Original Research

ABSTRACT

The purpose of this study is to assess the effect of insurance type (Medicaid, Medicare, private insurance) on the ability for patients with operative ankle fractures to access orthopedic traumatologists. The research team called 245 board-certified orthopedic surgeons specializing in orthopedic trauma within 8 representative states. The caller requested an appointment for their fictitious mother in order to be evaluated for an ankle fracture which was previously evaluated by her primary care physician and believed to require surgery. Each office was called 3 times to assess the response for each insurance type. For each call, information was documented regarding whether the patient was able to receive an appointment and the barriers the patient confronted to receive an appointment. Overall, 35.7% of offices scheduled an appointment for a patient with Medicaid, in comparison to 81.4%and 88.6% for Medicare and BlueCross, respectively (P < .0001). Medicaid patients confronted more barriers for receiving appointments. There was no statistically significant difference in access for Medicaid patients in states that had expanded Medicaid eligibility vs states that had not expanded Medicaid. Medicaid reimbursement for open reduction and internal fixation of an ankle fracture did not significantly correlate with appointment success rates or wait times. Despite the passage of the Affordable Care Act, patients with Medicaid have reduced access to orthopedic surgeons and more complex barriers to receiving appointments. A more robust strategy for increasing care-access for patients with Medicaid would be more equitable.

Continue to: In 2010, the Patient Protection and Affordable Care Act...

In 2010, the Patient Protection and Affordable Care Act (PPACA) expanded the eligibility criteria for Medicaid to all individuals with an income up to 138% of the poverty level.¹ A Supreme Court ruling stated that the decision to expand Medicaid was to be decided by individual states.² Currently, 31 states have chosen to expand Medicaid eligibility to their residents.² This expansion has allowed an additional 11.7 million people to enroll in Medicaid and the Children’s Health Insurance Program by May 2015.^3-5

Even with the passage of the PPACA, Medicaid patients seeking specialty orthopedic care have experienced more barriers to accessing care than Medicare or commercially-insured patients.^2,6-10 One major cited reason is Medicaid’s low reimbursement, which may discourage physicians from open panel participation in Medicaid.^11,12  

A common fundamental teaching for orthopedic traumatologists is the notion that they should be available to treat all injuries regardless of the patient’s ability to pay.¹³ This has resulted in both trauma centers and trauma surgeons becoming financially challenged due to the higher proportion of Medicaid and uninsured trauma patients and lower Medicaid reimbursement levels.^14,15

This study focuses on the effect of different types of insurance (Medicaid, Medicare, or commercial insurance) on the ability of patients to obtain care for operative ankle fractures. The purpose of this study is to evaluate, in the context of the PPACA, patient access to orthopedic surgeons for operative ankle fractures based on insurance-type. We hypothesized that patients with Medicaid would face a greater volume of obstacles when seeking appointments for an ankle fracture, even after the PPACA.

Continue to: MATERIALS AND METHODS...

MATERIALS AND METHODS

The study population included board-certified orthopedic surgeons who belonged to the Orthopaedic Trauma Association (OTA) from 8 representative states; 4 states with expanded Medicaid eligibility (California, Massachusetts, New York, Ohio) and 4 states without expanded Medicaid eligibility (Florida, North Carolina, Georgia, Texas). These states were selected due to their ability to represent diverse healthcare marketplaces throughout the country. Using the OTA website’s “Find a Surgeon” search tool,¹⁶ we created a list of surgeons for each state and matched each surgeon with a random number. The list of surgeons was ordered according to the value of the surgeon’s associated random number, and surgeons were called in ascending order. We excluded disconnected or inaccurate numbers from the calling list. Surgeons who did not manage ankle fractures were removed from the dataset. Approximately 30 orthopedic trauma surgeons per state were contacted.

Each office was called to make an appointment for the caller’s mother. Every surgeon’s office was specifically asked if the surgeon would accept the patient to be evaluated for an ankle fracture that occurred out-of-state. The caller had a standardized protocol to limit intra- and inter-office variations (Appendix). The scenario involved a request to be evaluated for an unstable ankle fracture, with the patient having Medicaid, Medicare, or BlueCross insurance. The scenario required 3 separate calls to the same surgeon in order to obtain data regarding each insurance-type. The calls were separated by at least 1 week to avoid caller recognition by the surgeon’s office.

Appendix

Scenario

1. Date of Birth: Medicaid–2/07/55; BlueCross PPO–2/09/55; Medicare–7/31/45.

2. Ankle fracture evaluated by primary care physician 1 or 2 days ago

3. Not seen previously by your clinic or hospital, she would be a new patient

4. Asked how early she could be scheduled for an appointment

5. Script:

“I’m calling for my mother who injured her ankle a few days ago. Her family doctor took an X-ray and believes she has a fracture and needs surgery. Is Dr. X accepting new patients for evaluation and treatment of ankle fractures?” If YES →

“I was wondering if you take Medicaid/Medicare/BlueCross plan?” If YES →

“When is your soonest available appointment?”

The date of each phone call and date of appointment, if provided, were recorded. If the office did not give an appointment, we asked for reasons why. If an appointment was denied for a patient with Medicaid, we asked for a referral to another office that accepted Medicaid. We considered barriers to obtaining an initial appointment, such as requiring a referral from a primary care physician (PCP), as an unsuccessful attempt at making an appointment. We determined the waiting period for an appointment by calculating the time between the date of the call and the date of the appointment. Appointments were not scheduled to ensure that actual patients were not disadvantaged. For both appointment success rates and waiting periods, we stratified the data into 2 groups: states with expanded Medicaid eligibility (California, Massachusetts, New York, Ohio) and states without expanded Medicaid eligibility (Florida, North Carolina, Georgia, Texas).

We obtained Medicaid reimbursement rates for open reduction and internal fixation of an ankle fracture by querying each state’s reimbursement rate using Current Procedural Terminology code 27822.

Chi-square test or Fisher’s exact test was used to analyze acceptance rate differences based on the patient’s type of insurance. To compare the waiting periods for an appointment, we used an independent samples t-test after applying natural log-transformation, as the data was not normally distributed. We performed logistic regression analysis to detect whether reimbursement was a significant predictor of successfully making an appointment for patients, and a linear regression analysis was used to evaluate whether reimbursement predicted waiting periods. Unless otherwise stated, all statistical testing was performed two-tailed at an alpha-level of 0.05.

This study was approved by the Institutional Review Board of Yale University School of Medicine (HIC No. 1363).

Continue to: RESULTS...

RESULTS

In total, 350 offices were contacted across 8 states (4 states with and 4 states without expanded Medicaid eligibility) of which we identified 245 orthopedic surgeons who would surgically treat ankle fractures. The 245 surgeons’ offices were called 3 times for each separate insurance-type.

Table 1. Appointment Success Rate

^aComparison to Medicaid.

The overall rate of successfully being offered an appointment with Medicaid was 35.7%, 81.4% for Medicare, and 88.6% for BlueCross (Table 1). For states with expanded Medicaid eligibility, the success rate for obtaining an appointment was 39.6%, 83.5%, and 89.2% for Medicaid, Medicare, and BlueCross, respectively. For states without expanded Medicaid eligibility, the success rate for obtaining an appointment was 31.9% for Medicaid, 79.4% for Medicare, and 87.9% for BlueCross. In all cases, the success rate for obtaining an appointment was significantly lower for Medicaid, compared to Medicare (P < .0001) or BlueCross (P < .0001). Medicaid appointment success rate was 39.6% in expanded states vs 31.9% in non-expanded states, however, the difference was not statistically significant (Table 2).

Table 2. Medicaid Appointment Success Rate in Expanded Vs Non-Expanded States

In 43.7% of occasions, patients with Medicaid did not have their insurance accepted, compared to 7.3% for Medicare and 0% for BlueCross. The majority of offices which did not accept Medicaid were not able to refer patients to another surgeon who would accept Medicaid. The requirement to have a primary care referral was the second most common reason for Medicaid patients not obtaining an appointment. No Medicare (10.4% vs 0.0%, P < .0001) or BlueCross (10.4% vs 0.0%, P < .0001) patients experienced this requirement (Table 3). There was no difference found between the percent of Medicaid patients who were required to have referrals in states with and without expanded Medicaid eligibility (Table 4).

Table 3. Referral Rate

^aComparison to Medicaid.

Table 4. Medicaid Referral Rates in Expanded Vs Non-Expanded States

Reimbursements for ankle fracture varied across states (Table 5). For Medicaid, Georgia paid the highest reimbursement ($1049.95) and Florida paid the lowest ($469.44). Logistic and linear regression analysis did not demonstrate a significant relationship between reimbursement and appointment success rate or waiting periods.

Table 5. Medicaid Reimbursements for Ankle Fracture Repair (CPT and HCPCS 27822) in 2014

^aStates with expanded Medicaid eligibility.

Abbreviations: CPT, Current Procedural Terminology; HCPCS, Healthcare Common Procedure Coding System.

Waiting periods (Table 6) varied significantly by the type of insurance (7.3 days for Medicaid, 6.0 days for Medicare, and 6.0 days for BlueCross; P = .002). For states with expanded Medicaid eligibility, waiting periods varied significantly by insurance (7.7 days for Medicaid, 6.2 days for Medicare, P = .003; and 6.1 days for BlueCross, P = .01). Waiting periods did not vary significantly for states without expanded Medicaid. Additionally, waiting periods did not differ significantly when comparing between states with and without Medicaid expansion.

Table 6. Waiting Period (Days) by Insurance Type.

Continue to: DISCUSSION...

DISCUSSION

This study assessed how insurance type (Medicaid, Medicare, and BlueCross) affects patient access to orthopedic trauma surgeons in 8 geographically representative states. We selected unstable ankle fractures as they are basic fractures treated by nearly all trauma surgeons and should often be surgically treated to prevent serious long-term consequences. Our hypothesis stated that despite the passage of the PPACA, patients with Medicaid would have reduced access to care. As the PPACA has changed the healthcare marketplace by increasing the number of Medicaid enrollees, it is important to ensure that patient access to care improves.

This nationwide survey of orthopedic trauma surgeons demonstrates that Medicaid patients experience added barriers to care that ultimately results in lower rates of successfully obtaining care. This is consistent with other investigations which have assessed Medicaid patient healthcare access.^6,8,10,17-19 This study did not demonstrate a statistically significant difference between Medicaid patients’ ability to obtain appointments in states with expanded Medicaid eligibility vs in states without expanded Medicaid eligibility (39.6% vs 31.9%, P < .18); this has been demonstrated in the literature.⁶  

A barrier that was unique to Medicaid patients was the requirement to have a PCP referral (Table 3). A PCP referral was not a barrier to receiving an appointment for patients with Medicare or BlueCross. One reason to explain why Medicaid patients may be required to have PCP referrals is due to their increased medical complexity, extra documentation requirements, and low reimbursement.⁴ Patients who have obtained a PCP referral may be characterized as being more medically compliant.

It is important to note that the Medicaid policies for 4 states included in this study (Massachusetts, North Carolina, Texas, and New York) required a PCP referral in order to see a specialist. However, we found that many orthopedic trauma practices in these states scheduled appointments for Medicaid patients without a PCP referral, suggesting that the decision depended on individual policy. In addition, the majority of offices within these states cited that they simply did not accept Medicaid as an insurance policy, and not that they required a referral.

Our regression analysis did not find a significant relationship between being able to successfully obtain an appointment to be evaluated for an ankle fracture and reimbursement rates for Medicaid. Although studies have stressed the importance of Medicaid reimbursements on physician participation, this result is consistent with previous studies regarding carpal tunnel release and total ankle replacements.^17,19 Long²⁰ suggested that although reimbursements may help, additional strategies for promoting Medicaid acceptance may be needed, including: lowering the costs of participating in Medicaid by simplifying administrative processes, speeding up reimbursement, and reducing the costs associated with caring for those patients.

Continue to: Previous studies have demonstrated...

Previous studies have demonstrated that more physicians may accept Medicaid if reimbursements increased.^4,12 Given the high percentage of trauma patients with Medicaid as their primary insurance or whom are emergently enrolled in Medicaid by hospital systems, it is concerning that the PPACA is reducing payments under the Medicare and Medicaid Disproportionate Share Hospital programs which provide hospitals for uncompensated care given to low-income and uninsured patients.²¹ Trauma centers generally operate at a deficit due to the higher proportion of Medicaid and uninsured patients.¹⁴ This is currently worsened by additional federal funding cuts for supporting trauma service’s humane mission.²¹

This study has several limitations. While the study evaluated access to care in 8 representative states, a thorough nationwide survey would be more representative. Some results may have become statistically significant if we had performed the study with a larger sample size. In addition, we were unable to control for many factors which could impact appointment wait times, such as physician call schedules and vacations. Socioeconomic factors can influence a patient’s ability to attend an appointment, such as transportation costs, time off from work, and childcare availability. In addition, this study did not assess access for the uninsured, who are predominantly the working poor who cannot afford health insurance, even with federal and state subsidies.

The authors apologize for inconveniencing these offices, however, data collection could not be achieved in a better manner. We hope that the value of this study compensates any inconvenience.

CONCLUSION

Overall, our results demonstrate that despite the ratification of the PPACA, Medicaid patients are confronted with more barriers to accessing care by comparison to patients with Medicare and BlueCross insurance. Medicaid patients have worse baseline health²² and are at an increased risk of complications. These disparities are thought to be due to decreased healthcare access,^23,24 as well as socioeconomic challenges. Interventions, such as increasing Medicaid’s reimbursement levels, reducing burdensome administrative responsibilities, and establishing partnerships between trauma centers and trauma surgeons, may enable underinsured patients to be appropriately cared for.

This paper will be judged for the Resident Writer’s Award.

ABSTRACT

The purpose of this study is to assess the effect of insurance type (Medicaid, Medicare, private insurance) on the ability for patients with operative ankle fractures to access orthopedic traumatologists. The research team called 245 board-certified orthopedic surgeons specializing in orthopedic trauma within 8 representative states. The caller requested an appointment for their fictitious mother in order to be evaluated for an ankle fracture which was previously evaluated by her primary care physician and believed to require surgery. Each office was called 3 times to assess the response for each insurance type. For each call, information was documented regarding whether the patient was able to receive an appointment and the barriers the patient confronted to receive an appointment. Overall, 35.7% of offices scheduled an appointment for a patient with Medicaid, in comparison to 81.4%and 88.6% for Medicare and BlueCross, respectively (P < .0001). Medicaid patients confronted more barriers for receiving appointments. There was no statistically significant difference in access for Medicaid patients in states that had expanded Medicaid eligibility vs states that had not expanded Medicaid. Medicaid reimbursement for open reduction and internal fixation of an ankle fracture did not significantly correlate with appointment success rates or wait times. Despite the passage of the Affordable Care Act, patients with Medicaid have reduced access to orthopedic surgeons and more complex barriers to receiving appointments. A more robust strategy for increasing care-access for patients with Medicaid would be more equitable.

Continue to: In 2010, the Patient Protection and Affordable Care Act...

In 2010, the Patient Protection and Affordable Care Act (PPACA) expanded the eligibility criteria for Medicaid to all individuals with an income up to 138% of the poverty level.¹ A Supreme Court ruling stated that the decision to expand Medicaid was to be decided by individual states.² Currently, 31 states have chosen to expand Medicaid eligibility to their residents.² This expansion has allowed an additional 11.7 million people to enroll in Medicaid and the Children’s Health Insurance Program by May 2015.^3-5

Even with the passage of the PPACA, Medicaid patients seeking specialty orthopedic care have experienced more barriers to accessing care than Medicare or commercially-insured patients.^2,6-10 One major cited reason is Medicaid’s low reimbursement, which may discourage physicians from open panel participation in Medicaid.^11,12  

A common fundamental teaching for orthopedic traumatologists is the notion that they should be available to treat all injuries regardless of the patient’s ability to pay.¹³ This has resulted in both trauma centers and trauma surgeons becoming financially challenged due to the higher proportion of Medicaid and uninsured trauma patients and lower Medicaid reimbursement levels.^14,15

This study focuses on the effect of different types of insurance (Medicaid, Medicare, or commercial insurance) on the ability of patients to obtain care for operative ankle fractures. The purpose of this study is to evaluate, in the context of the PPACA, patient access to orthopedic surgeons for operative ankle fractures based on insurance-type. We hypothesized that patients with Medicaid would face a greater volume of obstacles when seeking appointments for an ankle fracture, even after the PPACA.

Continue to: MATERIALS AND METHODS...

MATERIALS AND METHODS

The study population included board-certified orthopedic surgeons who belonged to the Orthopaedic Trauma Association (OTA) from 8 representative states; 4 states with expanded Medicaid eligibility (California, Massachusetts, New York, Ohio) and 4 states without expanded Medicaid eligibility (Florida, North Carolina, Georgia, Texas). These states were selected due to their ability to represent diverse healthcare marketplaces throughout the country. Using the OTA website’s “Find a Surgeon” search tool,¹⁶ we created a list of surgeons for each state and matched each surgeon with a random number. The list of surgeons was ordered according to the value of the surgeon’s associated random number, and surgeons were called in ascending order. We excluded disconnected or inaccurate numbers from the calling list. Surgeons who did not manage ankle fractures were removed from the dataset. Approximately 30 orthopedic trauma surgeons per state were contacted.

Each office was called to make an appointment for the caller’s mother. Every surgeon’s office was specifically asked if the surgeon would accept the patient to be evaluated for an ankle fracture that occurred out-of-state. The caller had a standardized protocol to limit intra- and inter-office variations (Appendix). The scenario involved a request to be evaluated for an unstable ankle fracture, with the patient having Medicaid, Medicare, or BlueCross insurance. The scenario required 3 separate calls to the same surgeon in order to obtain data regarding each insurance-type. The calls were separated by at least 1 week to avoid caller recognition by the surgeon’s office.

