, randomized controlled trials overwhelmingly support a restrictive transfusion strategy using a hemoglobin value threshold of 7 mg/dL, even in the postoperative setting.^23-25

The implications of mild to moderate HAA are less clear. The odds ratios for mild and moderate HAA, while not statistically significant, suggest a small increase in harm compared to not developing HAA. Furthermore, the upper boundary of the confidence intervals for mild and moderate HAA cannot exclude a possible 30% and 56% increase in the odds for the 30-day composite outcome, respectively. Thus, a better powered study, including more patients and extending the time interval for ascertaining postdischarge adverse events beyond 30 days, may reveal a harmful association. Lastly, our study assessed only the association of HAA with 30-day readmission and mortality. Examining the association between HAA and other patient-centered outcomes such as fatigue, functional impairment, and prolonged posthospitalization recovery time may uncover other important adverse effects of mild and moderate HAA, both of which occur far more frequently than severe HAA.

Our findings should be interpreted in the context of several limitations. First, although we included a diverse group of patients from a multihospital cohort, generalizability to other settings is uncertain. Second, as this was a retrospective study using EHR data, we had limited information to infer the precise mechanism of HAA for each patient. However, procedure codes and discharge diagnoses enabled us to assess which patients underwent a major procedure or had a hemorrhage or hemorrhagic disorder during the hospitalization. Third, given the relatively few number of patients with severe HAA in our cohort, we were unable to assess if the association of severe HAA differed by suspected etiology. Lastly, because we were unable to ascertain the timing of the hematocrit values within the first 24 hours of admission, we excluded both patients with preexisting anemia on admission and those who developed HAA within the first 24 hours of admission, which is not uncommon.²⁶ Thus, we were unable to assess the effect of acute on chronic anemia arising during hospitalization and HAA that develops within the first 24 hours, both of which may also be harmful.^18,27,28

In conclusion, severe HAA occurs in 1.4% of all medical hospitalizations and is associated with increased odds of death or readmission within 30 days. Since most patients with severe HAA had undergone a major procedure or had a discharge diagnosis of hemorrhage or a coagulation or hemorrhagic disorder, it is unclear if severe HAA is potentially preventable through preventing blood loss from phlebotomy or by reducing iatrogenic injury during procedures. Future research should assess the potential preventability of severe HAA, and examine other patient-centered outcomes potentially related to anemia, including fatigue, functional impairment, and trajectory of posthospital recovery.

Acknowledgments

The authors would like to acknowledge Ruben Amarasingham, MD, MBA, President and Chief Executive Officer of the Parkland Center for Clinical Innovation, and Ferdinand Velasco, MD, Chief Health Information Officer at Texas Health Resources, for their assistance in assembling the 6 hospital cohort used in this study. The authors would also like to thank Valy Fontil, MD, MAS, Assistant Professor of Medicine at the University of California San Francisco School of Medicine, and Elizabeth Rogers, MD, MAS, Assistant Professor of Internal Medicine and Pediatrics at the University of Minnesota Medical School, for their constructive feedback on an earlier version of this manuscript.

Disclosures

This work was supported by the Agency for Healthcare Research and Quality-funded UT Southwestern Center for Patient-Centered Outcomes Research (R24 HS022418-01); the Commonwealth Foundation (#20100323); the UT Southwestern KL2 Scholars Program supported by the National Institutes of Health (KL2 TR001103); the National Center for Advancing Translational Sciences at the National Institute of Health (U54 RFA-TR-12-006); and the National Institute on Aging (K23AG052603). The study sponsors had no role in design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript. The authors have no financial conflicts of interest to disclose.

Hospital-acquired anemia (HAA) is defined as having a normal hemoglobin value upon admission but developing anemia during the course of hospitalization. The condition is common, with an incidence ranging from approximately 25% when defined by using the hemoglobin value prior to discharge to 74% when using the nadir hemoglobin value during hospitalization.^1-5 While there are many potential etiologies for HAA, given that iatrogenic blood loss from phlebotomy may lead to its development,^6,7 HAA has been postulated to be a hazard of hospitalization that is potentially preventable.⁸ However, it is unclear whether the development of HAA portends worse outcomes after hospital discharge.

The limited number of studies on the association between HAA and postdischarge outcomes has been restricted to patients hospitalized for acute myocardial infarction (AMI).^3,9,10 Among this subpopulation, HAA is independently associated with greater morbidity and mortality following hospital discharge.^3,9,10 In a more broadly representative population of hospitalized adults, Koch et al.² found that the development of HAA is associated with greater length of stay (LOS), hospital charges, and inpatient mortality. However, given that HAA was defined by the lowest hemoglobin level during hospitalization (and not necessarily the last value prior to discharge), it is unclear if the worse outcomes observed were the cause of the HAA, rather than its effect, since hospital LOS is a robust predictor for the development of HAA, as well as a major driver of hospital costs and a prognostic marker for inpatient mortality.^3,9 Furthermore, this study evaluated outcomes only during the index hospitalization, so it is unclear if patients who develop HAA have worse clinical outcomes after discharge.

Therefore, in this study, we used clinically granular electronic health record (EHR) data from a diverse cohort of consecutive medicine inpatients hospitalized for any reason at 1 of 6 hospitals to: 1) describe the epidemiology of HAA; 2) identify predictors of its development; and 3) examine its association with 30-day postdischarge adverse outcomes. We hypothesized that the development of HAA would be independently associated with 30-day readmission and mortality in a dose-dependent fashion, with increasing severity of HAA associated with worse outcomes.

Study Design, Population, and Data Sources

We conducted a retrospective observational cohort study using EHR data collected from November 1, 2009 to October 30, 2010 from 6 hospitals in the north Texas region. One site was a university-affiliated safety-net hospital; the remaining 5 community hospitals were a mix of teaching and nonteaching sites. All hospitals used the Epic EHR system (Epic Systems Corporation, Verona, Wisconsin). Details of this cohort have been published.^11,12This study included consecutive hospitalizations among adults age 18 years or older who were discharged from a medicine inpatient service with any diagnosis. We excluded hospitalizations by individuals who were anemic within the first 24 hours of admission (hematocrit less than 36% for women and less than 40% for men), were missing a hematocrit value within the first 24 hours of hospitalization or a repeat hematocrit value prior to discharge, had a hospitalization in the preceding 30 days (ie, index hospitalization was considered a readmission), died in the hospital, were transferred to another hospital, or left against medical advice. For individuals with multiple eligible hospitalizations during the study period, we included only the first hospitalization. We also excluded those discharged to hospice, given that this population of individuals may have intentionally desired less aggressive care.

Definition of Hospital-Acquired Anemia

HAA was defined as having a normal hematocrit value (36% or greater for women and 40% or greater for men) within the first 24 hours of admission and a hematocrit value at the time of hospital discharge lower than the World Health Organization’s sex-specific cut points.¹³ If there was more than 1 hematocrit value on the same day, we chose the lowest value. Based on prior studies, HAA was further categorized by severity as mild (hematocrit greater than 33% and less than 36% in women; and greater than 33% and less than 40% in men), moderate (hematocrit greater than 27% and 33% or less for all), or severe (hematocrit 27% or less for all).^2,14

Characteristics

We extracted information on sociodemographic characteristics, comorbidities, LOS, procedures, blood transfusions, and laboratory values from the EHR. Hospitalizations in the 12 months preceding the index hospitalization were ascertained from the EHR and from an all-payer regional hospitalization database that captures hospitalizations from 75 acute care hospitals within a 100-mile radius of Dallas-Fort Worth. International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) discharge diagnosis codes were categorized according to the Agency for Healthcare Research and Quality (AHRQ) Clinical Classifications Software (CCS).¹⁵ We defined a diagnosis for hemorrhage and coagulation, and hemorrhagic disorder as the presence of any ICD-9-CM code (primary or secondary) that mapped to the AHRQ CCS diagnoses 60 and 153, and 62, respectively. Procedures were categorized as minor diagnostic, minor therapeutic, major diagnostic, and major therapeutic using the AHRQ Healthcare Cost and Utilization Procedure Classes tool.¹⁶

Outcomes

The primary outcome was a composite of death or readmission within 30 days of hospital discharge. Hospital readmissions were ascertained at the index hospital and at any of 75 acute care hospitals in the region as described earlier. Death was ascertained from each of the hospitals’ EHR and administrative data and the Social Security Death Index. Individuals who had both outcomes (eg, a 30-day readmission and death) were considered to have only 1 occurrence of the composite primary outcome measure. Our secondary outcomes were death and readmission within 30 days of discharge, considered as separate outcomes.

Statistical Analysis

We used logistic regression models to evaluate predictors of HAA and to estimate the association of HAA on subsequent 30-day adverse outcomes after hospital discharge. All models accounted for clustering of patients by hospital. For the outcomes analyses, models were adjusted for potential confounders based on prior literature and our group’s expertise, which included age, sex, race/ethnicity, Charlson comorbidity index, prior hospitalizations, nonelective admission status, creatinine level on admission, blood urea nitrogen (BUN) to creatinine ratio of more than 20:1 on admission, LOS, receipt of a major diagnostic or therapeutic procedure during the index hospitalization, a discharge diagnosis for hemorrhage, and a discharge diagnosis for a coagulation or hemorrhagic disorder. For the mortality analyses, given the limited number of 30-day deaths after hospital discharge in our cohort, we collapsed moderate and severe HAA into a single category. In sensitivity analyses, we repeated the adjusted model, but excluded patients in our cohort who had received at least 1 blood transfusion during the index hospitalization (2.6%) given its potential for harm, and patients with a primary discharge diagnosis for AMI (3.1%).¹⁷

The functional forms of continuous variables were assessed using restricted cubic splines and locally weighted scatterplot smoothing techniques. All analyses were performed using STATA statistical software version 12.0 (StataCorp, College Station, Texas). The University of Texas Southwestern Medical Center institutional review board approved this study.