Appendix

Scenario

1. Date of Birth: Medicaid–2/07/55; BlueCross PPO–2/09/55; Medicare–7/31/45.

2. Ankle fracture evaluated by primary care physician 1 or 2 days ago

3. Not seen previously by your clinic or hospital, she would be a new patient

4. Asked how early she could be scheduled for an appointment

5. Script:

“I’m calling for my mother who injured her ankle a few days ago. Her family doctor took an X-ray and believes she has a fracture and needs surgery. Is Dr. X accepting new patients for evaluation and treatment of ankle fractures?” If YES →

“I was wondering if you take Medicaid/Medicare/BlueCross plan?” If YES →

“When is your soonest available appointment?”

The date of each phone call and date of appointment, if provided, were recorded. If the office did not give an appointment, we asked for reasons why. If an appointment was denied for a patient with Medicaid, we asked for a referral to another office that accepted Medicaid. We considered barriers to obtaining an initial appointment, such as requiring a referral from a primary care physician (PCP), as an unsuccessful attempt at making an appointment. We determined the waiting period for an appointment by calculating the time between the date of the call and the date of the appointment. Appointments were not scheduled to ensure that actual patients were not disadvantaged. For both appointment success rates and waiting periods, we stratified the data into 2 groups: states with expanded Medicaid eligibility (California, Massachusetts, New York, Ohio) and states without expanded Medicaid eligibility (Florida, North Carolina, Georgia, Texas).

We obtained Medicaid reimbursement rates for open reduction and internal fixation of an ankle fracture by querying each state’s reimbursement rate using Current Procedural Terminology code 27822.

Chi-square test or Fisher’s exact test was used to analyze acceptance rate differences based on the patient’s type of insurance. To compare the waiting periods for an appointment, we used an independent samples t-test after applying natural log-transformation, as the data was not normally distributed. We performed logistic regression analysis to detect whether reimbursement was a significant predictor of successfully making an appointment for patients, and a linear regression analysis was used to evaluate whether reimbursement predicted waiting periods. Unless otherwise stated, all statistical testing was performed two-tailed at an alpha-level of 0.05.

This study was approved by the Institutional Review Board of Yale University School of Medicine (HIC No. 1363).

Continue to: RESULTS...

RESULTS

In total, 350 offices were contacted across 8 states (4 states with and 4 states without expanded Medicaid eligibility) of which we identified 245 orthopedic surgeons who would surgically treat ankle fractures. The 245 surgeons’ offices were called 3 times for each separate insurance-type.

Table 1. Appointment Success Rate

^aComparison to Medicaid.

The overall rate of successfully being offered an appointment with Medicaid was 35.7%, 81.4% for Medicare, and 88.6% for BlueCross (Table 1). For states with expanded Medicaid eligibility, the success rate for obtaining an appointment was 39.6%, 83.5%, and 89.2% for Medicaid, Medicare, and BlueCross, respectively. For states without expanded Medicaid eligibility, the success rate for obtaining an appointment was 31.9% for Medicaid, 79.4% for Medicare, and 87.9% for BlueCross. In all cases, the success rate for obtaining an appointment was significantly lower for Medicaid, compared to Medicare (P < .0001) or BlueCross (P < .0001). Medicaid appointment success rate was 39.6% in expanded states vs 31.9% in non-expanded states, however, the difference was not statistically significant (Table 2).

Table 2. Medicaid Appointment Success Rate in Expanded Vs Non-Expanded States

In 43.7% of occasions, patients with Medicaid did not have their insurance accepted, compared to 7.3% for Medicare and 0% for BlueCross. The majority of offices which did not accept Medicaid were not able to refer patients to another surgeon who would accept Medicaid. The requirement to have a primary care referral was the second most common reason for Medicaid patients not obtaining an appointment. No Medicare (10.4% vs 0.0%, P < .0001) or BlueCross (10.4% vs 0.0%, P < .0001) patients experienced this requirement (Table 3). There was no difference found between the percent of Medicaid patients who were required to have referrals in states with and without expanded Medicaid eligibility (Table 4).

Table 3. Referral Rate

^aComparison to Medicaid.

Table 4. Medicaid Referral Rates in Expanded Vs Non-Expanded States

Reimbursements for ankle fracture varied across states (Table 5). For Medicaid, Georgia paid the highest reimbursement ($1049.95) and Florida paid the lowest ($469.44). Logistic and linear regression analysis did not demonstrate a significant relationship between reimbursement and appointment success rate or waiting periods.

Table 5. Medicaid Reimbursements for Ankle Fracture Repair (CPT and HCPCS 27822) in 2014

^aStates with expanded Medicaid eligibility.

Abbreviations: CPT, Current Procedural Terminology; HCPCS, Healthcare Common Procedure Coding System.

Waiting periods (Table 6) varied significantly by the type of insurance (7.3 days for Medicaid, 6.0 days for Medicare, and 6.0 days for BlueCross; P = .002). For states with expanded Medicaid eligibility, waiting periods varied significantly by insurance (7.7 days for Medicaid, 6.2 days for Medicare, P = .003; and 6.1 days for BlueCross, P = .01). Waiting periods did not vary significantly for states without expanded Medicaid. Additionally, waiting periods did not differ significantly when comparing between states with and without Medicaid expansion.

Table 6. Waiting Period (Days) by Insurance Type.

Continue to: DISCUSSION...

DISCUSSION

This study assessed how insurance type (Medicaid, Medicare, and BlueCross) affects patient access to orthopedic trauma surgeons in 8 geographically representative states. We selected unstable ankle fractures as they are basic fractures treated by nearly all trauma surgeons and should often be surgically treated to prevent serious long-term consequences. Our hypothesis stated that despite the passage of the PPACA, patients with Medicaid would have reduced access to care. As the PPACA has changed the healthcare marketplace by increasing the number of Medicaid enrollees, it is important to ensure that patient access to care improves.

This nationwide survey of orthopedic trauma surgeons demonstrates that Medicaid patients experience added barriers to care that ultimately results in lower rates of successfully obtaining care. This is consistent with other investigations which have assessed Medicaid patient healthcare access.^6,8,10,17-19 This study did not demonstrate a statistically significant difference between Medicaid patients’ ability to obtain appointments in states with expanded Medicaid eligibility vs in states without expanded Medicaid eligibility (39.6% vs 31.9%, P < .18); this has been demonstrated in the literature.⁶  

A barrier that was unique to Medicaid patients was the requirement to have a PCP referral (Table 3). A PCP referral was not a barrier to receiving an appointment for patients with Medicare or BlueCross. One reason to explain why Medicaid patients may be required to have PCP referrals is due to their increased medical complexity, extra documentation requirements, and low reimbursement.⁴ Patients who have obtained a PCP referral may be characterized as being more medically compliant.

It is important to note that the Medicaid policies for 4 states included in this study (Massachusetts, North Carolina, Texas, and New York) required a PCP referral in order to see a specialist. However, we found that many orthopedic trauma practices in these states scheduled appointments for Medicaid patients without a PCP referral, suggesting that the decision depended on individual policy. In addition, the majority of offices within these states cited that they simply did not accept Medicaid as an insurance policy, and not that they required a referral.

Our regression analysis did not find a significant relationship between being able to successfully obtain an appointment to be evaluated for an ankle fracture and reimbursement rates for Medicaid. Although studies have stressed the importance of Medicaid reimbursements on physician participation, this result is consistent with previous studies regarding carpal tunnel release and total ankle replacements.^17,19 Long²⁰ suggested that although reimbursements may help, additional strategies for promoting Medicaid acceptance may be needed, including: lowering the costs of participating in Medicaid by simplifying administrative processes, speeding up reimbursement, and reducing the costs associated with caring for those patients.

Continue to: Previous studies have demonstrated...

Previous studies have demonstrated that more physicians may accept Medicaid if reimbursements increased.^4,12 Given the high percentage of trauma patients with Medicaid as their primary insurance or whom are emergently enrolled in Medicaid by hospital systems, it is concerning that the PPACA is reducing payments under the Medicare and Medicaid Disproportionate Share Hospital programs which provide hospitals for uncompensated care given to low-income and uninsured patients.²¹ Trauma centers generally operate at a deficit due to the higher proportion of Medicaid and uninsured patients.¹⁴ This is currently worsened by additional federal funding cuts for supporting trauma service’s humane mission.²¹

This study has several limitations. While the study evaluated access to care in 8 representative states, a thorough nationwide survey would be more representative. Some results may have become statistically significant if we had performed the study with a larger sample size. In addition, we were unable to control for many factors which could impact appointment wait times, such as physician call schedules and vacations. Socioeconomic factors can influence a patient’s ability to attend an appointment, such as transportation costs, time off from work, and childcare availability. In addition, this study did not assess access for the uninsured, who are predominantly the working poor who cannot afford health insurance, even with federal and state subsidies.

The authors apologize for inconveniencing these offices, however, data collection could not be achieved in a better manner. We hope that the value of this study compensates any inconvenience.

CONCLUSION

Overall, our results demonstrate that despite the ratification of the PPACA, Medicaid patients are confronted with more barriers to accessing care by comparison to patients with Medicare and BlueCross insurance. Medicaid patients have worse baseline health²² and are at an increased risk of complications. These disparities are thought to be due to decreased healthcare access,^23,24 as well as socioeconomic challenges. Interventions, such as increasing Medicaid’s reimbursement levels, reducing burdensome administrative responsibilities, and establishing partnerships between trauma centers and trauma surgeons, may enable underinsured patients to be appropriately cared for.

This paper will be judged for the Resident Writer’s Award.

ABSTRACT

The purpose of this study is to assess the effect of insurance type (Medicaid, Medicare, private insurance) on the ability for patients with operative ankle fractures to access orthopedic traumatologists. The research team called 245 board-certified orthopedic surgeons specializing in orthopedic trauma within 8 representative states. The caller requested an appointment for their fictitious mother in order to be evaluated for an ankle fracture which was previously evaluated by her primary care physician and believed to require surgery. Each office was called 3 times to assess the response for each insurance type. For each call, information was documented regarding whether the patient was able to receive an appointment and the barriers the patient confronted to receive an appointment. Overall, 35.7% of offices scheduled an appointment for a patient with Medicaid, in comparison to 81.4%and 88.6% for Medicare and BlueCross, respectively (P < .0001). Medicaid patients confronted more barriers for receiving appointments. There was no statistically significant difference in access for Medicaid patients in states that had expanded Medicaid eligibility vs states that had not expanded Medicaid. Medicaid reimbursement for open reduction and internal fixation of an ankle fracture did not significantly correlate with appointment success rates or wait times. Despite the passage of the Affordable Care Act, patients with Medicaid have reduced access to orthopedic surgeons and more complex barriers to receiving appointments. A more robust strategy for increasing care-access for patients with Medicaid would be more equitable.

Continue to: In 2010, the Patient Protection and Affordable Care Act...

In 2010, the Patient Protection and Affordable Care Act (PPACA) expanded the eligibility criteria for Medicaid to all individuals with an income up to 138% of the poverty level.¹ A Supreme Court ruling stated that the decision to expand Medicaid was to be decided by individual states.² Currently, 31 states have chosen to expand Medicaid eligibility to their residents.² This expansion has allowed an additional 11.7 million people to enroll in Medicaid and the Children’s Health Insurance Program by May 2015.^3-5

Even with the passage of the PPACA, Medicaid patients seeking specialty orthopedic care have experienced more barriers to accessing care than Medicare or commercially-insured patients.^2,6-10 One major cited reason is Medicaid’s low reimbursement, which may discourage physicians from open panel participation in Medicaid.^11,12  

A common fundamental teaching for orthopedic traumatologists is the notion that they should be available to treat all injuries regardless of the patient’s ability to pay.¹³ This has resulted in both trauma centers and trauma surgeons becoming financially challenged due to the higher proportion of Medicaid and uninsured trauma patients and lower Medicaid reimbursement levels.^14,15

This study focuses on the effect of different types of insurance (Medicaid, Medicare, or commercial insurance) on the ability of patients to obtain care for operative ankle fractures. The purpose of this study is to evaluate, in the context of the PPACA, patient access to orthopedic surgeons for operative ankle fractures based on insurance-type. We hypothesized that patients with Medicaid would face a greater volume of obstacles when seeking appointments for an ankle fracture, even after the PPACA.

Continue to: MATERIALS AND METHODS...

MATERIALS AND METHODS

The study population included board-certified orthopedic surgeons who belonged to the Orthopaedic Trauma Association (OTA) from 8 representative states; 4 states with expanded Medicaid eligibility (California, Massachusetts, New York, Ohio) and 4 states without expanded Medicaid eligibility (Florida, North Carolina, Georgia, Texas). These states were selected due to their ability to represent diverse healthcare marketplaces throughout the country. Using the OTA website’s “Find a Surgeon” search tool,¹⁶ we created a list of surgeons for each state and matched each surgeon with a random number. The list of surgeons was ordered according to the value of the surgeon’s associated random number, and surgeons were called in ascending order. We excluded disconnected or inaccurate numbers from the calling list. Surgeons who did not manage ankle fractures were removed from the dataset. Approximately 30 orthopedic trauma surgeons per state were contacted.

Each office was called to make an appointment for the caller’s mother. Every surgeon’s office was specifically asked if the surgeon would accept the patient to be evaluated for an ankle fracture that occurred out-of-state. The caller had a standardized protocol to limit intra- and inter-office variations (Appendix). The scenario involved a request to be evaluated for an unstable ankle fracture, with the patient having Medicaid, Medicare, or BlueCross insurance. The scenario required 3 separate calls to the same surgeon in order to obtain data regarding each insurance-type. The calls were separated by at least 1 week to avoid caller recognition by the surgeon’s office.

Appendix

Scenario

1. Date of Birth: Medicaid–2/07/55; BlueCross PPO–2/09/55; Medicare–7/31/45.

2. Ankle fracture evaluated by primary care physician 1 or 2 days ago

3. Not seen previously by your clinic or hospital, she would be a new patient

4. Asked how early she could be scheduled for an appointment

5. Script:

“I’m calling for my mother who injured her ankle a few days ago. Her family doctor took an X-ray and believes she has a fracture and needs surgery. Is Dr. X accepting new patients for evaluation and treatment of ankle fractures?” If YES →

“I was wondering if you take Medicaid/Medicare/BlueCross plan?” If YES →

“When is your soonest available appointment?”

The date of each phone call and date of appointment, if provided, were recorded. If the office did not give an appointment, we asked for reasons why. If an appointment was denied for a patient with Medicaid, we asked for a referral to another office that accepted Medicaid. We considered barriers to obtaining an initial appointment, such as requiring a referral from a primary care physician (PCP), as an unsuccessful attempt at making an appointment. We determined the waiting period for an appointment by calculating the time between the date of the call and the date of the appointment. Appointments were not scheduled to ensure that actual patients were not disadvantaged. For both appointment success rates and waiting periods, we stratified the data into 2 groups: states with expanded Medicaid eligibility (California, Massachusetts, New York, Ohio) and states without expanded Medicaid eligibility (Florida, North Carolina, Georgia, Texas).

We obtained Medicaid reimbursement rates for open reduction and internal fixation of an ankle fracture by querying each state’s reimbursement rate using Current Procedural Terminology code 27822.

Chi-square test or Fisher’s exact test was used to analyze acceptance rate differences based on the patient’s type of insurance. To compare the waiting periods for an appointment, we used an independent samples t-test after applying natural log-transformation, as the data was not normally distributed. We performed logistic regression analysis to detect whether reimbursement was a significant predictor of successfully making an appointment for patients, and a linear regression analysis was used to evaluate whether reimbursement predicted waiting periods. Unless otherwise stated, all statistical testing was performed two-tailed at an alpha-level of 0.05.

This study was approved by the Institutional Review Board of Yale University School of Medicine (HIC No. 1363).

Continue to: RESULTS...

RESULTS

In total, 350 offices were contacted across 8 states (4 states with and 4 states without expanded Medicaid eligibility) of which we identified 245 orthopedic surgeons who would surgically treat ankle fractures. The 245 surgeons’ offices were called 3 times for each separate insurance-type.

Table 1. Appointment Success Rate

^aComparison to Medicaid.

The overall rate of successfully being offered an appointment with Medicaid was 35.7%, 81.4% for Medicare, and 88.6% for BlueCross (Table 1). For states with expanded Medicaid eligibility, the success rate for obtaining an appointment was 39.6%, 83.5%, and 89.2% for Medicaid, Medicare, and BlueCross, respectively. For states without expanded Medicaid eligibility, the success rate for obtaining an appointment was 31.9% for Medicaid, 79.4% for Medicare, and 87.9% for BlueCross. In all cases, the success rate for obtaining an appointment was significantly lower for Medicaid, compared to Medicare (P < .0001) or BlueCross (P < .0001). Medicaid appointment success rate was 39.6% in expanded states vs 31.9% in non-expanded states, however, the difference was not statistically significant (Table 2).

Table 2. Medicaid Appointment Success Rate in Expanded Vs Non-Expanded States

In 43.7% of occasions, patients with Medicaid did not have their insurance accepted, compared to 7.3% for Medicare and 0% for BlueCross. The majority of offices which did not accept Medicaid were not able to refer patients to another surgeon who would accept Medicaid. The requirement to have a primary care referral was the second most common reason for Medicaid patients not obtaining an appointment. No Medicare (10.4% vs 0.0%, P < .0001) or BlueCross (10.4% vs 0.0%, P < .0001) patients experienced this requirement (Table 3). There was no difference found between the percent of Medicaid patients who were required to have referrals in states with and without expanded Medicaid eligibility (Table 4).

Table 3. Referral Rate

^aComparison to Medicaid.

Table 4. Medicaid Referral Rates in Expanded Vs Non-Expanded States

Reimbursements for ankle fracture varied across states (Table 5). For Medicaid, Georgia paid the highest reimbursement ($1049.95) and Florida paid the lowest ($469.44). Logistic and linear regression analysis did not demonstrate a significant relationship between reimbursement and appointment success rate or waiting periods.

Table 5. Medicaid Reimbursements for Ankle Fracture Repair (CPT and HCPCS 27822) in 2014

^aStates with expanded Medicaid eligibility.

Abbreviations: CPT, Current Procedural Terminology; HCPCS, Healthcare Common Procedure Coding System.

Waiting periods (Table 6) varied significantly by the type of insurance (7.3 days for Medicaid, 6.0 days for Medicare, and 6.0 days for BlueCross; P = .002). For states with expanded Medicaid eligibility, waiting periods varied significantly by insurance (7.7 days for Medicaid, 6.2 days for Medicare, P = .003; and 6.1 days for BlueCross, P = .01). Waiting periods did not vary significantly for states without expanded Medicaid. Additionally, waiting periods did not differ significantly when comparing between states with and without Medicaid expansion.

Table 6. Waiting Period (Days) by Insurance Type.

Continue to: DISCUSSION...

DISCUSSION

This study assessed how insurance type (Medicaid, Medicare, and BlueCross) affects patient access to orthopedic trauma surgeons in 8 geographically representative states. We selected unstable ankle fractures as they are basic fractures treated by nearly all trauma surgeons and should often be surgically treated to prevent serious long-term consequences. Our hypothesis stated that despite the passage of the PPACA, patients with Medicaid would have reduced access to care. As the PPACA has changed the healthcare marketplace by increasing the number of Medicaid enrollees, it is important to ensure that patient access to care improves.

This nationwide survey of orthopedic trauma surgeons demonstrates that Medicaid patients experience added barriers to care that ultimately results in lower rates of successfully obtaining care. This is consistent with other investigations which have assessed Medicaid patient healthcare access.^6,8,10,17-19 This study did not demonstrate a statistically significant difference between Medicaid patients’ ability to obtain appointments in states with expanded Medicaid eligibility vs in states without expanded Medicaid eligibility (39.6% vs 31.9%, P < .18); this has been demonstrated in the literature.⁶  

A barrier that was unique to Medicaid patients was the requirement to have a PCP referral (Table 3). A PCP referral was not a barrier to receiving an appointment for patients with Medicare or BlueCross. One reason to explain why Medicaid patients may be required to have PCP referrals is due to their increased medical complexity, extra documentation requirements, and low reimbursement.⁴ Patients who have obtained a PCP referral may be characterized as being more medically compliant.

It is important to note that the Medicaid policies for 4 states included in this study (Massachusetts, North Carolina, Texas, and New York) required a PCP referral in order to see a specialist. However, we found that many orthopedic trauma practices in these states scheduled appointments for Medicaid patients without a PCP referral, suggesting that the decision depended on individual policy. In addition, the majority of offices within these states cited that they simply did not accept Medicaid as an insurance policy, and not that they required a referral.

Our regression analysis did not find a significant relationship between being able to successfully obtain an appointment to be evaluated for an ankle fracture and reimbursement rates for Medicaid. Although studies have stressed the importance of Medicaid reimbursements on physician participation, this result is consistent with previous studies regarding carpal tunnel release and total ankle replacements.^17,19 Long²⁰ suggested that although reimbursements may help, additional strategies for promoting Medicaid acceptance may be needed, including: lowering the costs of participating in Medicaid by simplifying administrative processes, speeding up reimbursement, and reducing the costs associated with caring for those patients.

Continue to: Previous studies have demonstrated...

Previous studies have demonstrated that more physicians may accept Medicaid if reimbursements increased.^4,12 Given the high percentage of trauma patients with Medicaid as their primary insurance or whom are emergently enrolled in Medicaid by hospital systems, it is concerning that the PPACA is reducing payments under the Medicare and Medicaid Disproportionate Share Hospital programs which provide hospitals for uncompensated care given to low-income and uninsured patients.²¹ Trauma centers generally operate at a deficit due to the higher proportion of Medicaid and uninsured patients.¹⁴ This is currently worsened by additional federal funding cuts for supporting trauma service’s humane mission.²¹

This study has several limitations. While the study evaluated access to care in 8 representative states, a thorough nationwide survey would be more representative. Some results may have become statistically significant if we had performed the study with a larger sample size. In addition, we were unable to control for many factors which could impact appointment wait times, such as physician call schedules and vacations. Socioeconomic factors can influence a patient’s ability to attend an appointment, such as transportation costs, time off from work, and childcare availability. In addition, this study did not assess access for the uninsured, who are predominantly the working poor who cannot afford health insurance, even with federal and state subsidies.

The authors apologize for inconveniencing these offices, however, data collection could not be achieved in a better manner. We hope that the value of this study compensates any inconvenience.

CONCLUSION

Overall, our results demonstrate that despite the ratification of the PPACA, Medicaid patients are confronted with more barriers to accessing care by comparison to patients with Medicare and BlueCross insurance. Medicaid patients have worse baseline health²² and are at an increased risk of complications. These disparities are thought to be due to decreased healthcare access,^23,24 as well as socioeconomic challenges. Interventions, such as increasing Medicaid’s reimbursement levels, reducing burdensome administrative responsibilities, and establishing partnerships between trauma centers and trauma surgeons, may enable underinsured patients to be appropriately cared for.

This paper will be judged for the Resident Writer’s Award.

Care transitions between hospitals and skilled nursing facilities (SNFs) are a vulnerable time for patients. The current health care climate of decreasing hospital length of stay, readmission penalties, and increasing patient complexity has made hospital care transitions an important safety concern. Suboptimal transitions across clinical settings can result in adverse events, inadequately controlled comorbidities, deficient patient and caregiver preparation for discharge, medication errors, relocation stress, and overall increased morbidity and mortality.^1,2 Such care transitions also may generate unnecessary spending, including avoidable readmissions, emergency department utilization, and duplicative laboratory and imaging studies. Approximately 23% of patients admitted to SNFs are readmitted to acute care hospitals within 30 days, and these patients have increased mortality rates in risk-adjusted analyses. ^3,4

Compounding the magnitude of this risk and vulnerability is the significant growth in the number of patients discharged to SNFs over the past 30 years. In 2013, more than 20% of Medicare patients discharged from acute care hospitals were destined for SNFs.^5,6 Paradoxically, despite the increasing need for SNF providers, there is a shortage of clinicians with training in geriatrics or nursing home care.⁷ The result is a growing need to identify organizational systems to optimize physician practice in these settings, enhance quality of care, especially around transitions, and increase educational training opportunities in SNFs for future practitioners.

Many SNFs today are staffed by physicians and other licensed clinicians whose exclusive practice location is the nursing facility or possibly several such facilities. This prevailing model of care can isolate the physicians, depriving them of interaction with clinicians in other specialties, and can contribute to burnout.⁸ This model does not lend itself to academic scholarship, quality improvement (QI), and student or resident training, as each of these endeavors depends on interprofessional collaboration as well as access to an academic medical center with additional resources.⁹

Few studies have described innovative hospitalist rotation models from acute to subacute care. The Cleveland Clinic implemented the Connected Care model where hospital-employed physicians and advanced practice professionals integrated into postacute care and reduced the 30-day hospital readmission rate from SNFs from 28% to 22%.¹⁰ Goth and colleagues performed a comparative effectiveness trial between a postacute care hospitalist (PACH) model and a community-based physician model of nursing home care. They found that the institution of a PACH model in a nursing home was associated with a significant increase in laboratory costs, nonsignificant reduction in medication errors and pharmacy costs, and no improvement in fall rates.¹¹ The conclusion was that the PACH model may lead to greater clinician involvement and that the potential decrease in pharmacy costs and medications errors may offset the costs associated with additional laboratory testing. Overall, there has been a lack of studies on the impact of these hospitalist rotation models from acute to subacute care on educational programs, QI activities, and the interprofessional environment.

To achieve a system in which physicians in a SNF can excel in these areas, Veterans Affairs Boston Healthcare System (VABHS) adopted a staffing model in which academic hospitalist physicians rotate between the inpatient hospital and subacute settings. This report describes the model structure, the varying roles of the physicians, and early indicators of its positive effects on educational programs, QI activities, and the interprofessional environment.

Methods

The VABHS consists of a 159-bed acute care hospital in West Roxbury, Massachusetts; and a 110-bed SNF in Brockton, Massachusetts, with 3 units: a 65-bed transitional care unit (TCU), a 30-bed long-term care unit, and a 15-bed palliative care/hospice unit. The majority of patients admitted to the SNF are transferred from the acute care hospital in West Roxbury and other regional hospitals. Prior to 2015, the TCU was staffed with full-time clinicians who exclusively practiced in the SNF.

In the new staffing model, 6 hospitalist physicians divide their clinical time between the acute care hospital’s inpatient medical service and the TCU. The hospitalists come from varied backgrounds in terms of years in practice and advanced training (Table 1). 

The amount of nonclinical (protected) time and clinical time on the acute inpatient service and the TCU varies for each physician. For example, a physician serves as principal investigator for several major research grants and has a hospital-wide administrative leadership role; as a result, the principal investigator has fewer months of clinical responsibility. Physicians are expected to use the protected time for scholarship, educational program development and teaching, QI, and administrative responsibilities. The VABHS leadership determines the amount of protected time based on individualized benchmarks for research, education, and administrative responsibilities that follow VA national and local institutional guidelines. These metrics and time allocations are negotiated at the time of recruitment and then are reviewed annually.

The TCU also is staffed with 4 full-time clinicians (2 physicians and 2 physician assistants) who provide additional continuity of care. The new hospitalist staffing model only required an approximate 10% increase in TCU clinical staffing full-time equivalents. Patients and admissions are divided equally among clinicians on service (census per clinician 12-15 patients), with redistribution of patients at times of transition from clinical to nonclinical time. Blocks of clinical time are scheduled for greater than 2 weeks at a time to preserve continuity. In addition, the new staffing model allocates assignment of clinical responsibilities that allows for clinicians to take leave without resultant shortages in clinical coverage.

To facilitate communication among physicians serving in the acute inpatient facility and the TCU, leaders of both of these programs meet monthly and ad hoc to review the transitions of care between the 2 settings. The description of this model and its assessment have been reviewed and deemed exempt from oversight by the VA Boston Healthcare System Research and Development Committee.

Results

Since the implementation of this staffing model in 2015, the system has grown considerably in the breadth and depth of educational programming, QI, and systems redesign in the TCU and, more broadly, in the SNF. The TCU, which previously had limited training opportunities, has experienced marked expansion of educational offerings. It is now a site for core general medicine rotations for first-year psychiatry residents and physician assistant students. The TCU also has expanded as a clinical site for transitions-in-care internal medicine resident curricula and electives, as well as a clinical site for a geriatrics fellowship.

A hospitalist developed and implemented a 4-week interprofessional curriculum for all clinical trainees and students, which occurs continuously. The curriculum includes a monthly academic conference and 12 didactic lectures and is taught by 16 interprofessional faculty from the TCU and the Palliative Care/Hospice Unit, including medicine, geriatric and palliative care physicians, physician assistants, social workers, physical and occupational therapists, pharmacists, and a geriatric psychologist. The goal of the curriculum is to provide learners the knowledge, attitudes, and skills necessary to perform effective, efficient, and safe transfers between clinical settings as well as education in transitional care. In addition, using a team of interprofessional faculty, the curriculum develops the interprofessional competencies of teamwork and communication. The curriculum also has provided a significant opportunity for interprofessional collaboration among faculty who have volunteered their teaching time in the development and teaching of the curriculum, with potential for improved clinical staff knowledge of other disciplines.

Quality improvement and system redesign projects in care transitions also have expanded (Table 2). 

Early assessment indicates that the new staffing model is having positive effects on the clinical environment of the TCU. A survey was conducted of a convenience sample of all physicians, nurse managers, social workers, and other members of the clinical team in the TCU (N=24)(Table 3), with response categories ranging on a Likert scale from 1 (very negative) to 5 (very positive). 

Although not rigorously analyzed using qualitative research methods, comments from respondents have consistently indicated that this staffing model increases the transfer of clinical and logistical knowledge among staff members working in the acute inpatient facility and the TCU. 

Discussion

With greater numbers of increasingly complex patients transitioning from the hospital to SNF, health care systems need to expand the capacity of their skilled nursing systems, not only to provide clinical care, but also to support QI and medical education. The VABHS developed a physician staffing model with the goal of enriching physician practice and enhancing QI and educational opportunities in its SNF. The model offers an opportunity to improve transitions in care as physicians gain a greater knowledge of both the hospital and subacute clinical settings. This hospitalist rotation model may improve the knowledge necessary for caring for patients moving across care settings, as well as improve communication between settings. It also has served as a foundation for systematic innovation in QI and education at this institution. Clinical staff in the transitional care setting have reported positive effects of this model on clinical skills and patient care, educational opportunities, as well as a desire for replication in other health care systems.

The potential generalizability of this model requires careful consideration. The VABHS is a tertiary care integrated health care system, enabling physicians to work in multiple clinical settings. Other settings may not have the staffing or clinical volume to sustain such a model. In addition, this model may increase discontinuity in patient care as hospitalists move between acute and subacute settings and nonclinical roles. This loss of continuity may be a greater concern in the SNF setting, as the inpatient hospitalist model generally involves high provider turnover as shift work. Our survey included nurse managers, and not floor nurses due to survey administration limitations, and feedback may not have captured a comprehensive view from CLC staff. Moreover, some of the perceived positive impacts also may be related to professional and personal attributes of the physicians rather than the actual model of care. In addition, the survey response rate was 86%. However, the nature of the improvement work (focused on care transitions) and educational opportunities (interprofessional care) would likely not occur had the physicians been based in one clinical setting.

Other new physician staffing models have been designed to improve the continuity between the hospital, subacute, and outpatient settings. For example, the University of Chicago Comprehensive Care model pairs patients with trained hospitalists who provide both inpatient and outpatient care, thereby optimizing continuity between these settings.¹⁴ At CareMore Health System, high-risk patients also are paired with hospitalists, referred to as “extensivists,” who lead care teams that follow patients between settings and provide acute, postacute, and outpatient care.¹⁵ In these models, a single physician takes responsibility for the patient throughout transitions of care and through various care settings. Both models have shown reduction in hospital readmissions. One concern with such models is that the treatment teams need to coexist in the various settings of care, and the ability to impact and create systematic change within each environment is limited. This may limit QI, educational opportunities, and system level impact within each environment of care.

In comparison, the “transitionalist” model proposed here features hospitalist physicians rotating between the acute inpatient hospital and subacute care with dedicated time in each environment. This innovative organizational structure may enhance physician practice and enrich QI and educational opportunities in SNFs. Further evaluation will include the impact on quality metrics of patient care and patient satisfaction, as this model has the potential to influence quality, cost, and overall health outcomes.

Acknowledgments
We would like to thank Shivani Jindal, Matthew Russell, Matthew Ronan, Juman Hijab, Wei Shen, Sandra Vilbrun-Bruno, and Jack Earnshaw for their significant contributions to this staffing model. We would also like to thank Paul Conlin, Jay Orlander, and the leadership team of Veterans Affairs Boston Healthcare System for supporting this staffing model.

Care transitions between hospitals and skilled nursing facilities (SNFs) are a vulnerable time for patients. The current health care climate of decreasing hospital length of stay, readmission penalties, and increasing patient complexity has made hospital care transitions an important safety concern. Suboptimal transitions across clinical settings can result in adverse events, inadequately controlled comorbidities, deficient patient and caregiver preparation for discharge, medication errors, relocation stress, and overall increased morbidity and mortality.^1,2 Such care transitions also may generate unnecessary spending, including avoidable readmissions, emergency department utilization, and duplicative laboratory and imaging studies. Approximately 23% of patients admitted to SNFs are readmitted to acute care hospitals within 30 days, and these patients have increased mortality rates in risk-adjusted analyses. ^3,4

Compounding the magnitude of this risk and vulnerability is the significant growth in the number of patients discharged to SNFs over the past 30 years. In 2013, more than 20% of Medicare patients discharged from acute care hospitals were destined for SNFs.^5,6 Paradoxically, despite the increasing need for SNF providers, there is a shortage of clinicians with training in geriatrics or nursing home care.⁷ The result is a growing need to identify organizational systems to optimize physician practice in these settings, enhance quality of care, especially around transitions, and increase educational training opportunities in SNFs for future practitioners.

Many SNFs today are staffed by physicians and other licensed clinicians whose exclusive practice location is the nursing facility or possibly several such facilities. This prevailing model of care can isolate the physicians, depriving them of interaction with clinicians in other specialties, and can contribute to burnout.⁸ This model does not lend itself to academic scholarship, quality improvement (QI), and student or resident training, as each of these endeavors depends on interprofessional collaboration as well as access to an academic medical center with additional resources.⁹

Few studies have described innovative hospitalist rotation models from acute to subacute care. The Cleveland Clinic implemented the Connected Care model where hospital-employed physicians and advanced practice professionals integrated into postacute care and reduced the 30-day hospital readmission rate from SNFs from 28% to 22%.¹⁰ Goth and colleagues performed a comparative effectiveness trial between a postacute care hospitalist (PACH) model and a community-based physician model of nursing home care. They found that the institution of a PACH model in a nursing home was associated with a significant increase in laboratory costs, nonsignificant reduction in medication errors and pharmacy costs, and no improvement in fall rates.¹¹ The conclusion was that the PACH model may lead to greater clinician involvement and that the potential decrease in pharmacy costs and medications errors may offset the costs associated with additional laboratory testing. Overall, there has been a lack of studies on the impact of these hospitalist rotation models from acute to subacute care on educational programs, QI activities, and the interprofessional environment.

To achieve a system in which physicians in a SNF can excel in these areas, Veterans Affairs Boston Healthcare System (VABHS) adopted a staffing model in which academic hospitalist physicians rotate between the inpatient hospital and subacute settings. This report describes the model structure, the varying roles of the physicians, and early indicators of its positive effects on educational programs, QI activities, and the interprofessional environment.

Methods

The VABHS consists of a 159-bed acute care hospital in West Roxbury, Massachusetts; and a 110-bed SNF in Brockton, Massachusetts, with 3 units: a 65-bed transitional care unit (TCU), a 30-bed long-term care unit, and a 15-bed palliative care/hospice unit. The majority of patients admitted to the SNF are transferred from the acute care hospital in West Roxbury and other regional hospitals. Prior to 2015, the TCU was staffed with full-time clinicians who exclusively practiced in the SNF.