Of 53,995 consecutive medicine hospitalizations among adults age 18 years or older during our study period, 11,309 index hospitalizations were included in our study cohort (Supplemental Figure 1). The majority of patients excluded were because of having documented anemia within the first 24 hours of admission (n=24,950). With increasing severity of HAA, patients were older, more likely to be female, non-Hispanic white, electively admitted, have fewer comorbidities, less likely to be hospitalized in the past year, more likely to have had a major procedure, receive a blood transfusion, have a longer LOS, and have a primary or secondary discharge diagnosis for a hemorrhage or a coagulation or hemorrhagic disorder (Table 1).

Epidemiology of HAA

Among this cohort of patients without anemia on admission, the median hematocrit value on admission was 40.6 g/dL and on discharge was 38.9 g/dL. One-third of patients with normal hematocrit value at admission developed HAA, with 21.6% developing mild HAA, 10.1% developing moderate HAA, and 1.4% developing severe HAA. The median discharge hematocrit value was 36 g/dL (interquartile range [IQR]), 35-38 g/dL) for the group of patients who developed mild HAA, 31 g/dL (IQR, 30-32 g/dL) for moderate HAA, and 26 g/dL (IQR, 25-27 g/dL) for severe HAA (Supplemental Figure 2). Among the severe HAA group, 135 of the 159 patients (85%) had a major procedure (n=123, accounting for 219 unique major procedures), a diagnosis for hemorrhage (n=30), and/or a diagnosis for a coagulation or hemorrhagic disorder (n=23) during the index hospitalization. Of the 219 major procedures among patients with severe HAA, most were musculoskeletal (92 procedures), cardiovascular (61 procedures), or digestive system-related (41 procedures). The most common types of procedures were coronary artery bypass graft (36 procedures), hip replacement (25 procedures), knee replacement (17 procedures), and femur fracture reduction (15 procedures). The 10 most common principal discharge diagnoses of the index hospitalization by HAA group are shown in Supplemental Table 1. For the severe HAA group, the most common diagnosis was hip fracture (20.8%).

Predictors of HAA

Compared to no or mild HAA, female sex, elective admission status, serum creatinine on admission, BUN to creatinine ratio greater than 20 to 1, hospital LOS, and undergoing a major diagnostic or therapeutic procedure were predictors for the development of moderate or severe HAA (Table 2). The model explained 23% of the variance (McFadden’s pseudo R²).

Incidence of Postdischarge Outcomes by Severity of HAA

The severity of HAA was associated with a dose-dependent increase in the incidence of 30-day adverse outcomes, such that patients with increasing severity of HAA had greater 30-day composite, mortality, and readmission outcomes (P < 0.001; Figure). The 30-day postdischarge composite outcome was primarily driven by hospital readmissions given the low mortality rate in our cohort. Patients who did not develop HAA had an incidence of 9.7% for the composite outcome, whereas patients with severe HAA had an incidence of 16.4%. Among the 24 patients with severe HAA but who had not undergone a major procedure or had a discharge diagnosis for hemorrhage or for a coagulation or hemorrhagic disorder, only 3 (12.5%) had a composite postdischarge adverse outcome (2 readmissions and 1 death). The median time to readmission was similar between groups, but more patients with severe HAA had an early readmission within 7 days of hospital discharge than patients who did not develop HAA (6.9% vs. 2.9%, P = 0.001; Supplemental Table 2).

Association of HAA and Postdischarge Outcomes

In unadjusted analyses, compared to not developing HAA, mild, moderate, and severe HAA were associated with a 29%, 61%, and 81% increase in the odds for a composite outcome, respectively (Table 3). After adjustment for confounders, the effect size for HAA attenuated and was no longer statistically significant for mild and moderate HAA. However, severe HAA was significantly associated with a 39% increase in the odds for the composite outcome and a 41% increase in the odds for 30-day readmission (P = 0.008 and P = 0.02, respectively).

In sensitivity analyses, the exclusion of individuals who received at least 1 blood transfusion during the index hospitalization (n=298) and individuals who had a primary discharge diagnosis for AMI (n=353) did not substantively change the estimates of the association between severe HAA and postdischarge outcomes (Supplemental Tables 3 and 4). However, because of the fewer number of adverse events for each analysis, the confidence intervals were wider and the association of severe HAA and the composite outcome and readmission were no longer statistically significant in these subcohorts.

In this large and diverse sample of medical inpatients, we found that HAA occurs in one-third of adults with normal hematocrit value at admission, where 10.1% of the cohort developed moderately severe HAA and 1.4% developed severe HAA by the time of discharge. Length of stay and undergoing a major diagnostic or therapeutic procedure were the 2 strongest potentially modifiable predictors of developing moderate or severe HAA. Severe HAA was independently associated with a 39% increase in the odds of being readmitted or dying within 30 days after hospital discharge compared to not developing HAA. However, the associations between mild or moderate HAA with adverse outcomes were attenuated after adjusting for confounders and were no longer statistically significant.

To our knowledge, this is the first study on the postdischarge adverse outcomes of HAA among a diverse cohort of medical inpatients hospitalized for any reason. In a more restricted population, Salisbury et al.³ found that patients hospitalized for AMI who developed moderate to severe HAA (hemoglobin value at discharge of 11 g/dL or less) had greater 1-year mortality than those without HAA (8.4% vs. 2.6%, P < 0.001), and had an 82% increase in the hazard for mortality (95% confidence interval, hazard ratio 1.11-2.98). Others have similarly shown that HAA is common among patients hospitalized with AMI and is associated with greater mortality.^5,9,18 Our study extends upon this prior research by showing that severe HAA increases the risk for adverse outcomes for all adult inpatients, not only those hospitalized for AMI or among those receiving blood transfusions.

Despite the increased harm associated with severe HAA, it is unclear whether HAA is a preventable hazard of hospitalization, as suggested by others.^6,8 Most patients in our cohort who developed severe HAA underwent a major procedure, had a discharge diagnosis for hemorrhage, and/or had a discharge diagnosis for a coagulation or hemorrhagic disorder. Thus, blood loss due to phlebotomy, 1 of the more modifiable etiologies of HAA, was unlikely to have been the primary driver for most patients who developed severe HAA. Since it has been estimated to take 15 days of daily phlebotomy of 53 mL of whole blood in females of average body weight (and 20 days for average weight males) with no bone marrow synthesis for severe anemia to develop, it is even less likely that phlebotomy was the principal etiology given an 8-day median LOS among patients with severe HAA.^19,20 However, since the etiology of HAA can be multifactorial, limiting blood loss due to phlebotomy by using smaller volume tubes, blood conservation devices, or reducing unnecessary testing may mitigate the development of severe HAA.^21,22 Additionally, since more than three-quarters of patients who developed severe HAA underwent a major procedure, more care and attention to minimizing operative blood loss could lessen the severity of HAA and facilitate better recovery. If minimizing blood loss is not feasible, in the absence of symptoms related to anemia or ongoing blood loss, randomized controlled trials overwhelmingly support a restrictive transfusion strategy using a hemoglobin value threshold of 7 mg/dL, even in the postoperative setting.^23-25

The implications of mild to moderate HAA are less clear. The odds ratios for mild and moderate HAA, while not statistically significant, suggest a small increase in harm compared to not developing HAA. Furthermore, the upper boundary of the confidence intervals for mild and moderate HAA cannot exclude a possible 30% and 56% increase in the odds for the 30-day composite outcome, respectively. Thus, a better powered study, including more patients and extending the time interval for ascertaining postdischarge adverse events beyond 30 days, may reveal a harmful association. Lastly, our study assessed only the association of HAA with 30-day readmission and mortality. Examining the association between HAA and other patient-centered outcomes such as fatigue, functional impairment, and prolonged posthospitalization recovery time may uncover other important adverse effects of mild and moderate HAA, both of which occur far more frequently than severe HAA.

Our findings should be interpreted in the context of several limitations. First, although we included a diverse group of patients from a multihospital cohort, generalizability to other settings is uncertain. Second, as this was a retrospective study using EHR data, we had limited information to infer the precise mechanism of HAA for each patient. However, procedure codes and discharge diagnoses enabled us to assess which patients underwent a major procedure or had a hemorrhage or hemorrhagic disorder during the hospitalization. Third, given the relatively few number of patients with severe HAA in our cohort, we were unable to assess if the association of severe HAA differed by suspected etiology. Lastly, because we were unable to ascertain the timing of the hematocrit values within the first 24 hours of admission, we excluded both patients with preexisting anemia on admission and those who developed HAA within the first 24 hours of admission, which is not uncommon.²⁶ Thus, we were unable to assess the effect of acute on chronic anemia arising during hospitalization and HAA that develops within the first 24 hours, both of which may also be harmful.^18,27,28

In conclusion, severe HAA occurs in 1.4% of all medical hospitalizations and is associated with increased odds of death or readmission within 30 days. Since most patients with severe HAA had undergone a major procedure or had a discharge diagnosis of hemorrhage or a coagulation or hemorrhagic disorder, it is unclear if severe HAA is potentially preventable through preventing blood loss from phlebotomy or by reducing iatrogenic injury during procedures. Future research should assess the potential preventability of severe HAA, and examine other patient-centered outcomes potentially related to anemia, including fatigue, functional impairment, and trajectory of posthospital recovery.