In the new staffing model, 6 hospitalist physicians divide their clinical time between the acute care hospital’s inpatient medical service and the TCU. The hospitalists come from varied backgrounds in terms of years in practice and advanced training (Table 1). 

The amount of nonclinical (protected) time and clinical time on the acute inpatient service and the TCU varies for each physician. For example, a physician serves as principal investigator for several major research grants and has a hospital-wide administrative leadership role; as a result, the principal investigator has fewer months of clinical responsibility. Physicians are expected to use the protected time for scholarship, educational program development and teaching, QI, and administrative responsibilities. The VABHS leadership determines the amount of protected time based on individualized benchmarks for research, education, and administrative responsibilities that follow VA national and local institutional guidelines. These metrics and time allocations are negotiated at the time of recruitment and then are reviewed annually.

The TCU also is staffed with 4 full-time clinicians (2 physicians and 2 physician assistants) who provide additional continuity of care. The new hospitalist staffing model only required an approximate 10% increase in TCU clinical staffing full-time equivalents. Patients and admissions are divided equally among clinicians on service (census per clinician 12-15 patients), with redistribution of patients at times of transition from clinical to nonclinical time. Blocks of clinical time are scheduled for greater than 2 weeks at a time to preserve continuity. In addition, the new staffing model allocates assignment of clinical responsibilities that allows for clinicians to take leave without resultant shortages in clinical coverage.

To facilitate communication among physicians serving in the acute inpatient facility and the TCU, leaders of both of these programs meet monthly and ad hoc to review the transitions of care between the 2 settings. The description of this model and its assessment have been reviewed and deemed exempt from oversight by the VA Boston Healthcare System Research and Development Committee.

Results

Since the implementation of this staffing model in 2015, the system has grown considerably in the breadth and depth of educational programming, QI, and systems redesign in the TCU and, more broadly, in the SNF. The TCU, which previously had limited training opportunities, has experienced marked expansion of educational offerings. It is now a site for core general medicine rotations for first-year psychiatry residents and physician assistant students. The TCU also has expanded as a clinical site for transitions-in-care internal medicine resident curricula and electives, as well as a clinical site for a geriatrics fellowship.

A hospitalist developed and implemented a 4-week interprofessional curriculum for all clinical trainees and students, which occurs continuously. The curriculum includes a monthly academic conference and 12 didactic lectures and is taught by 16 interprofessional faculty from the TCU and the Palliative Care/Hospice Unit, including medicine, geriatric and palliative care physicians, physician assistants, social workers, physical and occupational therapists, pharmacists, and a geriatric psychologist. The goal of the curriculum is to provide learners the knowledge, attitudes, and skills necessary to perform effective, efficient, and safe transfers between clinical settings as well as education in transitional care. In addition, using a team of interprofessional faculty, the curriculum develops the interprofessional competencies of teamwork and communication. The curriculum also has provided a significant opportunity for interprofessional collaboration among faculty who have volunteered their teaching time in the development and teaching of the curriculum, with potential for improved clinical staff knowledge of other disciplines.

Quality improvement and system redesign projects in care transitions also have expanded (Table 2). 

Early assessment indicates that the new staffing model is having positive effects on the clinical environment of the TCU. A survey was conducted of a convenience sample of all physicians, nurse managers, social workers, and other members of the clinical team in the TCU (N=24)(Table 3), with response categories ranging on a Likert scale from 1 (very negative) to 5 (very positive). 

Although not rigorously analyzed using qualitative research methods, comments from respondents have consistently indicated that this staffing model increases the transfer of clinical and logistical knowledge among staff members working in the acute inpatient facility and the TCU. 

Discussion

With greater numbers of increasingly complex patients transitioning from the hospital to SNF, health care systems need to expand the capacity of their skilled nursing systems, not only to provide clinical care, but also to support QI and medical education. The VABHS developed a physician staffing model with the goal of enriching physician practice and enhancing QI and educational opportunities in its SNF. The model offers an opportunity to improve transitions in care as physicians gain a greater knowledge of both the hospital and subacute clinical settings. This hospitalist rotation model may improve the knowledge necessary for caring for patients moving across care settings, as well as improve communication between settings. It also has served as a foundation for systematic innovation in QI and education at this institution. Clinical staff in the transitional care setting have reported positive effects of this model on clinical skills and patient care, educational opportunities, as well as a desire for replication in other health care systems.

The potential generalizability of this model requires careful consideration. The VABHS is a tertiary care integrated health care system, enabling physicians to work in multiple clinical settings. Other settings may not have the staffing or clinical volume to sustain such a model. In addition, this model may increase discontinuity in patient care as hospitalists move between acute and subacute settings and nonclinical roles. This loss of continuity may be a greater concern in the SNF setting, as the inpatient hospitalist model generally involves high provider turnover as shift work. Our survey included nurse managers, and not floor nurses due to survey administration limitations, and feedback may not have captured a comprehensive view from CLC staff. Moreover, some of the perceived positive impacts also may be related to professional and personal attributes of the physicians rather than the actual model of care. In addition, the survey response rate was 86%. However, the nature of the improvement work (focused on care transitions) and educational opportunities (interprofessional care) would likely not occur had the physicians been based in one clinical setting.

Other new physician staffing models have been designed to improve the continuity between the hospital, subacute, and outpatient settings. For example, the University of Chicago Comprehensive Care model pairs patients with trained hospitalists who provide both inpatient and outpatient care, thereby optimizing continuity between these settings.¹⁴ At CareMore Health System, high-risk patients also are paired with hospitalists, referred to as “extensivists,” who lead care teams that follow patients between settings and provide acute, postacute, and outpatient care.¹⁵ In these models, a single physician takes responsibility for the patient throughout transitions of care and through various care settings. Both models have shown reduction in hospital readmissions. One concern with such models is that the treatment teams need to coexist in the various settings of care, and the ability to impact and create systematic change within each environment is limited. This may limit QI, educational opportunities, and system level impact within each environment of care.

In comparison, the “transitionalist” model proposed here features hospitalist physicians rotating between the acute inpatient hospital and subacute care with dedicated time in each environment. This innovative organizational structure may enhance physician practice and enrich QI and educational opportunities in SNFs. Further evaluation will include the impact on quality metrics of patient care and patient satisfaction, as this model has the potential to influence quality, cost, and overall health outcomes.

Acknowledgments
We would like to thank Shivani Jindal, Matthew Russell, Matthew Ronan, Juman Hijab, Wei Shen, Sandra Vilbrun-Bruno, and Jack Earnshaw for their significant contributions to this staffing model. We would also like to thank Paul Conlin, Jay Orlander, and the leadership team of Veterans Affairs Boston Healthcare System for supporting this staffing model.

Care transitions between hospitals and skilled nursing facilities (SNFs) are a vulnerable time for patients. The current health care climate of decreasing hospital length of stay, readmission penalties, and increasing patient complexity has made hospital care transitions an important safety concern. Suboptimal transitions across clinical settings can result in adverse events, inadequately controlled comorbidities, deficient patient and caregiver preparation for discharge, medication errors, relocation stress, and overall increased morbidity and mortality.^1,2 Such care transitions also may generate unnecessary spending, including avoidable readmissions, emergency department utilization, and duplicative laboratory and imaging studies. Approximately 23% of patients admitted to SNFs are readmitted to acute care hospitals within 30 days, and these patients have increased mortality rates in risk-adjusted analyses. ^3,4

Compounding the magnitude of this risk and vulnerability is the significant growth in the number of patients discharged to SNFs over the past 30 years. In 2013, more than 20% of Medicare patients discharged from acute care hospitals were destined for SNFs.^5,6 Paradoxically, despite the increasing need for SNF providers, there is a shortage of clinicians with training in geriatrics or nursing home care.⁷ The result is a growing need to identify organizational systems to optimize physician practice in these settings, enhance quality of care, especially around transitions, and increase educational training opportunities in SNFs for future practitioners.

Many SNFs today are staffed by physicians and other licensed clinicians whose exclusive practice location is the nursing facility or possibly several such facilities. This prevailing model of care can isolate the physicians, depriving them of interaction with clinicians in other specialties, and can contribute to burnout.⁸ This model does not lend itself to academic scholarship, quality improvement (QI), and student or resident training, as each of these endeavors depends on interprofessional collaboration as well as access to an academic medical center with additional resources.⁹

Few studies have described innovative hospitalist rotation models from acute to subacute care. The Cleveland Clinic implemented the Connected Care model where hospital-employed physicians and advanced practice professionals integrated into postacute care and reduced the 30-day hospital readmission rate from SNFs from 28% to 22%.¹⁰ Goth and colleagues performed a comparative effectiveness trial between a postacute care hospitalist (PACH) model and a community-based physician model of nursing home care. They found that the institution of a PACH model in a nursing home was associated with a significant increase in laboratory costs, nonsignificant reduction in medication errors and pharmacy costs, and no improvement in fall rates.¹¹ The conclusion was that the PACH model may lead to greater clinician involvement and that the potential decrease in pharmacy costs and medications errors may offset the costs associated with additional laboratory testing. Overall, there has been a lack of studies on the impact of these hospitalist rotation models from acute to subacute care on educational programs, QI activities, and the interprofessional environment.

To achieve a system in which physicians in a SNF can excel in these areas, Veterans Affairs Boston Healthcare System (VABHS) adopted a staffing model in which academic hospitalist physicians rotate between the inpatient hospital and subacute settings. This report describes the model structure, the varying roles of the physicians, and early indicators of its positive effects on educational programs, QI activities, and the interprofessional environment.

Methods

The VABHS consists of a 159-bed acute care hospital in West Roxbury, Massachusetts; and a 110-bed SNF in Brockton, Massachusetts, with 3 units: a 65-bed transitional care unit (TCU), a 30-bed long-term care unit, and a 15-bed palliative care/hospice unit. The majority of patients admitted to the SNF are transferred from the acute care hospital in West Roxbury and other regional hospitals. Prior to 2015, the TCU was staffed with full-time clinicians who exclusively practiced in the SNF.

In the new staffing model, 6 hospitalist physicians divide their clinical time between the acute care hospital’s inpatient medical service and the TCU. The hospitalists come from varied backgrounds in terms of years in practice and advanced training (Table 1). 

The amount of nonclinical (protected) time and clinical time on the acute inpatient service and the TCU varies for each physician. For example, a physician serves as principal investigator for several major research grants and has a hospital-wide administrative leadership role; as a result, the principal investigator has fewer months of clinical responsibility. Physicians are expected to use the protected time for scholarship, educational program development and teaching, QI, and administrative responsibilities. The VABHS leadership determines the amount of protected time based on individualized benchmarks for research, education, and administrative responsibilities that follow VA national and local institutional guidelines. These metrics and time allocations are negotiated at the time of recruitment and then are reviewed annually.

The TCU also is staffed with 4 full-time clinicians (2 physicians and 2 physician assistants) who provide additional continuity of care. The new hospitalist staffing model only required an approximate 10% increase in TCU clinical staffing full-time equivalents. Patients and admissions are divided equally among clinicians on service (census per clinician 12-15 patients), with redistribution of patients at times of transition from clinical to nonclinical time. Blocks of clinical time are scheduled for greater than 2 weeks at a time to preserve continuity. In addition, the new staffing model allocates assignment of clinical responsibilities that allows for clinicians to take leave without resultant shortages in clinical coverage.

To facilitate communication among physicians serving in the acute inpatient facility and the TCU, leaders of both of these programs meet monthly and ad hoc to review the transitions of care between the 2 settings. The description of this model and its assessment have been reviewed and deemed exempt from oversight by the VA Boston Healthcare System Research and Development Committee.

Results

Since the implementation of this staffing model in 2015, the system has grown considerably in the breadth and depth of educational programming, QI, and systems redesign in the TCU and, more broadly, in the SNF. The TCU, which previously had limited training opportunities, has experienced marked expansion of educational offerings. It is now a site for core general medicine rotations for first-year psychiatry residents and physician assistant students. The TCU also has expanded as a clinical site for transitions-in-care internal medicine resident curricula and electives, as well as a clinical site for a geriatrics fellowship.

A hospitalist developed and implemented a 4-week interprofessional curriculum for all clinical trainees and students, which occurs continuously. The curriculum includes a monthly academic conference and 12 didactic lectures and is taught by 16 interprofessional faculty from the TCU and the Palliative Care/Hospice Unit, including medicine, geriatric and palliative care physicians, physician assistants, social workers, physical and occupational therapists, pharmacists, and a geriatric psychologist. The goal of the curriculum is to provide learners the knowledge, attitudes, and skills necessary to perform effective, efficient, and safe transfers between clinical settings as well as education in transitional care. In addition, using a team of interprofessional faculty, the curriculum develops the interprofessional competencies of teamwork and communication. The curriculum also has provided a significant opportunity for interprofessional collaboration among faculty who have volunteered their teaching time in the development and teaching of the curriculum, with potential for improved clinical staff knowledge of other disciplines.

Quality improvement and system redesign projects in care transitions also have expanded (Table 2). 

Early assessment indicates that the new staffing model is having positive effects on the clinical environment of the TCU. A survey was conducted of a convenience sample of all physicians, nurse managers, social workers, and other members of the clinical team in the TCU (N=24)(Table 3), with response categories ranging on a Likert scale from 1 (very negative) to 5 (very positive). 

Although not rigorously analyzed using qualitative research methods, comments from respondents have consistently indicated that this staffing model increases the transfer of clinical and logistical knowledge among staff members working in the acute inpatient facility and the TCU. 

Discussion

With greater numbers of increasingly complex patients transitioning from the hospital to SNF, health care systems need to expand the capacity of their skilled nursing systems, not only to provide clinical care, but also to support QI and medical education. The VABHS developed a physician staffing model with the goal of enriching physician practice and enhancing QI and educational opportunities in its SNF. The model offers an opportunity to improve transitions in care as physicians gain a greater knowledge of both the hospital and subacute clinical settings. This hospitalist rotation model may improve the knowledge necessary for caring for patients moving across care settings, as well as improve communication between settings. It also has served as a foundation for systematic innovation in QI and education at this institution. Clinical staff in the transitional care setting have reported positive effects of this model on clinical skills and patient care, educational opportunities, as well as a desire for replication in other health care systems.

The potential generalizability of this model requires careful consideration. The VABHS is a tertiary care integrated health care system, enabling physicians to work in multiple clinical settings. Other settings may not have the staffing or clinical volume to sustain such a model. In addition, this model may increase discontinuity in patient care as hospitalists move between acute and subacute settings and nonclinical roles. This loss of continuity may be a greater concern in the SNF setting, as the inpatient hospitalist model generally involves high provider turnover as shift work. Our survey included nurse managers, and not floor nurses due to survey administration limitations, and feedback may not have captured a comprehensive view from CLC staff. Moreover, some of the perceived positive impacts also may be related to professional and personal attributes of the physicians rather than the actual model of care. In addition, the survey response rate was 86%. However, the nature of the improvement work (focused on care transitions) and educational opportunities (interprofessional care) would likely not occur had the physicians been based in one clinical setting.

Other new physician staffing models have been designed to improve the continuity between the hospital, subacute, and outpatient settings. For example, the University of Chicago Comprehensive Care model pairs patients with trained hospitalists who provide both inpatient and outpatient care, thereby optimizing continuity between these settings.¹⁴ At CareMore Health System, high-risk patients also are paired with hospitalists, referred to as “extensivists,” who lead care teams that follow patients between settings and provide acute, postacute, and outpatient care.¹⁵ In these models, a single physician takes responsibility for the patient throughout transitions of care and through various care settings. Both models have shown reduction in hospital readmissions. One concern with such models is that the treatment teams need to coexist in the various settings of care, and the ability to impact and create systematic change within each environment is limited. This may limit QI, educational opportunities, and system level impact within each environment of care.

In comparison, the “transitionalist” model proposed here features hospitalist physicians rotating between the acute inpatient hospital and subacute care with dedicated time in each environment. This innovative organizational structure may enhance physician practice and enrich QI and educational opportunities in SNFs. Further evaluation will include the impact on quality metrics of patient care and patient satisfaction, as this model has the potential to influence quality, cost, and overall health outcomes.

Acknowledgments
We would like to thank Shivani Jindal, Matthew Russell, Matthew Ronan, Juman Hijab, Wei Shen, Sandra Vilbrun-Bruno, and Jack Earnshaw for their significant contributions to this staffing model. We would also like to thank Paul Conlin, Jay Orlander, and the leadership team of Veterans Affairs Boston Healthcare System for supporting this staffing model.

ABSTRACT

The morbidity and mortality after hip fracture in the elderly are influenced by non-modifiable comorbidities. Time-to-surgery is a modifiable factor that may play a role in postoperative morbidity. This study investigates the outcomes and complications in the elderly hip fracture surgery as a function of time-to-surgery.

Using the American College of Surgeons-National Surgical Quality Improvement Program data from 2011 to 2012, a study population was generated using the Current Procedural Terminology codes for percutaneous or open treatment of femoral neck fractures (27235, 27236) and fixation with a screw and side plate or intramedullary fixation (27244, 27245) for peritrochanteric fractures. Three time-to-surgery groups (<24 hours to surgical intervention, 24-48 hours, and >48 hours) were created and matched for surgery type, sex, age, and American Society of Anesthesiologists class. Time-to-surgery was then studied for its effect on the post-surgical outcomes using the adjusted regression modeling.

A study population of 6036 hip fractures was created, and 2012 patients were assigned to each matched time-to-surgery group. The unadjusted models showed that the earlier surgical intervention groups (<24 hours and 24-48 hours) exhibited a lower overall complication rate (P = .034) compared with the group waiting for surgery >48 hours. The unadjusted mortality rates increased with delay to surgical intervention (P = .039). Time-to-surgery caused no effect on the return to the operating room rate (P = .554) nor readmission rate (P = .285). Compared with other time-to-surgeries, the time-to-surgery of >48 hours was associated with prolonged total hospital length of stay (10.9 days) (P < .001) and a longer surgery-to-discharge time (hazard ratio, 95% confidence interval: 0.74, 0.69-0.79) (P < .001). Adjusted analyses showed no time-to-surgery related difference in complications (P = .143) but presented an increase in the total length of stay (P < .001) and surgery-to-discharge time (P < .001).

Timeliness of surgical intervention in a comorbidity-adjusted population of elderly hip fracture patients causes no effect on the overall complications, readmissions, nor 30-day mortality. However, time-to-surgery of >48 hours is associated with costly increase in the total length of stay, including an increased post-surgery-to-discharge time.

Continue to: Despite the best efforts to optimize surgical care...

Despite the best efforts to optimize surgical care and postoperative rehabilitation following hip fracture, elderly patients feature alarmingly high in-hospital and 1-year mortality rates of 4.35% to 9.2%^1-4 and 36%,⁵ respectively. Those who survive are unlikely to return to independent living, with only 17% of the patients following hip fracture being able to walk independently 6 months postoperatively, and 12% being able to climb stairs⁶. Possibly, these poor outcomes reflect a preoperative medical comorbidity burden rather than a measure of medical or surgical quality. Given the absence of consensus regarding optimal time-to-surgery, treating physicians often opt to delay surgical intervention for the purposes of medically optimizing highly comorbid patients without significant data to suggest clinical benefit of such practice.

Numerous investigators have attempted to identify the modifiable risk factors for complication after surgical care of elderly hip fracture patients. However, consensus guidelines of care are missing. This condition is largely due to the difficulties in effectively modifying preoperative demographic and medical comorbidities on a semi-urgent basis. However, timeliness to surgery is one area for study that the care team can affect. Although time-to-surgery is dependent on multiple factors, including time of presentation, day of week of admission, difficulties with scheduling, and administrative delays, the care team plays a role in hastening or retarding time-to-surgery. Several studies have considered various time cut-offs (24, 48, 72, and 120 hours) to define early intervention, but none have defined a specific role for early or delayed surgery. Several investigators have discovered a positive association between delayed time-to-surgery and mortality;^4,8-14 however, the most rigorously conducted studies that stringently control for preoperative comorbidities and demographics conclude that variance in time-to-surgery causes no effect on the in-hospital or 1-year mortality risk.^1-3,15-18  

Other investigators have shown that with early surgical intervention for hip fracture, patients experience shorter hospital stays,^{1,3,16,17,19-22} less days in pain,¹⁹ decreased risk of decubitus ulcers,^15,17,19,22 and an increased likelihood of independence following fracture,^22-25 regardless of preoperative medical status. Despite this evidence of improved outcomes with early surgery, 40% to 54% of hip fracture patients in the United States experience surgical delays of more than 24 to 48 hours. Additionally, with the recent (2013) national estimates of cost per day spent in the hospital falling between $1791 to $2289,²⁶ minimizing the days spent in the hospital would likely lead to significant cost-savings, presuming no adverse effect on health-related outcomes. To this end, we hypothesize that the value (outcomes per associated cost)⁷ of hip fracture surgical care can be positively influenced by minimizing surgical wait-times. We assessed the effect of early surgical intervention, within 24 or 48 hours of presentation, on 30-day mortality, postoperative morbidity, hospital length of stay, and readmission rates in a comorbidity-adjusted population from a nationally representative cohort.

Continue to: METHODS AND MATERIALS...

METHODS AND MATERIALS 

This study used the data from the American College of Surgeon-National Surgical Quality Improvement Program (ACS-NSQIP) database. With over 258 participating hospitals, this database has been widely used to identify national trends in various surgical specialties.^27-34 The database includes information from participants in 43 states with hospitals ranging from rural community hospitals to large academic centers. Each site employs surgical clinical reviewers who are rigorously trained to collect data through chart review and discussion with the treating surgeon and/or patient,³⁵ allowing for the use of robust and quality data with proven inter-rater reliability.^36,37

Using the 2011 to 2012 NSQIP database, we used primary Current Procedural Terminology codes to identify all patients who underwent percutaneous (27235) or open (27236) fixation of femoral neck fractures; and fixation with a screw and side plate (27244) or intramedullary fixation (27245) for peritrochanteric fractures. The sample was divided into 3 time-to-surgery groups (<24 hours from presentation to surgery, 24-48 hours, and >48 hours) which were matched for fracture type (femoral neck or peritrochanteric), sex, age (under 75 years or ≥75 years), and American Society of Anesthesiologists (ASA) class used as a surrogate for severity of medical infirmary. The subjects were randomly matched 1:1:1 to create 3 statistically equivalent time-to-surgery groups using Proc SurveySelect (SAS version 9.2, SAS Institute).

Generalized linear models using logit link function for binary variables and identity link function for normally distributed characteristics were used to compare the 3 time-to-surgery groups. Descriptive statistics are presented as counts and percentages or least-square means with standard deviations. Preoperative lab values that were not normally distributed were log transformed and presented in their original scales with median values and 25th to 75th percentiles. Outcomes were similarly modeled.  

Total hospital stay was modeled with a negative binomial distribution. Proportional hazards models were used to model the time from operating room (OR) to discharge, censoring patients who died before discharge, with results presented as hazard ratios (HR) and 95% confidence intervals (CI) (Figure). The assumption of the proportional hazards was tested using a Wald test. Using this model, a HR of <1 denotes a longer postoperative hospital stay, as a longer hospital stay decreases the “risk” for discharge.

All models were adjusted for confounders, including race, body mass index (BMI), hypertension, chronic obstructive pulmonary disease, cancer, bleeding disorders, transfusion within 72 hours before surgery, preoperative levels of creatinine, platelet count, white blood cells (WBCs), hematocrit anesthesia type, and wound infection. These covariates were selected based upon their observed relationship to the studied outcomes and time-to-surgery groups, and were evaluated across the models for all outcomes for consistency and clarity. All statistical analyses were run at a type I error rate of 5% and performed in SAS version 9.2 software.

Continue to: RESULTS...

RESULTS

A study population of 6036 hip fractures was identified and divided into 3 groups of 2012 subjects each based upon time-to-surgery. The groups were successfully matched for surgery type, age (≥75 years old), gender, and ASA class. In each group, 594 of the 2012 (29.5%) patients were male, 1525 (75.8%) were ≥75 years of age, 9 (.5%) were ASA Class I, 269 (13.4%) were ASA Class II, 1424 (70.8%) were ASA class III, and 309 (15.4%) were ASA class IV.

Significant differences in preoperative comorbidity burden and preoperative lab values were identified between the 3 cohorts. Increased time-to-surgery was associated with differences in race (P < .001), elevated BMI (P = .010), higher rates of congestive heart failure (P < .001), hypertension medication (P = .020), bleeding disorders (P < .001), blood transfusion within 72 hours of surgery (P < .001), and systemic sepsis (P = .001). Delay to surgery was also associated with lower preoperative sodium (P = .005), blood urea nitrogen (P = .013), serum WBC (P < .001), hematocrit (P < .001), and platelets (P < .001) (Table 1).

The unadjusted analyses revealed no association between time-to-surgery and return to OR (P = .554) nor readmission (P = .285). However, increasing time-to-surgery was associated with an increase in overall complications (P = .034), total length of hospital stay (P < .001), and 30-day mortality (P = .039) (Table 2).

Table 2. Estimated Event Rates from Matched Cohorts (Unadjusted)

Abbreviation: OR, operating room.

The adjusted analysis controlling for preoperative demographic and comorbidity variables revealed trends toward the increased overall complications and 30-day mortality with increased time-to-surgery; these trends showed no statistical significance (P = .143 and P = .08). No statistical relationship was observed between return to OR nor readmission and time-to-surgery. Increasing time-to-surgery remained significantly associated with the increased total length of hospital stay (P < .001). The adjusted analysis also revealed that the delay of >48 hours in time-to-surgery resulted in a longer surgery-to-discharge time (P < .001) (Table 3). No evidence of violation of the proportional hazards assumption was observed in the unadjusted nor adjusted clustered proportional hazards models (Wald test, P = .27 and P = .25, respectively).

Table 3. Estimated Event Rates from Matched Cohorts (Adjusted^a)

^aModel adjusted for race, hypertension medication, cancer, bleeding disorders, transfusion within 72 hours before surgery, emergency status, wound infection, anesthesia type (general), body mass index (18.5-25), history of chronic obstructive pulmonary disease, and preoperative levels of creatinine, platelet count, white blood cell count, and hematocrit.

Continue to: DISCUSSION...

DISCUSSION

Previous research has demonstrated an association between age,^3,4,25 comorbidity burden,^1,3,25 gender,^3,4 and ASA class^4,18,21 with outcomes following hip fractures and serves as the basis of our matched analysis statistical methodology in assessing the effect of time-to-surgery on the outcome following hip fracture surgery. Prior investigators have also established the positive correlation between increased preoperative comorbidity burden and delay in time-to-surgery.^10,15 This finding was confirmed in our unadjusted comparison of 3 time-to-surgery groups. However, prior investigations have not established a clear association between time-to-surgical intervention and postoperative morbidity and mortality.^{1,15,16,18,20,38} This study utilized a nationally representative dataset known for its data integrity and from which 6036 patients with surgically treated hip fractures, matched for surgery type, age, gender, and ASA class (a surrogate for severity of medical infirmary), were studied using adjusted regression modeling to afford an isolated statistical assessment of the effect of time-to-surgery on outcomes following hip fracture surgery.

Despite a large sample size and rigorous statistical methodology, for many outcome measures, our results show no support for the early or late operative intervention following hip fracture. We found no difference in 30-day mortality, readmission rate, nor total complication rate between the 3 time-to-surgery cohorts. This result indicates that the care of elderly patients following hip fracture is inherently complicated and that perioperative complication risk is probably only modestly modifiable by best medical practices, including optimizing time from clinical presentation to surgery.

As expected, patients who experienced longer delays from presentation to surgery were on average, more comorbid and more likely to yield abnormal preoperative lab values. However, in the adjusted analysis, delay in time-to-surgery, presumably for medical management, was not found to be associated with improved outcomes. In the same adjusted analysis, we uniquely identified that in the patients whose surgeries were delayed for more than 48 hours, the time from surgery-to-discharge was significantly increased. As a result, these patients spent extra days in the hospital both preoperatively and postoperatively, but without any corollary improvement in the outcomes.

Continue to: Recent estimates of the cost of hospital admission...

Recent estimates of the cost of hospital admission is approximated nationally at $2000/day.²⁶ Although our data fail to support the formal cost-analysis of the effect of time-to-surgery in hip fracture care, a simple value-based analysis indicates that quality is preserved (no difference in outcome), whereas costly hospital days are eliminated with earlier surgery. The value in elderly hip fracture care. defined as the outcomes relative to the costs,⁷ is ultimately optimized by earlier time-to-surgery.

Although using a large, multi-institutional database is advantageous for finding population-based trends that are representative of a large cohort, using the ACS-NSQIP database features its limitations. Our analysis was limited to the defined scope of NSQIP and nature of the injury, whereas root cause for delay was not available for study. We were unable to identify which patients were delayed for administrative reasons or surgical convenience and which were delayed for medical optimization. Participation in the ACS-NSQIP database is voluntary, and no randomized hospital sampling was conducted. Participating hospitals were de-identified in the database. As expected, we were unable to identify the specific institution-based hip fracture protocols that may affect the outcomes following treatment for these fractures. Further, socioeconomic information and payer-status are unavailable for the study. Additionally, observations are limited to 30 days postoperative, and we cannot comment on longer-term outcomes. Finally, discharge disposition and functional outcome data are not represented, and we were unable to correlate time-to-surgery and functional recovery. However, previous studies have established that delay in time-to-surgery following hip fractures is negatively correlated with functional outcomes.^22-25

Nevertheless, the ACS-NSQIP database remains one of the largest American surgical databases available, and includes care centers from nearly every state with variable demographics including rural, urban, and academic centers. The ACS performs broad-based inter-rater reliability audits on every participating site and has found an overall disagreement rate of only 1.8%. As such, although discrepancies exist between the complete patient chart and the data entered, the data found in the ACS-NSQIP database are reliable and considered a valid source of study.^34,35 The large sample size, quality of data collection, wide geographic representation, and varied hospital types within the dataset possibly make our findings relevant in the majority of American healthcare settings.

CONCLUSION

This study demonstrates an associated increased length of hospital stay, including the increased time from surgery-to-discharge, in patients with hip fractures whose surgical intervention is delayed for >48 hours after presentation. Given the prior evidence that early surgical intervention improves the functional outcomes and the current evidence that surgical delay for any cause increases costly hospital length of stay without corollary improvement in the outcomes, a value-based assessment of hip fracture care argues for early surgical intervention whenever possible. Our findings should inform physician, institution, and policy maker value-based decision making regarding the best practices in geriatric hip fracture care.

ABSTRACT

The morbidity and mortality after hip fracture in the elderly are influenced by non-modifiable comorbidities. Time-to-surgery is a modifiable factor that may play a role in postoperative morbidity. This study investigates the outcomes and complications in the elderly hip fracture surgery as a function of time-to-surgery.

Using the American College of Surgeons-National Surgical Quality Improvement Program data from 2011 to 2012, a study population was generated using the Current Procedural Terminology codes for percutaneous or open treatment of femoral neck fractures (27235, 27236) and fixation with a screw and side plate or intramedullary fixation (27244, 27245) for peritrochanteric fractures. Three time-to-surgery groups (<24 hours to surgical intervention, 24-48 hours, and >48 hours) were created and matched for surgery type, sex, age, and American Society of Anesthesiologists class. Time-to-surgery was then studied for its effect on the post-surgical outcomes using the adjusted regression modeling.

A study population of 6036 hip fractures was created, and 2012 patients were assigned to each matched time-to-surgery group. The unadjusted models showed that the earlier surgical intervention groups (<24 hours and 24-48 hours) exhibited a lower overall complication rate (P = .034) compared with the group waiting for surgery >48 hours. The unadjusted mortality rates increased with delay to surgical intervention (P = .039). Time-to-surgery caused no effect on the return to the operating room rate (P = .554) nor readmission rate (P = .285). Compared with other time-to-surgeries, the time-to-surgery of >48 hours was associated with prolonged total hospital length of stay (10.9 days) (P < .001) and a longer surgery-to-discharge time (hazard ratio, 95% confidence interval: 0.74, 0.69-0.79) (P < .001). Adjusted analyses showed no time-to-surgery related difference in complications (P = .143) but presented an increase in the total length of stay (P < .001) and surgery-to-discharge time (P < .001).

Timeliness of surgical intervention in a comorbidity-adjusted population of elderly hip fracture patients causes no effect on the overall complications, readmissions, nor 30-day mortality. However, time-to-surgery of >48 hours is associated with costly increase in the total length of stay, including an increased post-surgery-to-discharge time.

Continue to: Despite the best efforts to optimize surgical care...

Despite the best efforts to optimize surgical care and postoperative rehabilitation following hip fracture, elderly patients feature alarmingly high in-hospital and 1-year mortality rates of 4.35% to 9.2%^1-4 and 36%,⁵ respectively. Those who survive are unlikely to return to independent living, with only 17% of the patients following hip fracture being able to walk independently 6 months postoperatively, and 12% being able to climb stairs⁶. Possibly, these poor outcomes reflect a preoperative medical comorbidity burden rather than a measure of medical or surgical quality. Given the absence of consensus regarding optimal time-to-surgery, treating physicians often opt to delay surgical intervention for the purposes of medically optimizing highly comorbid patients without significant data to suggest clinical benefit of such practice.

Numerous investigators have attempted to identify the modifiable risk factors for complication after surgical care of elderly hip fracture patients. However, consensus guidelines of care are missing. This condition is largely due to the difficulties in effectively modifying preoperative demographic and medical comorbidities on a semi-urgent basis. However, timeliness to surgery is one area for study that the care team can affect. Although time-to-surgery is dependent on multiple factors, including time of presentation, day of week of admission, difficulties with scheduling, and administrative delays, the care team plays a role in hastening or retarding time-to-surgery. Several studies have considered various time cut-offs (24, 48, 72, and 120 hours) to define early intervention, but none have defined a specific role for early or delayed surgery. Several investigators have discovered a positive association between delayed time-to-surgery and mortality;^4,8-14 however, the most rigorously conducted studies that stringently control for preoperative comorbidities and demographics conclude that variance in time-to-surgery causes no effect on the in-hospital or 1-year mortality risk.^1-3,15-18  

Other investigators have shown that with early surgical intervention for hip fracture, patients experience shorter hospital stays,^{1,3,16,17,19-22} less days in pain,¹⁹ decreased risk of decubitus ulcers,^15,17,19,22 and an increased likelihood of independence following fracture,^22-25 regardless of preoperative medical status. Despite this evidence of improved outcomes with early surgery, 40% to 54% of hip fracture patients in the United States experience surgical delays of more than 24 to 48 hours. Additionally, with the recent (2013) national estimates of cost per day spent in the hospital falling between $1791 to $2289,²⁶ minimizing the days spent in the hospital would likely lead to significant cost-savings, presuming no adverse effect on health-related outcomes. To this end, we hypothesize that the value (outcomes per associated cost)⁷ of hip fracture surgical care can be positively influenced by minimizing surgical wait-times. We assessed the effect of early surgical intervention, within 24 or 48 hours of presentation, on 30-day mortality, postoperative morbidity, hospital length of stay, and readmission rates in a comorbidity-adjusted population from a nationally representative cohort.

Continue to: METHODS AND MATERIALS...

METHODS AND MATERIALS 

This study used the data from the American College of Surgeon-National Surgical Quality Improvement Program (ACS-NSQIP) database. With over 258 participating hospitals, this database has been widely used to identify national trends in various surgical specialties.^27-34 The database includes information from participants in 43 states with hospitals ranging from rural community hospitals to large academic centers. Each site employs surgical clinical reviewers who are rigorously trained to collect data through chart review and discussion with the treating surgeon and/or patient,³⁵ allowing for the use of robust and quality data with proven inter-rater reliability.^36,37

Using the 2011 to 2012 NSQIP database, we used primary Current Procedural Terminology codes to identify all patients who underwent percutaneous (27235) or open (27236) fixation of femoral neck fractures; and fixation with a screw and side plate (27244) or intramedullary fixation (27245) for peritrochanteric fractures. The sample was divided into 3 time-to-surgery groups (<24 hours from presentation to surgery, 24-48 hours, and >48 hours) which were matched for fracture type (femoral neck or peritrochanteric), sex, age (under 75 years or ≥75 years), and American Society of Anesthesiologists (ASA) class used as a surrogate for severity of medical infirmary. The subjects were randomly matched 1:1:1 to create 3 statistically equivalent time-to-surgery groups using Proc SurveySelect (SAS version 9.2, SAS Institute).