Acknowledgments

The authors would like to acknowledge Ruben Amarasingham, MD, MBA, President and Chief Executive Officer of the Parkland Center for Clinical Innovation, and Ferdinand Velasco, MD, Chief Health Information Officer at Texas Health Resources, for their assistance in assembling the 6 hospital cohort used in this study. The authors would also like to thank Valy Fontil, MD, MAS, Assistant Professor of Medicine at the University of California San Francisco School of Medicine, and Elizabeth Rogers, MD, MAS, Assistant Professor of Internal Medicine and Pediatrics at the University of Minnesota Medical School, for their constructive feedback on an earlier version of this manuscript.

Disclosures

This work was supported by the Agency for Healthcare Research and Quality-funded UT Southwestern Center for Patient-Centered Outcomes Research (R24 HS022418-01); the Commonwealth Foundation (#20100323); the UT Southwestern KL2 Scholars Program supported by the National Institutes of Health (KL2 TR001103); the National Center for Advancing Translational Sciences at the National Institute of Health (U54 RFA-TR-12-006); and the National Institute on Aging (K23AG052603). The study sponsors had no role in design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript. The authors have no financial conflicts of interest to disclose.

Hospital-acquired anemia (HAA) is defined as having a normal hemoglobin value upon admission but developing anemia during the course of hospitalization. The condition is common, with an incidence ranging from approximately 25% when defined by using the hemoglobin value prior to discharge to 74% when using the nadir hemoglobin value during hospitalization.^1-5 While there are many potential etiologies for HAA, given that iatrogenic blood loss from phlebotomy may lead to its development,^6,7 HAA has been postulated to be a hazard of hospitalization that is potentially preventable.⁸ However, it is unclear whether the development of HAA portends worse outcomes after hospital discharge.

The limited number of studies on the association between HAA and postdischarge outcomes has been restricted to patients hospitalized for acute myocardial infarction (AMI).^3,9,10 Among this subpopulation, HAA is independently associated with greater morbidity and mortality following hospital discharge.^3,9,10 In a more broadly representative population of hospitalized adults, Koch et al.² found that the development of HAA is associated with greater length of stay (LOS), hospital charges, and inpatient mortality. However, given that HAA was defined by the lowest hemoglobin level during hospitalization (and not necessarily the last value prior to discharge), it is unclear if the worse outcomes observed were the cause of the HAA, rather than its effect, since hospital LOS is a robust predictor for the development of HAA, as well as a major driver of hospital costs and a prognostic marker for inpatient mortality.^3,9 Furthermore, this study evaluated outcomes only during the index hospitalization, so it is unclear if patients who develop HAA have worse clinical outcomes after discharge.

Therefore, in this study, we used clinically granular electronic health record (EHR) data from a diverse cohort of consecutive medicine inpatients hospitalized for any reason at 1 of 6 hospitals to: 1) describe the epidemiology of HAA; 2) identify predictors of its development; and 3) examine its association with 30-day postdischarge adverse outcomes. We hypothesized that the development of HAA would be independently associated with 30-day readmission and mortality in a dose-dependent fashion, with increasing severity of HAA associated with worse outcomes.

Study Design, Population, and Data Sources

We conducted a retrospective observational cohort study using EHR data collected from November 1, 2009 to October 30, 2010 from 6 hospitals in the north Texas region. One site was a university-affiliated safety-net hospital; the remaining 5 community hospitals were a mix of teaching and nonteaching sites. All hospitals used the Epic EHR system (Epic Systems Corporation, Verona, Wisconsin). Details of this cohort have been published.^11,12This study included consecutive hospitalizations among adults age 18 years or older who were discharged from a medicine inpatient service with any diagnosis. We excluded hospitalizations by individuals who were anemic within the first 24 hours of admission (hematocrit less than 36% for women and less than 40% for men), were missing a hematocrit value within the first 24 hours of hospitalization or a repeat hematocrit value prior to discharge, had a hospitalization in the preceding 30 days (ie, index hospitalization was considered a readmission), died in the hospital, were transferred to another hospital, or left against medical advice. For individuals with multiple eligible hospitalizations during the study period, we included only the first hospitalization. We also excluded those discharged to hospice, given that this population of individuals may have intentionally desired less aggressive care.

Definition of Hospital-Acquired Anemia

HAA was defined as having a normal hematocrit value (36% or greater for women and 40% or greater for men) within the first 24 hours of admission and a hematocrit value at the time of hospital discharge lower than the World Health Organization’s sex-specific cut points.¹³ If there was more than 1 hematocrit value on the same day, we chose the lowest value. Based on prior studies, HAA was further categorized by severity as mild (hematocrit greater than 33% and less than 36% in women; and greater than 33% and less than 40% in men), moderate (hematocrit greater than 27% and 33% or less for all), or severe (hematocrit 27% or less for all).^2,14

Characteristics

We extracted information on sociodemographic characteristics, comorbidities, LOS, procedures, blood transfusions, and laboratory values from the EHR. Hospitalizations in the 12 months preceding the index hospitalization were ascertained from the EHR and from an all-payer regional hospitalization database that captures hospitalizations from 75 acute care hospitals within a 100-mile radius of Dallas-Fort Worth. International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) discharge diagnosis codes were categorized according to the Agency for Healthcare Research and Quality (AHRQ) Clinical Classifications Software (CCS).¹⁵ We defined a diagnosis for hemorrhage and coagulation, and hemorrhagic disorder as the presence of any ICD-9-CM code (primary or secondary) that mapped to the AHRQ CCS diagnoses 60 and 153, and 62, respectively. Procedures were categorized as minor diagnostic, minor therapeutic, major diagnostic, and major therapeutic using the AHRQ Healthcare Cost and Utilization Procedure Classes tool.¹⁶

Outcomes

The primary outcome was a composite of death or readmission within 30 days of hospital discharge. Hospital readmissions were ascertained at the index hospital and at any of 75 acute care hospitals in the region as described earlier. Death was ascertained from each of the hospitals’ EHR and administrative data and the Social Security Death Index. Individuals who had both outcomes (eg, a 30-day readmission and death) were considered to have only 1 occurrence of the composite primary outcome measure. Our secondary outcomes were death and readmission within 30 days of discharge, considered as separate outcomes.

Statistical Analysis

We used logistic regression models to evaluate predictors of HAA and to estimate the association of HAA on subsequent 30-day adverse outcomes after hospital discharge. All models accounted for clustering of patients by hospital. For the outcomes analyses, models were adjusted for potential confounders based on prior literature and our group’s expertise, which included age, sex, race/ethnicity, Charlson comorbidity index, prior hospitalizations, nonelective admission status, creatinine level on admission, blood urea nitrogen (BUN) to creatinine ratio of more than 20:1 on admission, LOS, receipt of a major diagnostic or therapeutic procedure during the index hospitalization, a discharge diagnosis for hemorrhage, and a discharge diagnosis for a coagulation or hemorrhagic disorder. For the mortality analyses, given the limited number of 30-day deaths after hospital discharge in our cohort, we collapsed moderate and severe HAA into a single category. In sensitivity analyses, we repeated the adjusted model, but excluded patients in our cohort who had received at least 1 blood transfusion during the index hospitalization (2.6%) given its potential for harm, and patients with a primary discharge diagnosis for AMI (3.1%).¹⁷

The functional forms of continuous variables were assessed using restricted cubic splines and locally weighted scatterplot smoothing techniques. All analyses were performed using STATA statistical software version 12.0 (StataCorp, College Station, Texas). The University of Texas Southwestern Medical Center institutional review board approved this study.

Of 53,995 consecutive medicine hospitalizations among adults age 18 years or older during our study period, 11,309 index hospitalizations were included in our study cohort (Supplemental Figure 1). The majority of patients excluded were because of having documented anemia within the first 24 hours of admission (n=24,950). With increasing severity of HAA, patients were older, more likely to be female, non-Hispanic white, electively admitted, have fewer comorbidities, less likely to be hospitalized in the past year, more likely to have had a major procedure, receive a blood transfusion, have a longer LOS, and have a primary or secondary discharge diagnosis for a hemorrhage or a coagulation or hemorrhagic disorder (Table 1).

Epidemiology of HAA

Among this cohort of patients without anemia on admission, the median hematocrit value on admission was 40.6 g/dL and on discharge was 38.9 g/dL. One-third of patients with normal hematocrit value at admission developed HAA, with 21.6% developing mild HAA, 10.1% developing moderate HAA, and 1.4% developing severe HAA. The median discharge hematocrit value was 36 g/dL (interquartile range [IQR]), 35-38 g/dL) for the group of patients who developed mild HAA, 31 g/dL (IQR, 30-32 g/dL) for moderate HAA, and 26 g/dL (IQR, 25-27 g/dL) for severe HAA (Supplemental Figure 2). Among the severe HAA group, 135 of the 159 patients (85%) had a major procedure (n=123, accounting for 219 unique major procedures), a diagnosis for hemorrhage (n=30), and/or a diagnosis for a coagulation or hemorrhagic disorder (n=23) during the index hospitalization. Of the 219 major procedures among patients with severe HAA, most were musculoskeletal (92 procedures), cardiovascular (61 procedures), or digestive system-related (41 procedures). The most common types of procedures were coronary artery bypass graft (36 procedures), hip replacement (25 procedures), knee replacement (17 procedures), and femur fracture reduction (15 procedures). The 10 most common principal discharge diagnoses of the index hospitalization by HAA group are shown in Supplemental Table 1. For the severe HAA group, the most common diagnosis was hip fracture (20.8%).