Generalized linear models using logit link function for binary variables and identity link function for normally distributed characteristics were used to compare the 3 time-to-surgery groups. Descriptive statistics are presented as counts and percentages or least-square means with standard deviations. Preoperative lab values that were not normally distributed were log transformed and presented in their original scales with median values and 25th to 75th percentiles. Outcomes were similarly modeled.  

Total hospital stay was modeled with a negative binomial distribution. Proportional hazards models were used to model the time from operating room (OR) to discharge, censoring patients who died before discharge, with results presented as hazard ratios (HR) and 95% confidence intervals (CI) (Figure). The assumption of the proportional hazards was tested using a Wald test. Using this model, a HR of <1 denotes a longer postoperative hospital stay, as a longer hospital stay decreases the “risk” for discharge.

All models were adjusted for confounders, including race, body mass index (BMI), hypertension, chronic obstructive pulmonary disease, cancer, bleeding disorders, transfusion within 72 hours before surgery, preoperative levels of creatinine, platelet count, white blood cells (WBCs), hematocrit anesthesia type, and wound infection. These covariates were selected based upon their observed relationship to the studied outcomes and time-to-surgery groups, and were evaluated across the models for all outcomes for consistency and clarity. All statistical analyses were run at a type I error rate of 5% and performed in SAS version 9.2 software.

Continue to: RESULTS...

RESULTS

A study population of 6036 hip fractures was identified and divided into 3 groups of 2012 subjects each based upon time-to-surgery. The groups were successfully matched for surgery type, age (≥75 years old), gender, and ASA class. In each group, 594 of the 2012 (29.5%) patients were male, 1525 (75.8%) were ≥75 years of age, 9 (.5%) were ASA Class I, 269 (13.4%) were ASA Class II, 1424 (70.8%) were ASA class III, and 309 (15.4%) were ASA class IV.

Significant differences in preoperative comorbidity burden and preoperative lab values were identified between the 3 cohorts. Increased time-to-surgery was associated with differences in race (P < .001), elevated BMI (P = .010), higher rates of congestive heart failure (P < .001), hypertension medication (P = .020), bleeding disorders (P < .001), blood transfusion within 72 hours of surgery (P < .001), and systemic sepsis (P = .001). Delay to surgery was also associated with lower preoperative sodium (P = .005), blood urea nitrogen (P = .013), serum WBC (P < .001), hematocrit (P < .001), and platelets (P < .001) (Table 1).

The unadjusted analyses revealed no association between time-to-surgery and return to OR (P = .554) nor readmission (P = .285). However, increasing time-to-surgery was associated with an increase in overall complications (P = .034), total length of hospital stay (P < .001), and 30-day mortality (P = .039) (Table 2).

Table 2. Estimated Event Rates from Matched Cohorts (Unadjusted)

Abbreviation: OR, operating room.

The adjusted analysis controlling for preoperative demographic and comorbidity variables revealed trends toward the increased overall complications and 30-day mortality with increased time-to-surgery; these trends showed no statistical significance (P = .143 and P = .08). No statistical relationship was observed between return to OR nor readmission and time-to-surgery. Increasing time-to-surgery remained significantly associated with the increased total length of hospital stay (P < .001). The adjusted analysis also revealed that the delay of >48 hours in time-to-surgery resulted in a longer surgery-to-discharge time (P < .001) (Table 3). No evidence of violation of the proportional hazards assumption was observed in the unadjusted nor adjusted clustered proportional hazards models (Wald test, P = .27 and P = .25, respectively).

Table 3. Estimated Event Rates from Matched Cohorts (Adjusted^a)

^aModel adjusted for race, hypertension medication, cancer, bleeding disorders, transfusion within 72 hours before surgery, emergency status, wound infection, anesthesia type (general), body mass index (18.5-25), history of chronic obstructive pulmonary disease, and preoperative levels of creatinine, platelet count, white blood cell count, and hematocrit.

Continue to: DISCUSSION...

DISCUSSION

Previous research has demonstrated an association between age,^3,4,25 comorbidity burden,^1,3,25 gender,^3,4 and ASA class^4,18,21 with outcomes following hip fractures and serves as the basis of our matched analysis statistical methodology in assessing the effect of time-to-surgery on the outcome following hip fracture surgery. Prior investigators have also established the positive correlation between increased preoperative comorbidity burden and delay in time-to-surgery.^10,15 This finding was confirmed in our unadjusted comparison of 3 time-to-surgery groups. However, prior investigations have not established a clear association between time-to-surgical intervention and postoperative morbidity and mortality.^{1,15,16,18,20,38} This study utilized a nationally representative dataset known for its data integrity and from which 6036 patients with surgically treated hip fractures, matched for surgery type, age, gender, and ASA class (a surrogate for severity of medical infirmary), were studied using adjusted regression modeling to afford an isolated statistical assessment of the effect of time-to-surgery on outcomes following hip fracture surgery.

Despite a large sample size and rigorous statistical methodology, for many outcome measures, our results show no support for the early or late operative intervention following hip fracture. We found no difference in 30-day mortality, readmission rate, nor total complication rate between the 3 time-to-surgery cohorts. This result indicates that the care of elderly patients following hip fracture is inherently complicated and that perioperative complication risk is probably only modestly modifiable by best medical practices, including optimizing time from clinical presentation to surgery.

As expected, patients who experienced longer delays from presentation to surgery were on average, more comorbid and more likely to yield abnormal preoperative lab values. However, in the adjusted analysis, delay in time-to-surgery, presumably for medical management, was not found to be associated with improved outcomes. In the same adjusted analysis, we uniquely identified that in the patients whose surgeries were delayed for more than 48 hours, the time from surgery-to-discharge was significantly increased. As a result, these patients spent extra days in the hospital both preoperatively and postoperatively, but without any corollary improvement in the outcomes.

Continue to: Recent estimates of the cost of hospital admission...

Recent estimates of the cost of hospital admission is approximated nationally at $2000/day.²⁶ Although our data fail to support the formal cost-analysis of the effect of time-to-surgery in hip fracture care, a simple value-based analysis indicates that quality is preserved (no difference in outcome), whereas costly hospital days are eliminated with earlier surgery. The value in elderly hip fracture care. defined as the outcomes relative to the costs,⁷ is ultimately optimized by earlier time-to-surgery.

Although using a large, multi-institutional database is advantageous for finding population-based trends that are representative of a large cohort, using the ACS-NSQIP database features its limitations. Our analysis was limited to the defined scope of NSQIP and nature of the injury, whereas root cause for delay was not available for study. We were unable to identify which patients were delayed for administrative reasons or surgical convenience and which were delayed for medical optimization. Participation in the ACS-NSQIP database is voluntary, and no randomized hospital sampling was conducted. Participating hospitals were de-identified in the database. As expected, we were unable to identify the specific institution-based hip fracture protocols that may affect the outcomes following treatment for these fractures. Further, socioeconomic information and payer-status are unavailable for the study. Additionally, observations are limited to 30 days postoperative, and we cannot comment on longer-term outcomes. Finally, discharge disposition and functional outcome data are not represented, and we were unable to correlate time-to-surgery and functional recovery. However, previous studies have established that delay in time-to-surgery following hip fractures is negatively correlated with functional outcomes.^22-25

Nevertheless, the ACS-NSQIP database remains one of the largest American surgical databases available, and includes care centers from nearly every state with variable demographics including rural, urban, and academic centers. The ACS performs broad-based inter-rater reliability audits on every participating site and has found an overall disagreement rate of only 1.8%. As such, although discrepancies exist between the complete patient chart and the data entered, the data found in the ACS-NSQIP database are reliable and considered a valid source of study.^34,35 The large sample size, quality of data collection, wide geographic representation, and varied hospital types within the dataset possibly make our findings relevant in the majority of American healthcare settings.

CONCLUSION

This study demonstrates an associated increased length of hospital stay, including the increased time from surgery-to-discharge, in patients with hip fractures whose surgical intervention is delayed for >48 hours after presentation. Given the prior evidence that early surgical intervention improves the functional outcomes and the current evidence that surgical delay for any cause increases costly hospital length of stay without corollary improvement in the outcomes, a value-based assessment of hip fracture care argues for early surgical intervention whenever possible. Our findings should inform physician, institution, and policy maker value-based decision making regarding the best practices in geriatric hip fracture care.

ABSTRACT

The morbidity and mortality after hip fracture in the elderly are influenced by non-modifiable comorbidities. Time-to-surgery is a modifiable factor that may play a role in postoperative morbidity. This study investigates the outcomes and complications in the elderly hip fracture surgery as a function of time-to-surgery.

Using the American College of Surgeons-National Surgical Quality Improvement Program data from 2011 to 2012, a study population was generated using the Current Procedural Terminology codes for percutaneous or open treatment of femoral neck fractures (27235, 27236) and fixation with a screw and side plate or intramedullary fixation (27244, 27245) for peritrochanteric fractures. Three time-to-surgery groups (<24 hours to surgical intervention, 24-48 hours, and >48 hours) were created and matched for surgery type, sex, age, and American Society of Anesthesiologists class. Time-to-surgery was then studied for its effect on the post-surgical outcomes using the adjusted regression modeling.

A study population of 6036 hip fractures was created, and 2012 patients were assigned to each matched time-to-surgery group. The unadjusted models showed that the earlier surgical intervention groups (<24 hours and 24-48 hours) exhibited a lower overall complication rate (P = .034) compared with the group waiting for surgery >48 hours. The unadjusted mortality rates increased with delay to surgical intervention (P = .039). Time-to-surgery caused no effect on the return to the operating room rate (P = .554) nor readmission rate (P = .285). Compared with other time-to-surgeries, the time-to-surgery of >48 hours was associated with prolonged total hospital length of stay (10.9 days) (P < .001) and a longer surgery-to-discharge time (hazard ratio, 95% confidence interval: 0.74, 0.69-0.79) (P < .001). Adjusted analyses showed no time-to-surgery related difference in complications (P = .143) but presented an increase in the total length of stay (P < .001) and surgery-to-discharge time (P < .001).

Timeliness of surgical intervention in a comorbidity-adjusted population of elderly hip fracture patients causes no effect on the overall complications, readmissions, nor 30-day mortality. However, time-to-surgery of >48 hours is associated with costly increase in the total length of stay, including an increased post-surgery-to-discharge time.

Continue to: Despite the best efforts to optimize surgical care...

Despite the best efforts to optimize surgical care and postoperative rehabilitation following hip fracture, elderly patients feature alarmingly high in-hospital and 1-year mortality rates of 4.35% to 9.2%^1-4 and 36%,⁵ respectively. Those who survive are unlikely to return to independent living, with only 17% of the patients following hip fracture being able to walk independently 6 months postoperatively, and 12% being able to climb stairs⁶. Possibly, these poor outcomes reflect a preoperative medical comorbidity burden rather than a measure of medical or surgical quality. Given the absence of consensus regarding optimal time-to-surgery, treating physicians often opt to delay surgical intervention for the purposes of medically optimizing highly comorbid patients without significant data to suggest clinical benefit of such practice.

Numerous investigators have attempted to identify the modifiable risk factors for complication after surgical care of elderly hip fracture patients. However, consensus guidelines of care are missing. This condition is largely due to the difficulties in effectively modifying preoperative demographic and medical comorbidities on a semi-urgent basis. However, timeliness to surgery is one area for study that the care team can affect. Although time-to-surgery is dependent on multiple factors, including time of presentation, day of week of admission, difficulties with scheduling, and administrative delays, the care team plays a role in hastening or retarding time-to-surgery. Several studies have considered various time cut-offs (24, 48, 72, and 120 hours) to define early intervention, but none have defined a specific role for early or delayed surgery. Several investigators have discovered a positive association between delayed time-to-surgery and mortality;^4,8-14 however, the most rigorously conducted studies that stringently control for preoperative comorbidities and demographics conclude that variance in time-to-surgery causes no effect on the in-hospital or 1-year mortality risk.^1-3,15-18  

Other investigators have shown that with early surgical intervention for hip fracture, patients experience shorter hospital stays,^{1,3,16,17,19-22} less days in pain,¹⁹ decreased risk of decubitus ulcers,^15,17,19,22 and an increased likelihood of independence following fracture,^22-25 regardless of preoperative medical status. Despite this evidence of improved outcomes with early surgery, 40% to 54% of hip fracture patients in the United States experience surgical delays of more than 24 to 48 hours. Additionally, with the recent (2013) national estimates of cost per day spent in the hospital falling between $1791 to $2289,²⁶ minimizing the days spent in the hospital would likely lead to significant cost-savings, presuming no adverse effect on health-related outcomes. To this end, we hypothesize that the value (outcomes per associated cost)⁷ of hip fracture surgical care can be positively influenced by minimizing surgical wait-times. We assessed the effect of early surgical intervention, within 24 or 48 hours of presentation, on 30-day mortality, postoperative morbidity, hospital length of stay, and readmission rates in a comorbidity-adjusted population from a nationally representative cohort.

Continue to: METHODS AND MATERIALS...

METHODS AND MATERIALS 

This study used the data from the American College of Surgeon-National Surgical Quality Improvement Program (ACS-NSQIP) database. With over 258 participating hospitals, this database has been widely used to identify national trends in various surgical specialties.^27-34 The database includes information from participants in 43 states with hospitals ranging from rural community hospitals to large academic centers. Each site employs surgical clinical reviewers who are rigorously trained to collect data through chart review and discussion with the treating surgeon and/or patient,³⁵ allowing for the use of robust and quality data with proven inter-rater reliability.^36,37

Using the 2011 to 2012 NSQIP database, we used primary Current Procedural Terminology codes to identify all patients who underwent percutaneous (27235) or open (27236) fixation of femoral neck fractures; and fixation with a screw and side plate (27244) or intramedullary fixation (27245) for peritrochanteric fractures. The sample was divided into 3 time-to-surgery groups (<24 hours from presentation to surgery, 24-48 hours, and >48 hours) which were matched for fracture type (femoral neck or peritrochanteric), sex, age (under 75 years or ≥75 years), and American Society of Anesthesiologists (ASA) class used as a surrogate for severity of medical infirmary. The subjects were randomly matched 1:1:1 to create 3 statistically equivalent time-to-surgery groups using Proc SurveySelect (SAS version 9.2, SAS Institute).

Generalized linear models using logit link function for binary variables and identity link function for normally distributed characteristics were used to compare the 3 time-to-surgery groups. Descriptive statistics are presented as counts and percentages or least-square means with standard deviations. Preoperative lab values that were not normally distributed were log transformed and presented in their original scales with median values and 25th to 75th percentiles. Outcomes were similarly modeled.  

Total hospital stay was modeled with a negative binomial distribution. Proportional hazards models were used to model the time from operating room (OR) to discharge, censoring patients who died before discharge, with results presented as hazard ratios (HR) and 95% confidence intervals (CI) (Figure). The assumption of the proportional hazards was tested using a Wald test. Using this model, a HR of <1 denotes a longer postoperative hospital stay, as a longer hospital stay decreases the “risk” for discharge.

All models were adjusted for confounders, including race, body mass index (BMI), hypertension, chronic obstructive pulmonary disease, cancer, bleeding disorders, transfusion within 72 hours before surgery, preoperative levels of creatinine, platelet count, white blood cells (WBCs), hematocrit anesthesia type, and wound infection. These covariates were selected based upon their observed relationship to the studied outcomes and time-to-surgery groups, and were evaluated across the models for all outcomes for consistency and clarity. All statistical analyses were run at a type I error rate of 5% and performed in SAS version 9.2 software.

Continue to: RESULTS...

RESULTS

A study population of 6036 hip fractures was identified and divided into 3 groups of 2012 subjects each based upon time-to-surgery. The groups were successfully matched for surgery type, age (≥75 years old), gender, and ASA class. In each group, 594 of the 2012 (29.5%) patients were male, 1525 (75.8%) were ≥75 years of age, 9 (.5%) were ASA Class I, 269 (13.4%) were ASA Class II, 1424 (70.8%) were ASA class III, and 309 (15.4%) were ASA class IV.

Significant differences in preoperative comorbidity burden and preoperative lab values were identified between the 3 cohorts. Increased time-to-surgery was associated with differences in race (P < .001), elevated BMI (P = .010), higher rates of congestive heart failure (P < .001), hypertension medication (P = .020), bleeding disorders (P < .001), blood transfusion within 72 hours of surgery (P < .001), and systemic sepsis (P = .001). Delay to surgery was also associated with lower preoperative sodium (P = .005), blood urea nitrogen (P = .013), serum WBC (P < .001), hematocrit (P < .001), and platelets (P < .001) (Table 1).

The unadjusted analyses revealed no association between time-to-surgery and return to OR (P = .554) nor readmission (P = .285). However, increasing time-to-surgery was associated with an increase in overall complications (P = .034), total length of hospital stay (P < .001), and 30-day mortality (P = .039) (Table 2).

Table 2. Estimated Event Rates from Matched Cohorts (Unadjusted)

Abbreviation: OR, operating room.

The adjusted analysis controlling for preoperative demographic and comorbidity variables revealed trends toward the increased overall complications and 30-day mortality with increased time-to-surgery; these trends showed no statistical significance (P = .143 and P = .08). No statistical relationship was observed between return to OR nor readmission and time-to-surgery. Increasing time-to-surgery remained significantly associated with the increased total length of hospital stay (P < .001). The adjusted analysis also revealed that the delay of >48 hours in time-to-surgery resulted in a longer surgery-to-discharge time (P < .001) (Table 3). No evidence of violation of the proportional hazards assumption was observed in the unadjusted nor adjusted clustered proportional hazards models (Wald test, P = .27 and P = .25, respectively).