Predictors of HAA

Compared to no or mild HAA, female sex, elective admission status, serum creatinine on admission, BUN to creatinine ratio greater than 20 to 1, hospital LOS, and undergoing a major diagnostic or therapeutic procedure were predictors for the development of moderate or severe HAA (Table 2). The model explained 23% of the variance (McFadden’s pseudo R²).

Incidence of Postdischarge Outcomes by Severity of HAA

The severity of HAA was associated with a dose-dependent increase in the incidence of 30-day adverse outcomes, such that patients with increasing severity of HAA had greater 30-day composite, mortality, and readmission outcomes (P < 0.001; Figure). The 30-day postdischarge composite outcome was primarily driven by hospital readmissions given the low mortality rate in our cohort. Patients who did not develop HAA had an incidence of 9.7% for the composite outcome, whereas patients with severe HAA had an incidence of 16.4%. Among the 24 patients with severe HAA but who had not undergone a major procedure or had a discharge diagnosis for hemorrhage or for a coagulation or hemorrhagic disorder, only 3 (12.5%) had a composite postdischarge adverse outcome (2 readmissions and 1 death). The median time to readmission was similar between groups, but more patients with severe HAA had an early readmission within 7 days of hospital discharge than patients who did not develop HAA (6.9% vs. 2.9%, P = 0.001; Supplemental Table 2).

Association of HAA and Postdischarge Outcomes

In unadjusted analyses, compared to not developing HAA, mild, moderate, and severe HAA were associated with a 29%, 61%, and 81% increase in the odds for a composite outcome, respectively (Table 3). After adjustment for confounders, the effect size for HAA attenuated and was no longer statistically significant for mild and moderate HAA. However, severe HAA was significantly associated with a 39% increase in the odds for the composite outcome and a 41% increase in the odds for 30-day readmission (P = 0.008 and P = 0.02, respectively).

In sensitivity analyses, the exclusion of individuals who received at least 1 blood transfusion during the index hospitalization (n=298) and individuals who had a primary discharge diagnosis for AMI (n=353) did not substantively change the estimates of the association between severe HAA and postdischarge outcomes (Supplemental Tables 3 and 4). However, because of the fewer number of adverse events for each analysis, the confidence intervals were wider and the association of severe HAA and the composite outcome and readmission were no longer statistically significant in these subcohorts.

In this large and diverse sample of medical inpatients, we found that HAA occurs in one-third of adults with normal hematocrit value at admission, where 10.1% of the cohort developed moderately severe HAA and 1.4% developed severe HAA by the time of discharge. Length of stay and undergoing a major diagnostic or therapeutic procedure were the 2 strongest potentially modifiable predictors of developing moderate or severe HAA. Severe HAA was independently associated with a 39% increase in the odds of being readmitted or dying within 30 days after hospital discharge compared to not developing HAA. However, the associations between mild or moderate HAA with adverse outcomes were attenuated after adjusting for confounders and were no longer statistically significant.

To our knowledge, this is the first study on the postdischarge adverse outcomes of HAA among a diverse cohort of medical inpatients hospitalized for any reason. In a more restricted population, Salisbury et al.³ found that patients hospitalized for AMI who developed moderate to severe HAA (hemoglobin value at discharge of 11 g/dL or less) had greater 1-year mortality than those without HAA (8.4% vs. 2.6%, P < 0.001), and had an 82% increase in the hazard for mortality (95% confidence interval, hazard ratio 1.11-2.98). Others have similarly shown that HAA is common among patients hospitalized with AMI and is associated with greater mortality.^5,9,18 Our study extends upon this prior research by showing that severe HAA increases the risk for adverse outcomes for all adult inpatients, not only those hospitalized for AMI or among those receiving blood transfusions.

Despite the increased harm associated with severe HAA, it is unclear whether HAA is a preventable hazard of hospitalization, as suggested by others.^6,8 Most patients in our cohort who developed severe HAA underwent a major procedure, had a discharge diagnosis for hemorrhage, and/or had a discharge diagnosis for a coagulation or hemorrhagic disorder. Thus, blood loss due to phlebotomy, 1 of the more modifiable etiologies of HAA, was unlikely to have been the primary driver for most patients who developed severe HAA. Since it has been estimated to take 15 days of daily phlebotomy of 53 mL of whole blood in females of average body weight (and 20 days for average weight males) with no bone marrow synthesis for severe anemia to develop, it is even less likely that phlebotomy was the principal etiology given an 8-day median LOS among patients with severe HAA.^19,20 However, since the etiology of HAA can be multifactorial, limiting blood loss due to phlebotomy by using smaller volume tubes, blood conservation devices, or reducing unnecessary testing may mitigate the development of severe HAA.^21,22 Additionally, since more than three-quarters of patients who developed severe HAA underwent a major procedure, more care and attention to minimizing operative blood loss could lessen the severity of HAA and facilitate better recovery. If minimizing blood loss is not feasible, in the absence of symptoms related to anemia or ongoing blood loss, randomized controlled trials overwhelmingly support a restrictive transfusion strategy using a hemoglobin value threshold of 7 mg/dL, even in the postoperative setting.^23-25

The implications of mild to moderate HAA are less clear. The odds ratios for mild and moderate HAA, while not statistically significant, suggest a small increase in harm compared to not developing HAA. Furthermore, the upper boundary of the confidence intervals for mild and moderate HAA cannot exclude a possible 30% and 56% increase in the odds for the 30-day composite outcome, respectively. Thus, a better powered study, including more patients and extending the time interval for ascertaining postdischarge adverse events beyond 30 days, may reveal a harmful association. Lastly, our study assessed only the association of HAA with 30-day readmission and mortality. Examining the association between HAA and other patient-centered outcomes such as fatigue, functional impairment, and prolonged posthospitalization recovery time may uncover other important adverse effects of mild and moderate HAA, both of which occur far more frequently than severe HAA.

Our findings should be interpreted in the context of several limitations. First, although we included a diverse group of patients from a multihospital cohort, generalizability to other settings is uncertain. Second, as this was a retrospective study using EHR data, we had limited information to infer the precise mechanism of HAA for each patient. However, procedure codes and discharge diagnoses enabled us to assess which patients underwent a major procedure or had a hemorrhage or hemorrhagic disorder during the hospitalization. Third, given the relatively few number of patients with severe HAA in our cohort, we were unable to assess if the association of severe HAA differed by suspected etiology. Lastly, because we were unable to ascertain the timing of the hematocrit values within the first 24 hours of admission, we excluded both patients with preexisting anemia on admission and those who developed HAA within the first 24 hours of admission, which is not uncommon.²⁶ Thus, we were unable to assess the effect of acute on chronic anemia arising during hospitalization and HAA that develops within the first 24 hours, both of which may also be harmful.^18,27,28

In conclusion, severe HAA occurs in 1.4% of all medical hospitalizations and is associated with increased odds of death or readmission within 30 days. Since most patients with severe HAA had undergone a major procedure or had a discharge diagnosis of hemorrhage or a coagulation or hemorrhagic disorder, it is unclear if severe HAA is potentially preventable through preventing blood loss from phlebotomy or by reducing iatrogenic injury during procedures. Future research should assess the potential preventability of severe HAA, and examine other patient-centered outcomes potentially related to anemia, including fatigue, functional impairment, and trajectory of posthospital recovery.

Acknowledgments

The authors would like to acknowledge Ruben Amarasingham, MD, MBA, President and Chief Executive Officer of the Parkland Center for Clinical Innovation, and Ferdinand Velasco, MD, Chief Health Information Officer at Texas Health Resources, for their assistance in assembling the 6 hospital cohort used in this study. The authors would also like to thank Valy Fontil, MD, MAS, Assistant Professor of Medicine at the University of California San Francisco School of Medicine, and Elizabeth Rogers, MD, MAS, Assistant Professor of Internal Medicine and Pediatrics at the University of Minnesota Medical School, for their constructive feedback on an earlier version of this manuscript.

Disclosures

This work was supported by the Agency for Healthcare Research and Quality-funded UT Southwestern Center for Patient-Centered Outcomes Research (R24 HS022418-01); the Commonwealth Foundation (#20100323); the UT Southwestern KL2 Scholars Program supported by the National Institutes of Health (KL2 TR001103); the National Center for Advancing Translational Sciences at the National Institute of Health (U54 RFA-TR-12-006); and the National Institute on Aging (K23AG052603). The study sponsors had no role in design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript. The authors have no financial conflicts of interest to disclose.

Diagnostic imaging is an integral part of patient evaluation in acute care settings. The use of imaging for presenting complaints of chest pain, abdominal pain, and injuries has increased in emergency departments across the United States without an increase in detection of acute pathologic conditions.^1,2 An unintended consequence of this increase in diagnostic imaging is the discovery of incidental findings (IFs).

Incidental findings are unexpected findings (eg, nodules) noted on diagnostic imaging that are not related to the presenting complaint.³ The increasing use of diagnostic imaging and increased sensitivity of these tests have led to a higher burden of radiologic IFs.⁴ In a tertiary level hospital, Lumbreras et al.⁵ found that the overall incidence of IFs for all radiologic imaging for inpatients and outpatients was 15%, while Orme et al.⁶ found that the incidence in imaging research was 39.8%. The existing evidence base suggests that the identification of radiologic IFs has financial,^5,7 clinical,⁶ ethical, and legal implications.⁸ Also, IFs increase workload for healthcare professionals, including that related to follow-up and surveillance.⁹

In the field of radiology, the burden of radiologic IFs is a well-accepted fact and various white papers have been published by the American College of Radiology on how to address them.^4,7 Hospitalized patients are a population that undergoes a substantial number of diagnostic tests. In the era of accountable care organizations¹⁰ with an emphasis on population health and high-value care, radiologic IFs pose a particular challenge to healthcare providers.

Chest pain is one of the most common reasons for emergency department visits in the United States.¹¹ In this study, we report on radiologic IFs and factors associated with these among patients hospitalized for chest pain of suspected cardiac origin, and we evaluate the hypothesis that radiologic IFs are associated with an increase in LOS in this population.

We conducted a secondary analysis of data from the Chest Pain and Cocaine Study (CPAC). The CPAC study is a cross sectional study of all patients hospitalized with chest pain to our urban academic medical center. Medical records were reviewed to generate a database of all such patients during the study period. The main focus of CPAC was to look at healthcare disparities and resource utilization in patients with or without a concomitant diagnosis of cocaine use.¹²

Study Population

The Figure shows the selection of the study sample for this analysis. The CPaC Study identified 1811 consecutive admissions for chest pain/angina pectoris (based on admitting diagnosis ICD-9-CM codes: 411.x; 413.x, 414.x; and 786.5x) over 24 months. Per the CPaC Study protocol, patients older than 65 years were excluded (n=567 admissions). After chart review, all admissions diagnosed with acute myocardial infarction (n=97) or noncardiac chest pain (n=655) were excluded. For this analysis, we excluded 39 additional admissions of patients who had known prior radiologic IFs, leading to a sample size of 453 admissions. Three hundred and seventy six patients had accounted for 453 admissions during the study period, and we included1 of these admissions in the analysis using the following process: If a patient had a radiologic IF on any admission during the study period, that patient was included in the “IF” group for the analysis, and data from the first admission with an IF were used for the analysis. If a patient had no radiologic IFs on any admission during the study period, that patient was included in the “no IF” group, and the data from the first admission in the database were used for analysis.

Measurements

Data collection was completed retrospectively by medical record review using a standardized CPaC Study protocol. The database was created and maintained using REDCap (Research Electronic Data Capture; Vanderbilt University, Knoxville, Tennessee) electronic data capture tool hosted at Johns Hopkins University.¹³ All data were manually abstracted into REDCap from electronic medical records. All missing values and inconsistent data were reviewed by multiple physicians to ensure data integrity.

We defined all diagnostic (noninterventional; nonlaboratory) testing done during a patient’s hospitalization as “diagnostic” tests, except cardiac stress testing and echocardiogram. We defined diagnostic tests as “primary” tests if they were done in response to patients’ presenting complaint. We defined diagnostic tests as “secondary” tests if they were done by providers due to IFs. Cardiac computed tomography was included in diagnostic tests. Cardiac testing (echocardiogram, cardiac stress testing, cardiac catheterization and pacemaker placement) was considered separate from the “diagnostic tests” since these were focused cardiac imaging that are interventional in nature with low yield on extra-cardiac radiologic IFs.

Incidental findings were defined as any unexpected findings on diagnostic imaging unrelated to the reason for admission, and were classified based on organ systems and their clinical significance as major, moderate, or minor using a classification previously published by Lumbreras et al.¹⁴ All radiologic IFs data underwent sequential dual review by investigators for accuracy of documentation. Individuals with multiple radiologic IFs belonging to more than one category of clinical significance were categorized with the IFs group of highest clinical significance. Ten percent of the patients with no IFs were reviewed again, and no errors found.

Demographic variables at the time of admission included age, sex, race, level of education, employment status, insurance status, body mass index (BMI), and smoking status. Comorbid conditions at the time of admission consisted of the following: hypertension, diabetes mellitus, chronic kidney disease (CKD), chronic obstructive pulmonary disease (COPD), history of myocardial infarction, cerebrovascular accident (CVA), congestive heart failure (CHF), drug use and malignancy or history of it. Initial laboratory values were extracted from electronic medical records and included hemoglobin, creatinine, blood urea nitrogen (BUN), aspartate transaminase, alanine transaminase, and alkaline phosphatase. We calculated the estimated glomerular filtration rate (eGFR) using the MDRD (Modification of Diet in Renal Disease) equation.¹⁵ Admission and discharge information as well as whether the patient had a primary care provider, were obtained from medical records. The length of hospital stay was calculated by subtracting date of admission from date of discharge.

Statistical Analysis

We conducted 2 main analyses: 1) a descriptive analysis of the association between patient characteristics (independent variables) and identification of IFs during admission (primary outcome) and 2) an analysis of the association between identification of incidental findings during admission (independent variable) and LOS (primary outcome).

For the descriptive analysis of radiologic IFs, we compared the characteristics of patients with and without radiologic IFs during admission using a t-test (for normally distributed continuous variables) or Mann-Whitney test (for nonnormally distributed continuous variables) and a chi-square or Fisher exact test for categorical variables based on the number of observations. We included variables significantly associated with the occurrence of radiologic IFs (P < 0.05) in a multiple logistic regression model to identify characteristics independently associated with presence of radiologic IFs.

Length of stay was right-skewed even after natural logarithm transformation and, therefore, we used negative binomial regression for the analysis of the association between the identification of radiologic IFs during admission and LOS. We included potential confounding variables in the multiple negative binomial regression model based on plausibility of confounding and association with both the exposure (identification of radiologic IFs during admission) and outcome (LOS) at a level of P < 0.3. Age, education level, history of drug use, history of CHF, history of CKD, lower eGFR, higher serum creatinine/BUN, hemoglobin, occurrence of cardiac catheterization, stress testing, and multiple admissions during the study period were identified as confounders. For correlated variables (eg, hemoglobin and hematocrit), the variable with the strongest statistical association (lowest P value) was included in the model. In sensitivity analysis, we dropped patients with extreme LOS (longer than 10 days). All analyses were performed using STATA 13 (Stata Statistical Software: Release 13; StataCorp., College Station, Texas).

Table 1 shows the characteristics of the 376 patients included in this study. Overall mean age was 50.5 years, 40% were females, 62% were Caucasian, 66% were unemployed, 84% identified a primary care provider upon admission, and 68% were cared for by a hospitalist. Overall median LOS was 2 days (interquartile range [IQR] = 2). Of the 376 patients in the study, 197 (52%) had new radiologic IFs. Comparing the patients with radiologic IFs and no IFs, it was evident that more radiological tests were performed in the IF group (2.2 tests per patient) in comparison with the no IF group (1.26 tests per patient). Looking at patient characteristics, patients with radiologic IFs were older (52 years vs. 48.8 years; P < 0.001), reported a lower education level and lower hemoglobin levels on admission (12.0 gm/dL vs. 13.4 gm/dL; P = 0.029), but were more likely to be unemployed (72% vs. 59%; P = 0.009), have COPD (19% vs. 10%; P = 0.007), and a history of malignancy (7% vs. 2%, P = 0.04). In addition, patients in the radiologic IF group had lower rates of cardiac catheterization (18% vs. 28%; P = 0.02), were more likely to be readmitted more than once during the study period (17% vs. 7%; P = 0.02) and be discharged by hospitalists (75% vs. 60%; P = 0.003; Supplemental Table 1).

Overall, 658 diagnostic tests were performed in the study population; of these, 268 (40.7%) tests revealed 364 new radiologic IFs (Supplement Table 2). Of these radiologic IFs, 27 (7.4%) were of major clinical significance, 154 (42%) were of moderate clinical significance, and 183 (50%) were of minor clinical significance (Supplement Table 3). Computed tomography (CT) scans yielded more IFs compared to any other imaging modalities. Of the radiologic IFs of major clinical significance, 3 malignant/premalignant lesions were found. While pulmonary nodules were the most common moderate clinically significant findings, atelectasis and spinal degenerative changes were the most common radiologic IFs of minor clinical significance (Supplement Table 4).

Radiologic IFs in patients admitted with chest pain of suspected cardiac origin are a common occurrence as shown in our study. Similar to prior studies, 41% of all radiologic tests done in our study population revealed IFs.⁶ The majority of the IFs were of minor to moderate clinical significance and, as reported in the literature, were more common with older age and CT imaging.^14,16 In addition, an IF diagnosed during admission for chest pain was associated with a 26% increase in length of hospital stay.

To our knowledge, we present the first study on the impact of identification of radiologic IFs in hospitalized patients on length of hospital stay and specifically in patients hospitalized with chest pain of suspected cardiac origin. Trends over the past decade have shown a decrease in LOS and hospitalizations but with an increase in health resource utilization.^17,18 Association of radiologic IFs with increase in LOS is significant as this potentially increases hospital-acquired conditions such as infections and resource utilization leading to increase in costs of hospitalizations.¹⁹ This in return is a concern for patient safety.

The positive association between LOS and radiologic IFs, interestingly, continued to exist despite sensitivity analysis. Incidental findings of major clinical significance were associated with longer LOS in the sensitivity analysis. Supplemental chart review of patients with major clinical findings suggested more extra-cardiac workup compared to patients with minor/moderate radiologic IFs. This could indicate that the presence of clinically significant radiologic IFs could have led to further inpatient work-up and consultations. The downstream healthcare expenditure associated with workup of IFs in individual radiologic tests is well established.²⁰ In case of cardiac CT, Goehler et al.²¹ found that the healthcare expenditure was high following incidentally detected pulmonary nodules with an overall small reduction in lung cancer mortality. Incidental findings also increase the burden of reporting and concern for medico-legal issues for providers.⁴ These concerns are likely valid for hospitalized patients as well.