Table 3. Estimated Event Rates from Matched Cohorts (Adjusted^a)

^aModel adjusted for race, hypertension medication, cancer, bleeding disorders, transfusion within 72 hours before surgery, emergency status, wound infection, anesthesia type (general), body mass index (18.5-25), history of chronic obstructive pulmonary disease, and preoperative levels of creatinine, platelet count, white blood cell count, and hematocrit.

Continue to: DISCUSSION...

DISCUSSION

Previous research has demonstrated an association between age,^3,4,25 comorbidity burden,^1,3,25 gender,^3,4 and ASA class^4,18,21 with outcomes following hip fractures and serves as the basis of our matched analysis statistical methodology in assessing the effect of time-to-surgery on the outcome following hip fracture surgery. Prior investigators have also established the positive correlation between increased preoperative comorbidity burden and delay in time-to-surgery.^10,15 This finding was confirmed in our unadjusted comparison of 3 time-to-surgery groups. However, prior investigations have not established a clear association between time-to-surgical intervention and postoperative morbidity and mortality.^{1,15,16,18,20,38} This study utilized a nationally representative dataset known for its data integrity and from which 6036 patients with surgically treated hip fractures, matched for surgery type, age, gender, and ASA class (a surrogate for severity of medical infirmary), were studied using adjusted regression modeling to afford an isolated statistical assessment of the effect of time-to-surgery on outcomes following hip fracture surgery.

Despite a large sample size and rigorous statistical methodology, for many outcome measures, our results show no support for the early or late operative intervention following hip fracture. We found no difference in 30-day mortality, readmission rate, nor total complication rate between the 3 time-to-surgery cohorts. This result indicates that the care of elderly patients following hip fracture is inherently complicated and that perioperative complication risk is probably only modestly modifiable by best medical practices, including optimizing time from clinical presentation to surgery.

As expected, patients who experienced longer delays from presentation to surgery were on average, more comorbid and more likely to yield abnormal preoperative lab values. However, in the adjusted analysis, delay in time-to-surgery, presumably for medical management, was not found to be associated with improved outcomes. In the same adjusted analysis, we uniquely identified that in the patients whose surgeries were delayed for more than 48 hours, the time from surgery-to-discharge was significantly increased. As a result, these patients spent extra days in the hospital both preoperatively and postoperatively, but without any corollary improvement in the outcomes.

Continue to: Recent estimates of the cost of hospital admission...

Recent estimates of the cost of hospital admission is approximated nationally at $2000/day.²⁶ Although our data fail to support the formal cost-analysis of the effect of time-to-surgery in hip fracture care, a simple value-based analysis indicates that quality is preserved (no difference in outcome), whereas costly hospital days are eliminated with earlier surgery. The value in elderly hip fracture care. defined as the outcomes relative to the costs,⁷ is ultimately optimized by earlier time-to-surgery.

Although using a large, multi-institutional database is advantageous for finding population-based trends that are representative of a large cohort, using the ACS-NSQIP database features its limitations. Our analysis was limited to the defined scope of NSQIP and nature of the injury, whereas root cause for delay was not available for study. We were unable to identify which patients were delayed for administrative reasons or surgical convenience and which were delayed for medical optimization. Participation in the ACS-NSQIP database is voluntary, and no randomized hospital sampling was conducted. Participating hospitals were de-identified in the database. As expected, we were unable to identify the specific institution-based hip fracture protocols that may affect the outcomes following treatment for these fractures. Further, socioeconomic information and payer-status are unavailable for the study. Additionally, observations are limited to 30 days postoperative, and we cannot comment on longer-term outcomes. Finally, discharge disposition and functional outcome data are not represented, and we were unable to correlate time-to-surgery and functional recovery. However, previous studies have established that delay in time-to-surgery following hip fractures is negatively correlated with functional outcomes.^22-25

Nevertheless, the ACS-NSQIP database remains one of the largest American surgical databases available, and includes care centers from nearly every state with variable demographics including rural, urban, and academic centers. The ACS performs broad-based inter-rater reliability audits on every participating site and has found an overall disagreement rate of only 1.8%. As such, although discrepancies exist between the complete patient chart and the data entered, the data found in the ACS-NSQIP database are reliable and considered a valid source of study.^34,35 The large sample size, quality of data collection, wide geographic representation, and varied hospital types within the dataset possibly make our findings relevant in the majority of American healthcare settings.

CONCLUSION

This study demonstrates an associated increased length of hospital stay, including the increased time from surgery-to-discharge, in patients with hip fractures whose surgical intervention is delayed for >48 hours after presentation. Given the prior evidence that early surgical intervention improves the functional outcomes and the current evidence that surgical delay for any cause increases costly hospital length of stay without corollary improvement in the outcomes, a value-based assessment of hip fracture care argues for early surgical intervention whenever possible. Our findings should inform physician, institution, and policy maker value-based decision making regarding the best practices in geriatric hip fracture care.

Benzodiazepines (BZDs) are among the most commonly prescribed medications. A recent study found that in 2008, more than 5% of Americans used a BZD, and the percentage was almost 9% among Americans aged ≥ 65 years.^1,2 Among veterans, BZD use is even higher, in part because of the high prevalence of posttraumatic stress disorder (PTSD). One study found that more than 30% of veterans with PTSD received at least 1 BZD prescription.³ The risks associated with BZD treatment for PTSD are compounded by concurrent use of other sedatives and opioids prescribed for co-occurring chronic pain and insomnia.³

Older adults metabolize long-acting BZDs more slowly and generally have an increased sensitivity to the adverse effects (AEs) of all BZDs.⁴ In older adults, BZD use has been associated with cognitive decline, dementia, falls and consequent fractures, and adverse respiratory outcomes.^5-12 The risk of most but not all of these AEs was increased with higher BZD dose or long-term BZD use, which this quality improvement project (QIP) defines as having at least a 60-day supply of BZD prescriptions dispensed within the past year.

Long-term BZD use increases with age. One study found that, among patients receiving a BZD, the rate of long-term BZD use was more than double in older adults (31.4%) than it was in adults aged between 18 and 35 years (14.7%).² For these reasons, the 2012 Beers criteria of the American Geriatrics Society recommend avoiding all types of BZDs in the treatment of insomnia, agitation, or delirium in patients aged > 65 years.¹³ Despite this recommendation, the prevalence of BZD use in older adults remains high.¹⁴

Some innovative approaches have been developed to address the inappropriate use, including overuse and misuse, of BZDs in older adults.¹⁵ In one approach, direct-to-consumer (DTC) information is used to empower patients to collaborate with their physician to manage their health. Results from several studies suggest that providing older patients with information on BZD risks and benefits increases patient–physician interaction and thereby decreases inappropriate BZD use and improves health outcomes.^4,16,17 One study found that perceptions of BZD risks increased 1 week after exposure to a DTC educational brochure (EB), with intention to discuss BZD discontinuation with their physician higher for patients who received the EB than it was for those who did not (83.1% vs 44.3%; P < .0001).¹⁶ The EMPOWER (Eliminating Medications Through Patient Ownership of End Results) cluster randomized controlled trial assessed the effectiveness of a DTC EB focused on BZD risks in older adults.¹⁷ In that seminal study, patients who received a DTC EB were more likely than were comparison patients to discontinue BZD within 6 months (27% vs 5%; risk difference, 23%; 95% CI, 14%-32%).

The Veterans Integrated Systems Network (VISN) 22 Academic Detailing Program is a pharmacy educational outreach program that uses unbiased clinical guidelines to promote physicians’ safety initiatives and align prescribing behavior with best practices.^18-20 With BZD use among older veterans remaining high, the VISN 22 program initiated a clinical QIP modeled on the EMPOWER trial. Veterans in VISN 22 received the DTC EB, which included information on BZD risks and encouraged them to discuss their BZD treatment with their health care provider. VISN 22 was the first VISN in the VHA to implement the EMPOWER protocol.

As this was a QIP, all eligible veterans in VISN 22 were mailed the DTC EB, thus making it difficult to estimate the impact of the EB on BZD discontinuation in this VISN. Therefore, DTC EB efficacy was estimated by comparing BZD discontinuation between VISN 22 and VISN 21, an adjacent VISN that did not mail the DTC EB. To reduce selection bias associated with different controls in the 2 VISNs, the authors performed propensity score matching (PSM) to balance the covariates and provide an unbiased estimate of the mean treatment effect of the DTC EB in VISN 22 veterans who were included in the initial descriptive QIP and received the EB; these veterans were compared with VISN 21 veterans who did not receive the EB.

Methods

Two QIPs were undertaken to determine the impact of DTC EB on BZD use in older veterans in the VHA.

Quality Improvement Project 1

Design. A retrospective cohort analysis was performed. The VISN 22 catchment area, which encompasses VA facilities and clinics in southern California and southern Nevada, serves about 500,000 veterans, a substantial proportion of whom are aged ≥ 65 years. Among these older veterans are active long-term BZD users, who were defined as having ≥ 60-day supply of BZD prescriptions dispensed within the past year. Each active long-term user with a BZD prescription released within 200 days before the index date (the date the user was to meet with the prescribing physician) was mailed an EB 2 to 8 weeks in advance of the visit. Excluded from analysis were veterans with a schizophrenia, spinal cord injury, or seizure disorder diagnosis recorded in both their inpatient and outpatient medical records; veterans seen by Palliative Care within the past year; and veterans who died before analysis was completed.

Education Brochure. The EB for VISN 22 (Figure 1, see

Patients. The sample consisted of all veterans identified as meeting the inclusion criteria and being enrolled in VISN 22. The EB was mailed once to veterans on a rolling basis from December 2014 to February 2016. Change in BZD use was analyzed only after 9 to 24 months had passed since the index appointment with the prescribing physician. This period included 12 weeks for BZD taper and then 6 months after taper.

Analysis. For each veteran, monthly mean lorazepam equivalent (LE) was calculated using as many as 12 fills before the index date. Average daily dose of LE was calculated by dividing the sum of LE from all included prescriptions by total number of days between the first fill and the index date. The BZD prescription fills were evaluated after the index date. Veterans who received at least 1 prescription after the index date but then had no BZD prescription activity in VA clinics for 3 consecutive months during the 9-month observation period were recorded as having tapered and then discontinued BZD. Veterans who had no BZD prescription activity in VA clinics after the index date and during the 9-month observation period were recorded as having discontinued BZD without tapering. For veterans who had BZD prescription activity in VA clinics after the index date and during the 9-month observation period, mean LE was calculated by dividing the total LE for BZD prescriptions after the index date by number of days from the first fill after the index date to the date of analysis.

Quality Improvement Project 2

Design. A retrospective cohort analysis using PSM was performed on a subgroup of the QIP-1 sample to evaluate the impact of EB on BZD prescribing in the VA during 2 periods: 6 to 9 months and 6 to 12 months after the index date. A secondary outcome was discontinuation 1 to 12 months after the index date. Veterans in the analysis were active long-term BZD users, had at least 1 BZD prescription released within 200 days before the index date, were aged ≥ 65 years, and had an appointment scheduled with their BZD prescriber within 2 to 8 weeks (Figure 2). 

Patients. VISN 22 implemented QIP-2, a real-world application of a modified EMPOWER program, by identifying eligible veterans on a rolling basis from December 2014 to August 2015. All veterans who were identified and sent an EB during this period were included in the case group. The index date was defined as the first of the month the EB was mailed. Veterans with a pending appointment were chosen because the lead time would allow them to receive the EB and prepare to discuss it with the physician during the visit.

A comparator group was drawn from the adjacent VISN 21 catchment area, which encompasses VA facilities and clinics in Hawaii, northern California, and northern Nevada. During the observation period, VISN 21 did not mail any EBs specifically addressing BZD risks. Veterans in the comparator group had an appointment scheduled with their BZD prescribing physician within 4 weeks, were aged ≥ 65 years on the index date (first of the month before the next appointment, coinciding with the date EBs were sent to VISN 22 veterans), were active long-term BZD users, and had at least 1 BZD prescription released within 200 days before the index date. All patients were followed for up to 12 months after the index date, with BZD discontinuation recorded 9 and 12 months after the index date.

Propensity Score Matching

Propensity score (PS) was estimated with logistic regression analysis with treatment as the dependent variable and baseline characteristics as the independent variables.^21,22 One-to-one matching on the PS was performed using the nearest neighbor approach without replacements. Independent variables related to outcome but unrelated to EB exposure were selected for PS development.²² These variables included year of birth; male sex; Hispanic ethnicity; annual income; service connection status; region; body mass index; Charlson Comorbidity Index category; total baseline BZD dose; and diagnosis of AIDS, nonmetastatic cancer, metastatic cancer, chronic obstructive pulmonary disease (COPD), congestive heart failure (CHF), dementia, diabetes mellitus (DM), DM with complications, gastroesophageal reflux disease (GERD), general anxiety disorder (GAD), hemiparaplegia, liver disease (mild), liver disease (moderate to severe), myocardial infarction (MI), Parkinson disease, peptic ulcer disease (PUD), psychosis, renal disease, rheumatoid arthritis (RA), or substance use disorder (SUD).

The EMPOWER cluster randomized controlled trial (RCT) demonstrated the effectiveness of EB exposure in a Canadian population of elderly patients who were long-term BZD users.¹⁷ Randomized controlled trials are the gold standard for clinical trials because they can establish causal inference.^23-25 Given ethical and practical concerns, however, RCTs cannot be applied to all clinical scenarios. Although EMPOWER is reported to be an effective tool in reducing BZD use in older adults, its application in a real-world, large, integrated health care system remains untested. Observational studies are often conducted as an alternative to RCTs but are subject to selection bias because of their lack of randomization.²⁶ Therefore, robust research methods are needed to generate unbiased estimates of the impact of an intervention on an outcome. Propensity score matching simulates an RCT by balancing the covariates across treatment groups.^21,22,27 Observed patient characteristics are used to estimate PS, the probability that treatment will be received. Logistic or probit regression is used to balance the potential confounding covariates between the treatment groups.Once PSs are known, mean treatment effect can be estimated without the mean model.²⁸ In other words, PSM methods can be used to generate an unbiased estimate of the treatment.

Propensity Score Analysis

Baseline characteristics were compared using Student t test (continuous variables) and χ² test (discrete variables). Results are presented as means and standard deviations (continuous variables) and frequency and percentage (discrete variables).

The main outcome was BZD discontinuation 9 and 12 months after the index date. A postindex lag of 6 months was used to capture any tapering (Figure 2). Discontinuation, defined as 3 consecutive months of no BZD prescription on hand, was measured for 2 periods: 6 to 9 months and 6 to 12 months after the index date. A secondary outcome was discontinuation 1 to 12 months after the index date. An estimate was made of the difference in the proportions of BZD discontinuers who received the EB and BZD discontinuers who did not receive the EB, where mean treatment (risk difference) was presented as the absolute risk difference with a 95% CI. Standard errors and 95% CIs for the risk differences were generated with biased-corrected CIs from 1,000 bootstrap samples.

Sensitivity Analyses

Naïve multivariate logistic regression analysis was performed to evaluate the association between EB exposure and BZD discontinuation while controlling for potential confounders. Results are presented as odds ratios (ORs) and 95% CIs. Confounders identified were the same covariates used to generate the PSs.

Several analyses were performed to test the sensitivity of the methods applied using PSM by changing caliber size while maintaining the nearest neighbor approach without replacement. Linear regression analysis was performed with robust standard errors to estimate the risk difference of BZD discontinuation between EB-exposed and EB-unexposed veterans.

Statistical significance was set at P < .05. All statistical analyses were performed with Stata/SE Version 13 (College Station, TX).

Results

Quality Improvement Project 1

On a rolling basis from December 2014 to February 2016, the EB was mailed once to 3,896 VISN 22 veterans 2 to 8 weeks before a clinic appointment with their BZD prescribing physician. 

Quality Improvement Project 2

Of all the VISN 22 and VISN 21 veterans, 24,420 met the inclusion and exclusion criteria. Of these 24,420 veterans, 2,020 (8.3%) were in VISN 22 and received the EB between December 2014 and August 2015 (QIP-1), and 22,400 (91.7%) were in VISN 21 and did not receive the EB.

Naïve Results Before PS Matching. In the naïve analyses, a larger proportion of EB-exposed vs unexposed veterans discontinued BZD; in addition, reductions were 6.6%, 7.4%, and 9.5% larger for 6 to 9 months, 6 to 12 months, and 1 to 12 months after the index date, respectively (P < .0001 for all comparisons; Table 2).

Propensity Score Matching. Before matching, there were significant differences in baseline characteristics of veterans who met the inclusion and exclusion criteria, with few exceptions (eAppendices 2 and 3, ).

Propensity Score Matching Results. Inspection of PSs revealed good coverage across treatment groups on a histogram plot and a kernel density plot (eAppendices 5 and 6).

Discussion

This QIP was the first to evaluate the impact of an EMPOWER-modeled DTC EB in a large, integrated health care system in the U.S. It was also the first to demonstrate potential benefits of a DTC EB designed for older veterans who are long-term BZD users. In this QIP, which mailed the EB to 3,896 veterans, 1,847 (47.4%) decreased their BZD dose, 458 (11.7%) tapered and then discontinued BZD, and 455 (11.7%) immediately discontinued BZD. The total percentage of veterans who discontinued BZD (23.4%; 913/3,896) was similar to the 27% reported in the EMPOWER trial.¹⁷ However, the risk difference between the 1,316 EB-exposed VISN 22 veterans (QIP-1) and the 1,316 EB-unexposed VISN 21 veterans in this QIP was significantly lower than the 23% risk difference in EMPOWER (though it still demonstrated a significantly larger reduction for EB-exposed veterans).¹⁷

Given this inclusion of all qualifying veterans from the catchment area studied in this QIP, and given the ethical and practical concerns, an RCT was not possible. Therefore, PSM methods were used to balance the covariates across treatment groups and thereby simulate an RCT.^21,22,27 With use of the PSM approach, findings from the descriptive analysis were confirmed and potential selection bias reduced.