The socioeconomic trends in the study population were consistent with data from the Bureau of Labor Statistics in that low education is associated with higher unemployment.²² Although, overall, gender, race and insurance mix were similar in both groups, we did see trends of socioeconomic differences in the patients with radiologic IFs of major clinical significance that might not have been statistically significant owing to the small sample size. Despite the population being relatively of younger age (given our cut off age was 65 years) there was still a positive association with age and presence of radiologic IFs. The higher number of patients with COPD or history of malignancy in the radiologic IF group suggests that an association with IFs could exist for these disease cohorts; however, after adjustment for multiple covariates, such an association did not transpire. Interestingly, patients with no radiologic IFs underwent cardiac catheterization or stress testing more often than patients with discovered IFs. This speaks of 2 possibilities; first, that both tests probably do not yield many extra-cardiac IFs, or, secondly, that these patients did not require further workup. More patients in the IF group had more than 1 admission during the study period, and this was associated with increased odds of detecting radiologic IFs. We hypothesize that this might have occurred because of the diagnostic dilemma in these patients who have multiple admissions for the same reason leading to wider array of diagnostic workup. Indeed, we did not note upon chart review alternative diagnoses in these patients but only more IFs. There are several study limitations to consider. First, the fact that this is a single center study sets limitations to interpretation and generalizability of the data. Second, we cannot exclude the possibility of residual confounding. Third, the small number of patients included in this study precludes definitive identification of more factors potentially associated with IFs. However, this study sheds light on a yet unidentified problem within the realm of inpatient management especially for the internists and hospitalists. We tried to limit bias to the extent possible by including only 1 presenting complaint and age-restricting the population.

Incidental findings are both clinical and financial challenges to the medical field. This study attempted to shed light on impact of radiologic IFs on care and resource utilization in patients admitted with chest pain of suspected cardiac origin. The positive association between radiologic IFs and length of hospital stay implies that the presence of IFs is associated with increase in LOS and indirectly a likely increase in overall healthcare expenditure. Given the high incidence of radiologic IFs, assuming that these will be present on radiologic tests, should be more a norm than an exception. Providers should know that radiologic testing, especially CT, is associated with detection of IFs.¹⁶ By avoiding inappropriate ordering of imaging, the issue of IFs could be mitigated.

While radiologists have recommendations about necessary follow-up for some IFs,⁷ no clear follow-up guidelines exist for most IFs arising in hospitalized patients. Further prospective and cost analysis studies are needed to assess the overall impact of IFs on other hospitalized patient populations and on the healthcare system in general.

Disclosure

The authors report no conflicts of interest.

Diagnostic imaging is an integral part of patient evaluation in acute care settings. The use of imaging for presenting complaints of chest pain, abdominal pain, and injuries has increased in emergency departments across the United States without an increase in detection of acute pathologic conditions.^1,2 An unintended consequence of this increase in diagnostic imaging is the discovery of incidental findings (IFs).

Incidental findings are unexpected findings (eg, nodules) noted on diagnostic imaging that are not related to the presenting complaint.³ The increasing use of diagnostic imaging and increased sensitivity of these tests have led to a higher burden of radiologic IFs.⁴ In a tertiary level hospital, Lumbreras et al.⁵ found that the overall incidence of IFs for all radiologic imaging for inpatients and outpatients was 15%, while Orme et al.⁶ found that the incidence in imaging research was 39.8%. The existing evidence base suggests that the identification of radiologic IFs has financial,^5,7 clinical,⁶ ethical, and legal implications.⁸ Also, IFs increase workload for healthcare professionals, including that related to follow-up and surveillance.⁹

In the field of radiology, the burden of radiologic IFs is a well-accepted fact and various white papers have been published by the American College of Radiology on how to address them.^4,7 Hospitalized patients are a population that undergoes a substantial number of diagnostic tests. In the era of accountable care organizations¹⁰ with an emphasis on population health and high-value care, radiologic IFs pose a particular challenge to healthcare providers.

Chest pain is one of the most common reasons for emergency department visits in the United States.¹¹ In this study, we report on radiologic IFs and factors associated with these among patients hospitalized for chest pain of suspected cardiac origin, and we evaluate the hypothesis that radiologic IFs are associated with an increase in LOS in this population.

We conducted a secondary analysis of data from the Chest Pain and Cocaine Study (CPAC). The CPAC study is a cross sectional study of all patients hospitalized with chest pain to our urban academic medical center. Medical records were reviewed to generate a database of all such patients during the study period. The main focus of CPAC was to look at healthcare disparities and resource utilization in patients with or without a concomitant diagnosis of cocaine use.¹²

Study Population

The Figure shows the selection of the study sample for this analysis. The CPaC Study identified 1811 consecutive admissions for chest pain/angina pectoris (based on admitting diagnosis ICD-9-CM codes: 411.x; 413.x, 414.x; and 786.5x) over 24 months. Per the CPaC Study protocol, patients older than 65 years were excluded (n=567 admissions). After chart review, all admissions diagnosed with acute myocardial infarction (n=97) or noncardiac chest pain (n=655) were excluded. For this analysis, we excluded 39 additional admissions of patients who had known prior radiologic IFs, leading to a sample size of 453 admissions. Three hundred and seventy six patients had accounted for 453 admissions during the study period, and we included1 of these admissions in the analysis using the following process: If a patient had a radiologic IF on any admission during the study period, that patient was included in the “IF” group for the analysis, and data from the first admission with an IF were used for the analysis. If a patient had no radiologic IFs on any admission during the study period, that patient was included in the “no IF” group, and the data from the first admission in the database were used for analysis.

Measurements

Data collection was completed retrospectively by medical record review using a standardized CPaC Study protocol. The database was created and maintained using REDCap (Research Electronic Data Capture; Vanderbilt University, Knoxville, Tennessee) electronic data capture tool hosted at Johns Hopkins University.¹³ All data were manually abstracted into REDCap from electronic medical records. All missing values and inconsistent data were reviewed by multiple physicians to ensure data integrity.

We defined all diagnostic (noninterventional; nonlaboratory) testing done during a patient’s hospitalization as “diagnostic” tests, except cardiac stress testing and echocardiogram. We defined diagnostic tests as “primary” tests if they were done in response to patients’ presenting complaint. We defined diagnostic tests as “secondary” tests if they were done by providers due to IFs. Cardiac computed tomography was included in diagnostic tests. Cardiac testing (echocardiogram, cardiac stress testing, cardiac catheterization and pacemaker placement) was considered separate from the “diagnostic tests” since these were focused cardiac imaging that are interventional in nature with low yield on extra-cardiac radiologic IFs.

Incidental findings were defined as any unexpected findings on diagnostic imaging unrelated to the reason for admission, and were classified based on organ systems and their clinical significance as major, moderate, or minor using a classification previously published by Lumbreras et al.¹⁴ All radiologic IFs data underwent sequential dual review by investigators for accuracy of documentation. Individuals with multiple radiologic IFs belonging to more than one category of clinical significance were categorized with the IFs group of highest clinical significance. Ten percent of the patients with no IFs were reviewed again, and no errors found.

Demographic variables at the time of admission included age, sex, race, level of education, employment status, insurance status, body mass index (BMI), and smoking status. Comorbid conditions at the time of admission consisted of the following: hypertension, diabetes mellitus, chronic kidney disease (CKD), chronic obstructive pulmonary disease (COPD), history of myocardial infarction, cerebrovascular accident (CVA), congestive heart failure (CHF), drug use and malignancy or history of it. Initial laboratory values were extracted from electronic medical records and included hemoglobin, creatinine, blood urea nitrogen (BUN), aspartate transaminase, alanine transaminase, and alkaline phosphatase. We calculated the estimated glomerular filtration rate (eGFR) using the MDRD (Modification of Diet in Renal Disease) equation.¹⁵ Admission and discharge information as well as whether the patient had a primary care provider, were obtained from medical records. The length of hospital stay was calculated by subtracting date of admission from date of discharge.

Statistical Analysis

We conducted 2 main analyses: 1) a descriptive analysis of the association between patient characteristics (independent variables) and identification of IFs during admission (primary outcome) and 2) an analysis of the association between identification of incidental findings during admission (independent variable) and LOS (primary outcome).

For the descriptive analysis of radiologic IFs, we compared the characteristics of patients with and without radiologic IFs during admission using a t-test (for normally distributed continuous variables) or Mann-Whitney test (for nonnormally distributed continuous variables) and a chi-square or Fisher exact test for categorical variables based on the number of observations. We included variables significantly associated with the occurrence of radiologic IFs (P < 0.05) in a multiple logistic regression model to identify characteristics independently associated with presence of radiologic IFs.

Length of stay was right-skewed even after natural logarithm transformation and, therefore, we used negative binomial regression for the analysis of the association between the identification of radiologic IFs during admission and LOS. We included potential confounding variables in the multiple negative binomial regression model based on plausibility of confounding and association with both the exposure (identification of radiologic IFs during admission) and outcome (LOS) at a level of P < 0.3. Age, education level, history of drug use, history of CHF, history of CKD, lower eGFR, higher serum creatinine/BUN, hemoglobin, occurrence of cardiac catheterization, stress testing, and multiple admissions during the study period were identified as confounders. For correlated variables (eg, hemoglobin and hematocrit), the variable with the strongest statistical association (lowest P value) was included in the model. In sensitivity analysis, we dropped patients with extreme LOS (longer than 10 days). All analyses were performed using STATA 13 (Stata Statistical Software: Release 13; StataCorp., College Station, Texas).

Table 1 shows the characteristics of the 376 patients included in this study. Overall mean age was 50.5 years, 40% were females, 62% were Caucasian, 66% were unemployed, 84% identified a primary care provider upon admission, and 68% were cared for by a hospitalist. Overall median LOS was 2 days (interquartile range [IQR] = 2). Of the 376 patients in the study, 197 (52%) had new radiologic IFs. Comparing the patients with radiologic IFs and no IFs, it was evident that more radiological tests were performed in the IF group (2.2 tests per patient) in comparison with the no IF group (1.26 tests per patient). Looking at patient characteristics, patients with radiologic IFs were older (52 years vs. 48.8 years; P < 0.001), reported a lower education level and lower hemoglobin levels on admission (12.0 gm/dL vs. 13.4 gm/dL; P = 0.029), but were more likely to be unemployed (72% vs. 59%; P = 0.009), have COPD (19% vs. 10%; P = 0.007), and a history of malignancy (7% vs. 2%, P = 0.04). In addition, patients in the radiologic IF group had lower rates of cardiac catheterization (18% vs. 28%; P = 0.02), were more likely to be readmitted more than once during the study period (17% vs. 7%; P = 0.02) and be discharged by hospitalists (75% vs. 60%; P = 0.003; Supplemental Table 1).

Overall, 658 diagnostic tests were performed in the study population; of these, 268 (40.7%) tests revealed 364 new radiologic IFs (Supplement Table 2). Of these radiologic IFs, 27 (7.4%) were of major clinical significance, 154 (42%) were of moderate clinical significance, and 183 (50%) were of minor clinical significance (Supplement Table 3). Computed tomography (CT) scans yielded more IFs compared to any other imaging modalities. Of the radiologic IFs of major clinical significance, 3 malignant/premalignant lesions were found. While pulmonary nodules were the most common moderate clinically significant findings, atelectasis and spinal degenerative changes were the most common radiologic IFs of minor clinical significance (Supplement Table 4).

Radiologic IFs in patients admitted with chest pain of suspected cardiac origin are a common occurrence as shown in our study. Similar to prior studies, 41% of all radiologic tests done in our study population revealed IFs.⁶ The majority of the IFs were of minor to moderate clinical significance and, as reported in the literature, were more common with older age and CT imaging.^14,16 In addition, an IF diagnosed during admission for chest pain was associated with a 26% increase in length of hospital stay.

To our knowledge, we present the first study on the impact of identification of radiologic IFs in hospitalized patients on length of hospital stay and specifically in patients hospitalized with chest pain of suspected cardiac origin. Trends over the past decade have shown a decrease in LOS and hospitalizations but with an increase in health resource utilization.^17,18 Association of radiologic IFs with increase in LOS is significant as this potentially increases hospital-acquired conditions such as infections and resource utilization leading to increase in costs of hospitalizations.¹⁹ This in return is a concern for patient safety.

The positive association between LOS and radiologic IFs, interestingly, continued to exist despite sensitivity analysis. Incidental findings of major clinical significance were associated with longer LOS in the sensitivity analysis. Supplemental chart review of patients with major clinical findings suggested more extra-cardiac workup compared to patients with minor/moderate radiologic IFs. This could indicate that the presence of clinically significant radiologic IFs could have led to further inpatient work-up and consultations. The downstream healthcare expenditure associated with workup of IFs in individual radiologic tests is well established.²⁰ In case of cardiac CT, Goehler et al.²¹ found that the healthcare expenditure was high following incidentally detected pulmonary nodules with an overall small reduction in lung cancer mortality. Incidental findings also increase the burden of reporting and concern for medico-legal issues for providers.⁴ These concerns are likely valid for hospitalized patients as well.

The socioeconomic trends in the study population were consistent with data from the Bureau of Labor Statistics in that low education is associated with higher unemployment.²² Although, overall, gender, race and insurance mix were similar in both groups, we did see trends of socioeconomic differences in the patients with radiologic IFs of major clinical significance that might not have been statistically significant owing to the small sample size. Despite the population being relatively of younger age (given our cut off age was 65 years) there was still a positive association with age and presence of radiologic IFs. The higher number of patients with COPD or history of malignancy in the radiologic IF group suggests that an association with IFs could exist for these disease cohorts; however, after adjustment for multiple covariates, such an association did not transpire. Interestingly, patients with no radiologic IFs underwent cardiac catheterization or stress testing more often than patients with discovered IFs. This speaks of 2 possibilities; first, that both tests probably do not yield many extra-cardiac IFs, or, secondly, that these patients did not require further workup. More patients in the IF group had more than 1 admission during the study period, and this was associated with increased odds of detecting radiologic IFs. We hypothesize that this might have occurred because of the diagnostic dilemma in these patients who have multiple admissions for the same reason leading to wider array of diagnostic workup. Indeed, we did not note upon chart review alternative diagnoses in these patients but only more IFs. There are several study limitations to consider. First, the fact that this is a single center study sets limitations to interpretation and generalizability of the data. Second, we cannot exclude the possibility of residual confounding. Third, the small number of patients included in this study precludes definitive identification of more factors potentially associated with IFs. However, this study sheds light on a yet unidentified problem within the realm of inpatient management especially for the internists and hospitalists. We tried to limit bias to the extent possible by including only 1 presenting complaint and age-restricting the population.

Incidental findings are both clinical and financial challenges to the medical field. This study attempted to shed light on impact of radiologic IFs on care and resource utilization in patients admitted with chest pain of suspected cardiac origin. The positive association between radiologic IFs and length of hospital stay implies that the presence of IFs is associated with increase in LOS and indirectly a likely increase in overall healthcare expenditure. Given the high incidence of radiologic IFs, assuming that these will be present on radiologic tests, should be more a norm than an exception. Providers should know that radiologic testing, especially CT, is associated with detection of IFs.¹⁶ By avoiding inappropriate ordering of imaging, the issue of IFs could be mitigated.

While radiologists have recommendations about necessary follow-up for some IFs,⁷ no clear follow-up guidelines exist for most IFs arising in hospitalized patients. Further prospective and cost analysis studies are needed to assess the overall impact of IFs on other hospitalized patient populations and on the healthcare system in general.

Disclosure

The authors report no conflicts of interest.

Diagnostic imaging is an integral part of patient evaluation in acute care settings. The use of imaging for presenting complaints of chest pain, abdominal pain, and injuries has increased in emergency departments across the United States without an increase in detection of acute pathologic conditions.^1,2 An unintended consequence of this increase in diagnostic imaging is the discovery of incidental findings (IFs).

Incidental findings are unexpected findings (eg, nodules) noted on diagnostic imaging that are not related to the presenting complaint.³ The increasing use of diagnostic imaging and increased sensitivity of these tests have led to a higher burden of radiologic IFs.⁴ In a tertiary level hospital, Lumbreras et al.⁵ found that the overall incidence of IFs for all radiologic imaging for inpatients and outpatients was 15%, while Orme et al.⁶ found that the incidence in imaging research was 39.8%. The existing evidence base suggests that the identification of radiologic IFs has financial,^5,7 clinical,⁶ ethical, and legal implications.⁸ Also, IFs increase workload for healthcare professionals, including that related to follow-up and surveillance.⁹

In the field of radiology, the burden of radiologic IFs is a well-accepted fact and various white papers have been published by the American College of Radiology on how to address them.^4,7 Hospitalized patients are a population that undergoes a substantial number of diagnostic tests. In the era of accountable care organizations¹⁰ with an emphasis on population health and high-value care, radiologic IFs pose a particular challenge to healthcare providers.

Chest pain is one of the most common reasons for emergency department visits in the United States.¹¹ In this study, we report on radiologic IFs and factors associated with these among patients hospitalized for chest pain of suspected cardiac origin, and we evaluate the hypothesis that radiologic IFs are associated with an increase in LOS in this population.

We conducted a secondary analysis of data from the Chest Pain and Cocaine Study (CPAC). The CPAC study is a cross sectional study of all patients hospitalized with chest pain to our urban academic medical center. Medical records were reviewed to generate a database of all such patients during the study period. The main focus of CPAC was to look at healthcare disparities and resource utilization in patients with or without a concomitant diagnosis of cocaine use.¹²

Study Population

The Figure shows the selection of the study sample for this analysis. The CPaC Study identified 1811 consecutive admissions for chest pain/angina pectoris (based on admitting diagnosis ICD-9-CM codes: 411.x; 413.x, 414.x; and 786.5x) over 24 months. Per the CPaC Study protocol, patients older than 65 years were excluded (n=567 admissions). After chart review, all admissions diagnosed with acute myocardial infarction (n=97) or noncardiac chest pain (n=655) were excluded. For this analysis, we excluded 39 additional admissions of patients who had known prior radiologic IFs, leading to a sample size of 453 admissions. Three hundred and seventy six patients had accounted for 453 admissions during the study period, and we included1 of these admissions in the analysis using the following process: If a patient had a radiologic IF on any admission during the study period, that patient was included in the “IF” group for the analysis, and data from the first admission with an IF were used for the analysis. If a patient had no radiologic IFs on any admission during the study period, that patient was included in the “no IF” group, and the data from the first admission in the database were used for analysis.

Measurements

Data collection was completed retrospectively by medical record review using a standardized CPaC Study protocol. The database was created and maintained using REDCap (Research Electronic Data Capture; Vanderbilt University, Knoxville, Tennessee) electronic data capture tool hosted at Johns Hopkins University.¹³ All data were manually abstracted into REDCap from electronic medical records. All missing values and inconsistent data were reviewed by multiple physicians to ensure data integrity.

We defined all diagnostic (noninterventional; nonlaboratory) testing done during a patient’s hospitalization as “diagnostic” tests, except cardiac stress testing and echocardiogram. We defined diagnostic tests as “primary” tests if they were done in response to patients’ presenting complaint. We defined diagnostic tests as “secondary” tests if they were done by providers due to IFs. Cardiac computed tomography was included in diagnostic tests. Cardiac testing (echocardiogram, cardiac stress testing, cardiac catheterization and pacemaker placement) was considered separate from the “diagnostic tests” since these were focused cardiac imaging that are interventional in nature with low yield on extra-cardiac radiologic IFs.

Incidental findings were defined as any unexpected findings on diagnostic imaging unrelated to the reason for admission, and were classified based on organ systems and their clinical significance as major, moderate, or minor using a classification previously published by Lumbreras et al.¹⁴ All radiologic IFs data underwent sequential dual review by investigators for accuracy of documentation. Individuals with multiple radiologic IFs belonging to more than one category of clinical significance were categorized with the IFs group of highest clinical significance. Ten percent of the patients with no IFs were reviewed again, and no errors found.

Demographic variables at the time of admission included age, sex, race, level of education, employment status, insurance status, body mass index (BMI), and smoking status. Comorbid conditions at the time of admission consisted of the following: hypertension, diabetes mellitus, chronic kidney disease (CKD), chronic obstructive pulmonary disease (COPD), history of myocardial infarction, cerebrovascular accident (CVA), congestive heart failure (CHF), drug use and malignancy or history of it. Initial laboratory values were extracted from electronic medical records and included hemoglobin, creatinine, blood urea nitrogen (BUN), aspartate transaminase, alanine transaminase, and alkaline phosphatase. We calculated the estimated glomerular filtration rate (eGFR) using the MDRD (Modification of Diet in Renal Disease) equation.¹⁵ Admission and discharge information as well as whether the patient had a primary care provider, were obtained from medical records. The length of hospital stay was calculated by subtracting date of admission from date of discharge.

Statistical Analysis

We conducted 2 main analyses: 1) a descriptive analysis of the association between patient characteristics (independent variables) and identification of IFs during admission (primary outcome) and 2) an analysis of the association between identification of incidental findings during admission (independent variable) and LOS (primary outcome).

For the descriptive analysis of radiologic IFs, we compared the characteristics of patients with and without radiologic IFs during admission using a t-test (for normally distributed continuous variables) or Mann-Whitney test (for nonnormally distributed continuous variables) and a chi-square or Fisher exact test for categorical variables based on the number of observations. We included variables significantly associated with the occurrence of radiologic IFs (P < 0.05) in a multiple logistic regression model to identify characteristics independently associated with presence of radiologic IFs.

Length of stay was right-skewed even after natural logarithm transformation and, therefore, we used negative binomial regression for the analysis of the association between the identification of radiologic IFs during admission and LOS. We included potential confounding variables in the multiple negative binomial regression model based on plausibility of confounding and association with both the exposure (identification of radiologic IFs during admission) and outcome (LOS) at a level of P < 0.3. Age, education level, history of drug use, history of CHF, history of CKD, lower eGFR, higher serum creatinine/BUN, hemoglobin, occurrence of cardiac catheterization, stress testing, and multiple admissions during the study period were identified as confounders. For correlated variables (eg, hemoglobin and hematocrit), the variable with the strongest statistical association (lowest P value) was included in the model. In sensitivity analysis, we dropped patients with extreme LOS (longer than 10 days). All analyses were performed using STATA 13 (Stata Statistical Software: Release 13; StataCorp., College Station, Texas).

Table 1 shows the characteristics of the 376 patients included in this study. Overall mean age was 50.5 years, 40% were females, 62% were Caucasian, 66% were unemployed, 84% identified a primary care provider upon admission, and 68% were cared for by a hospitalist. Overall median LOS was 2 days (interquartile range [IQR] = 2). Of the 376 patients in the study, 197 (52%) had new radiologic IFs. Comparing the patients with radiologic IFs and no IFs, it was evident that more radiological tests were performed in the IF group (2.2 tests per patient) in comparison with the no IF group (1.26 tests per patient). Looking at patient characteristics, patients with radiologic IFs were older (52 years vs. 48.8 years; P < 0.001), reported a lower education level and lower hemoglobin levels on admission (12.0 gm/dL vs. 13.4 gm/dL; P = 0.029), but were more likely to be unemployed (72% vs. 59%; P = 0.009), have COPD (19% vs. 10%; P = 0.007), and a history of malignancy (7% vs. 2%, P = 0.04). In addition, patients in the radiologic IF group had lower rates of cardiac catheterization (18% vs. 28%; P = 0.02), were more likely to be readmitted more than once during the study period (17% vs. 7%; P = 0.02) and be discharged by hospitalists (75% vs. 60%; P = 0.003; Supplemental Table 1).

Overall, 658 diagnostic tests were performed in the study population; of these, 268 (40.7%) tests revealed 364 new radiologic IFs (Supplement Table 2). Of these radiologic IFs, 27 (7.4%) were of major clinical significance, 154 (42%) were of moderate clinical significance, and 183 (50%) were of minor clinical significance (Supplement Table 3). Computed tomography (CT) scans yielded more IFs compared to any other imaging modalities. Of the radiologic IFs of major clinical significance, 3 malignant/premalignant lesions were found. While pulmonary nodules were the most common moderate clinically significant findings, atelectasis and spinal degenerative changes were the most common radiologic IFs of minor clinical significance (Supplement Table 4).

Radiologic IFs in patients admitted with chest pain of suspected cardiac origin are a common occurrence as shown in our study. Similar to prior studies, 41% of all radiologic tests done in our study population revealed IFs.⁶ The majority of the IFs were of minor to moderate clinical significance and, as reported in the literature, were more common with older age and CT imaging.^14,16 In addition, an IF diagnosed during admission for chest pain was associated with a 26% increase in length of hospital stay.

To our knowledge, we present the first study on the impact of identification of radiologic IFs in hospitalized patients on length of hospital stay and specifically in patients hospitalized with chest pain of suspected cardiac origin. Trends over the past decade have shown a decrease in LOS and hospitalizations but with an increase in health resource utilization.^17,18 Association of radiologic IFs with increase in LOS is significant as this potentially increases hospital-acquired conditions such as infections and resource utilization leading to increase in costs of hospitalizations.¹⁹ This in return is a concern for patient safety.

The positive association between LOS and radiologic IFs, interestingly, continued to exist despite sensitivity analysis. Incidental findings of major clinical significance were associated with longer LOS in the sensitivity analysis. Supplemental chart review of patients with major clinical findings suggested more extra-cardiac workup compared to patients with minor/moderate radiologic IFs. This could indicate that the presence of clinically significant radiologic IFs could have led to further inpatient work-up and consultations. The downstream healthcare expenditure associated with workup of IFs in individual radiologic tests is well established.²⁰ In case of cardiac CT, Goehler et al.²¹ found that the healthcare expenditure was high following incidentally detected pulmonary nodules with an overall small reduction in lung cancer mortality. Incidental findings also increase the burden of reporting and concern for medico-legal issues for providers.⁴ These concerns are likely valid for hospitalized patients as well.

The socioeconomic trends in the study population were consistent with data from the Bureau of Labor Statistics in that low education is associated with higher unemployment.²² Although, overall, gender, race and insurance mix were similar in both groups, we did see trends of socioeconomic differences in the patients with radiologic IFs of major clinical significance that might not have been statistically significant owing to the small sample size. Despite the population being relatively of younger age (given our cut off age was 65 years) there was still a positive association with age and presence of radiologic IFs. The higher number of patients with COPD or history of malignancy in the radiologic IF group suggests that an association with IFs could exist for these disease cohorts; however, after adjustment for multiple covariates, such an association did not transpire. Interestingly, patients with no radiologic IFs underwent cardiac catheterization or stress testing more often than patients with discovered IFs. This speaks of 2 possibilities; first, that both tests probably do not yield many extra-cardiac IFs, or, secondly, that these patients did not require further workup. More patients in the IF group had more than 1 admission during the study period, and this was associated with increased odds of detecting radiologic IFs. We hypothesize that this might have occurred because of the diagnostic dilemma in these patients who have multiple admissions for the same reason leading to wider array of diagnostic workup. Indeed, we did not note upon chart review alternative diagnoses in these patients but only more IFs. There are several study limitations to consider. First, the fact that this is a single center study sets limitations to interpretation and generalizability of the data. Second, we cannot exclude the possibility of residual confounding. Third, the small number of patients included in this study precludes definitive identification of more factors potentially associated with IFs. However, this study sheds light on a yet unidentified problem within the realm of inpatient management especially for the internists and hospitalists. We tried to limit bias to the extent possible by including only 1 presenting complaint and age-restricting the population.

Incidental findings are both clinical and financial challenges to the medical field. This study attempted to shed light on impact of radiologic IFs on care and resource utilization in patients admitted with chest pain of suspected cardiac origin. The positive association between radiologic IFs and length of hospital stay implies that the presence of IFs is associated with increase in LOS and indirectly a likely increase in overall healthcare expenditure. Given the high incidence of radiologic IFs, assuming that these will be present on radiologic tests, should be more a norm than an exception. Providers should know that radiologic testing, especially CT, is associated with detection of IFs.¹⁶ By avoiding inappropriate ordering of imaging, the issue of IFs could be mitigated.

While radiologists have recommendations about necessary follow-up for some IFs,⁷ no clear follow-up guidelines exist for most IFs arising in hospitalized patients. Further prospective and cost analysis studies are needed to assess the overall impact of IFs on other hospitalized patient populations and on the healthcare system in general.

Disclosure

The authors report no conflicts of interest.