Study Limitations

The less robust risk difference found in this QIP has several possible explanations. The authors’ use of a DTC EB coincided with a national VA effort to reduce older veterans’ use of BZDs and other inappropriate medications. For instance, during the study period, academic detailing was being implemented to reduce use of BZDs, particularly in combination with opioids, across VHA facilities and clinics. (Academic detailing is a pharmacy educational outreach program that uses unbiased clinical guidelines to promote physicians’ safety initiatives and align prescribing behavior with best practices.^18-20) However, QIP-2 results and PS analysis of a subgroup of the original sample suggest that EB-exposed veterans were significantly more likely than were their unexposed counterparts were to discontinue BZD. To an extent, this analysis controlled for these other efforts to reduce BZD use in VHA clinics and can be considered a study strength.

Another limitation is the study design, which lacked a control group and did not consider the possibility that some facility or clinic physicians might influence others. Although the region variable was controlled for in PSM, the authors did not capture facility characteristics, including frequency of prescribing BZD and use of a protocol for enforcing the Beers criteria. Such confounders might have influenced outcomes. Unlike the EMPOWER trial,¹⁷ this QIP did not assess or exclude cognitively impaired veterans. It is reasonable to assume that these veterans might not understand some EB messages and consequently might fail to engage their physicians. Failure to initiate discussion with a physician would attenuate the impact of the EB.

Study Strengths

A strength of this QIP was its use of a DTC EB in a large, regional sample of older veterans in a real-world clinical setting. In addition, the study group (EB-exposed veterans) and the comparator group (EB-unexposed veterans) were from similar geographic areas (primarily California and Nevada).

Conclusion

Results of this study suggest that a DTC EB, designed to reduce BZD use among older veterans, was effective in helping patients lower their BZD dose and discontinue BZD. The likelihood of discontinuing BZD 9 and 12 months after the index date was significantly higher for veterans who received an EB modeled on the EMPOWER educational brochure than for a comparator group of veterans who did not receive the EB and were receiving care during the same observation period. In the future, it would be beneficial to use a design that controls for physician exposure to academic detailing focused on BZD reduction and that accounts for the cluster effects of facility practice. Despite these limitations, this QIP is the first real-world empirical example of using an EMPOWER-modeled DTC EB to decrease BZD use among older veterans. Furthermore, these results suggest that a DTC EB can be used to target other high-risk prescription drugs, such as opioids, particularly if alternative treatment options can be provided.

Acknowledgments
Dr. Hauser thanks Cathy, Anika, Katia, and Max Hauser, and Alba and Kevin Quinlan, for their support. In memory of Jirina Hauser, who died on Mother’s Day, May 14, 2017, at the age of 100.

Benzodiazepines (BZDs) are among the most commonly prescribed medications. A recent study found that in 2008, more than 5% of Americans used a BZD, and the percentage was almost 9% among Americans aged ≥ 65 years.^1,2 Among veterans, BZD use is even higher, in part because of the high prevalence of posttraumatic stress disorder (PTSD). One study found that more than 30% of veterans with PTSD received at least 1 BZD prescription.³ The risks associated with BZD treatment for PTSD are compounded by concurrent use of other sedatives and opioids prescribed for co-occurring chronic pain and insomnia.³

Older adults metabolize long-acting BZDs more slowly and generally have an increased sensitivity to the adverse effects (AEs) of all BZDs.⁴ In older adults, BZD use has been associated with cognitive decline, dementia, falls and consequent fractures, and adverse respiratory outcomes.^5-12 The risk of most but not all of these AEs was increased with higher BZD dose or long-term BZD use, which this quality improvement project (QIP) defines as having at least a 60-day supply of BZD prescriptions dispensed within the past year.

Long-term BZD use increases with age. One study found that, among patients receiving a BZD, the rate of long-term BZD use was more than double in older adults (31.4%) than it was in adults aged between 18 and 35 years (14.7%).² For these reasons, the 2012 Beers criteria of the American Geriatrics Society recommend avoiding all types of BZDs in the treatment of insomnia, agitation, or delirium in patients aged > 65 years.¹³ Despite this recommendation, the prevalence of BZD use in older adults remains high.¹⁴

Some innovative approaches have been developed to address the inappropriate use, including overuse and misuse, of BZDs in older adults.¹⁵ In one approach, direct-to-consumer (DTC) information is used to empower patients to collaborate with their physician to manage their health. Results from several studies suggest that providing older patients with information on BZD risks and benefits increases patient–physician interaction and thereby decreases inappropriate BZD use and improves health outcomes.^4,16,17 One study found that perceptions of BZD risks increased 1 week after exposure to a DTC educational brochure (EB), with intention to discuss BZD discontinuation with their physician higher for patients who received the EB than it was for those who did not (83.1% vs 44.3%; P < .0001).¹⁶ The EMPOWER (Eliminating Medications Through Patient Ownership of End Results) cluster randomized controlled trial assessed the effectiveness of a DTC EB focused on BZD risks in older adults.¹⁷ In that seminal study, patients who received a DTC EB were more likely than were comparison patients to discontinue BZD within 6 months (27% vs 5%; risk difference, 23%; 95% CI, 14%-32%).

The Veterans Integrated Systems Network (VISN) 22 Academic Detailing Program is a pharmacy educational outreach program that uses unbiased clinical guidelines to promote physicians’ safety initiatives and align prescribing behavior with best practices.^18-20 With BZD use among older veterans remaining high, the VISN 22 program initiated a clinical QIP modeled on the EMPOWER trial. Veterans in VISN 22 received the DTC EB, which included information on BZD risks and encouraged them to discuss their BZD treatment with their health care provider. VISN 22 was the first VISN in the VHA to implement the EMPOWER protocol.

As this was a QIP, all eligible veterans in VISN 22 were mailed the DTC EB, thus making it difficult to estimate the impact of the EB on BZD discontinuation in this VISN. Therefore, DTC EB efficacy was estimated by comparing BZD discontinuation between VISN 22 and VISN 21, an adjacent VISN that did not mail the DTC EB. To reduce selection bias associated with different controls in the 2 VISNs, the authors performed propensity score matching (PSM) to balance the covariates and provide an unbiased estimate of the mean treatment effect of the DTC EB in VISN 22 veterans who were included in the initial descriptive QIP and received the EB; these veterans were compared with VISN 21 veterans who did not receive the EB.

Methods

Two QIPs were undertaken to determine the impact of DTC EB on BZD use in older veterans in the VHA.

Quality Improvement Project 1

Design. A retrospective cohort analysis was performed. The VISN 22 catchment area, which encompasses VA facilities and clinics in southern California and southern Nevada, serves about 500,000 veterans, a substantial proportion of whom are aged ≥ 65 years. Among these older veterans are active long-term BZD users, who were defined as having ≥ 60-day supply of BZD prescriptions dispensed within the past year. Each active long-term user with a BZD prescription released within 200 days before the index date (the date the user was to meet with the prescribing physician) was mailed an EB 2 to 8 weeks in advance of the visit. Excluded from analysis were veterans with a schizophrenia, spinal cord injury, or seizure disorder diagnosis recorded in both their inpatient and outpatient medical records; veterans seen by Palliative Care within the past year; and veterans who died before analysis was completed.

Education Brochure. The EB for VISN 22 (Figure 1, see

Patients. The sample consisted of all veterans identified as meeting the inclusion criteria and being enrolled in VISN 22. The EB was mailed once to veterans on a rolling basis from December 2014 to February 2016. Change in BZD use was analyzed only after 9 to 24 months had passed since the index appointment with the prescribing physician. This period included 12 weeks for BZD taper and then 6 months after taper.

Analysis. For each veteran, monthly mean lorazepam equivalent (LE) was calculated using as many as 12 fills before the index date. Average daily dose of LE was calculated by dividing the sum of LE from all included prescriptions by total number of days between the first fill and the index date. The BZD prescription fills were evaluated after the index date. Veterans who received at least 1 prescription after the index date but then had no BZD prescription activity in VA clinics for 3 consecutive months during the 9-month observation period were recorded as having tapered and then discontinued BZD. Veterans who had no BZD prescription activity in VA clinics after the index date and during the 9-month observation period were recorded as having discontinued BZD without tapering. For veterans who had BZD prescription activity in VA clinics after the index date and during the 9-month observation period, mean LE was calculated by dividing the total LE for BZD prescriptions after the index date by number of days from the first fill after the index date to the date of analysis.

Quality Improvement Project 2

Design. A retrospective cohort analysis using PSM was performed on a subgroup of the QIP-1 sample to evaluate the impact of EB on BZD prescribing in the VA during 2 periods: 6 to 9 months and 6 to 12 months after the index date. A secondary outcome was discontinuation 1 to 12 months after the index date. Veterans in the analysis were active long-term BZD users, had at least 1 BZD prescription released within 200 days before the index date, were aged ≥ 65 years, and had an appointment scheduled with their BZD prescriber within 2 to 8 weeks (Figure 2). 

Patients. VISN 22 implemented QIP-2, a real-world application of a modified EMPOWER program, by identifying eligible veterans on a rolling basis from December 2014 to August 2015. All veterans who were identified and sent an EB during this period were included in the case group. The index date was defined as the first of the month the EB was mailed. Veterans with a pending appointment were chosen because the lead time would allow them to receive the EB and prepare to discuss it with the physician during the visit.

A comparator group was drawn from the adjacent VISN 21 catchment area, which encompasses VA facilities and clinics in Hawaii, northern California, and northern Nevada. During the observation period, VISN 21 did not mail any EBs specifically addressing BZD risks. Veterans in the comparator group had an appointment scheduled with their BZD prescribing physician within 4 weeks, were aged ≥ 65 years on the index date (first of the month before the next appointment, coinciding with the date EBs were sent to VISN 22 veterans), were active long-term BZD users, and had at least 1 BZD prescription released within 200 days before the index date. All patients were followed for up to 12 months after the index date, with BZD discontinuation recorded 9 and 12 months after the index date.

Propensity Score Matching

Propensity score (PS) was estimated with logistic regression analysis with treatment as the dependent variable and baseline characteristics as the independent variables.^21,22 One-to-one matching on the PS was performed using the nearest neighbor approach without replacements. Independent variables related to outcome but unrelated to EB exposure were selected for PS development.²² These variables included year of birth; male sex; Hispanic ethnicity; annual income; service connection status; region; body mass index; Charlson Comorbidity Index category; total baseline BZD dose; and diagnosis of AIDS, nonmetastatic cancer, metastatic cancer, chronic obstructive pulmonary disease (COPD), congestive heart failure (CHF), dementia, diabetes mellitus (DM), DM with complications, gastroesophageal reflux disease (GERD), general anxiety disorder (GAD), hemiparaplegia, liver disease (mild), liver disease (moderate to severe), myocardial infarction (MI), Parkinson disease, peptic ulcer disease (PUD), psychosis, renal disease, rheumatoid arthritis (RA), or substance use disorder (SUD).

The EMPOWER cluster randomized controlled trial (RCT) demonstrated the effectiveness of EB exposure in a Canadian population of elderly patients who were long-term BZD users.¹⁷ Randomized controlled trials are the gold standard for clinical trials because they can establish causal inference.^23-25 Given ethical and practical concerns, however, RCTs cannot be applied to all clinical scenarios. Although EMPOWER is reported to be an effective tool in reducing BZD use in older adults, its application in a real-world, large, integrated health care system remains untested. Observational studies are often conducted as an alternative to RCTs but are subject to selection bias because of their lack of randomization.²⁶ Therefore, robust research methods are needed to generate unbiased estimates of the impact of an intervention on an outcome. Propensity score matching simulates an RCT by balancing the covariates across treatment groups.^21,22,27 Observed patient characteristics are used to estimate PS, the probability that treatment will be received. Logistic or probit regression is used to balance the potential confounding covariates between the treatment groups.Once PSs are known, mean treatment effect can be estimated without the mean model.²⁸ In other words, PSM methods can be used to generate an unbiased estimate of the treatment.

Propensity Score Analysis

Baseline characteristics were compared using Student t test (continuous variables) and χ² test (discrete variables). Results are presented as means and standard deviations (continuous variables) and frequency and percentage (discrete variables).

The main outcome was BZD discontinuation 9 and 12 months after the index date. A postindex lag of 6 months was used to capture any tapering (Figure 2). Discontinuation, defined as 3 consecutive months of no BZD prescription on hand, was measured for 2 periods: 6 to 9 months and 6 to 12 months after the index date. A secondary outcome was discontinuation 1 to 12 months after the index date. An estimate was made of the difference in the proportions of BZD discontinuers who received the EB and BZD discontinuers who did not receive the EB, where mean treatment (risk difference) was presented as the absolute risk difference with a 95% CI. Standard errors and 95% CIs for the risk differences were generated with biased-corrected CIs from 1,000 bootstrap samples.

Sensitivity Analyses

Naïve multivariate logistic regression analysis was performed to evaluate the association between EB exposure and BZD discontinuation while controlling for potential confounders. Results are presented as odds ratios (ORs) and 95% CIs. Confounders identified were the same covariates used to generate the PSs.

Several analyses were performed to test the sensitivity of the methods applied using PSM by changing caliber size while maintaining the nearest neighbor approach without replacement. Linear regression analysis was performed with robust standard errors to estimate the risk difference of BZD discontinuation between EB-exposed and EB-unexposed veterans.

Statistical significance was set at P < .05. All statistical analyses were performed with Stata/SE Version 13 (College Station, TX).

Results

Quality Improvement Project 1

On a rolling basis from December 2014 to February 2016, the EB was mailed once to 3,896 VISN 22 veterans 2 to 8 weeks before a clinic appointment with their BZD prescribing physician. 

Quality Improvement Project 2

Of all the VISN 22 and VISN 21 veterans, 24,420 met the inclusion and exclusion criteria. Of these 24,420 veterans, 2,020 (8.3%) were in VISN 22 and received the EB between December 2014 and August 2015 (QIP-1), and 22,400 (91.7%) were in VISN 21 and did not receive the EB.

Naïve Results Before PS Matching. In the naïve analyses, a larger proportion of EB-exposed vs unexposed veterans discontinued BZD; in addition, reductions were 6.6%, 7.4%, and 9.5% larger for 6 to 9 months, 6 to 12 months, and 1 to 12 months after the index date, respectively (P < .0001 for all comparisons; Table 2).

Propensity Score Matching. Before matching, there were significant differences in baseline characteristics of veterans who met the inclusion and exclusion criteria, with few exceptions (eAppendices 2 and 3, ).

Propensity Score Matching Results. Inspection of PSs revealed good coverage across treatment groups on a histogram plot and a kernel density plot (eAppendices 5 and 6).

Discussion

This QIP was the first to evaluate the impact of an EMPOWER-modeled DTC EB in a large, integrated health care system in the U.S. It was also the first to demonstrate potential benefits of a DTC EB designed for older veterans who are long-term BZD users. In this QIP, which mailed the EB to 3,896 veterans, 1,847 (47.4%) decreased their BZD dose, 458 (11.7%) tapered and then discontinued BZD, and 455 (11.7%) immediately discontinued BZD. The total percentage of veterans who discontinued BZD (23.4%; 913/3,896) was similar to the 27% reported in the EMPOWER trial.¹⁷ However, the risk difference between the 1,316 EB-exposed VISN 22 veterans (QIP-1) and the 1,316 EB-unexposed VISN 21 veterans in this QIP was significantly lower than the 23% risk difference in EMPOWER (though it still demonstrated a significantly larger reduction for EB-exposed veterans).¹⁷

Given this inclusion of all qualifying veterans from the catchment area studied in this QIP, and given the ethical and practical concerns, an RCT was not possible. Therefore, PSM methods were used to balance the covariates across treatment groups and thereby simulate an RCT.^21,22,27 With use of the PSM approach, findings from the descriptive analysis were confirmed and potential selection bias reduced.

Study Limitations

The less robust risk difference found in this QIP has several possible explanations. The authors’ use of a DTC EB coincided with a national VA effort to reduce older veterans’ use of BZDs and other inappropriate medications. For instance, during the study period, academic detailing was being implemented to reduce use of BZDs, particularly in combination with opioids, across VHA facilities and clinics. (Academic detailing is a pharmacy educational outreach program that uses unbiased clinical guidelines to promote physicians’ safety initiatives and align prescribing behavior with best practices.^18-20) However, QIP-2 results and PS analysis of a subgroup of the original sample suggest that EB-exposed veterans were significantly more likely than were their unexposed counterparts were to discontinue BZD. To an extent, this analysis controlled for these other efforts to reduce BZD use in VHA clinics and can be considered a study strength.

Another limitation is the study design, which lacked a control group and did not consider the possibility that some facility or clinic physicians might influence others. Although the region variable was controlled for in PSM, the authors did not capture facility characteristics, including frequency of prescribing BZD and use of a protocol for enforcing the Beers criteria. Such confounders might have influenced outcomes. Unlike the EMPOWER trial,¹⁷ this QIP did not assess or exclude cognitively impaired veterans. It is reasonable to assume that these veterans might not understand some EB messages and consequently might fail to engage their physicians. Failure to initiate discussion with a physician would attenuate the impact of the EB.

Study Strengths

A strength of this QIP was its use of a DTC EB in a large, regional sample of older veterans in a real-world clinical setting. In addition, the study group (EB-exposed veterans) and the comparator group (EB-unexposed veterans) were from similar geographic areas (primarily California and Nevada).

Conclusion

Results of this study suggest that a DTC EB, designed to reduce BZD use among older veterans, was effective in helping patients lower their BZD dose and discontinue BZD. The likelihood of discontinuing BZD 9 and 12 months after the index date was significantly higher for veterans who received an EB modeled on the EMPOWER educational brochure than for a comparator group of veterans who did not receive the EB and were receiving care during the same observation period. In the future, it would be beneficial to use a design that controls for physician exposure to academic detailing focused on BZD reduction and that accounts for the cluster effects of facility practice. Despite these limitations, this QIP is the first real-world empirical example of using an EMPOWER-modeled DTC EB to decrease BZD use among older veterans. Furthermore, these results suggest that a DTC EB can be used to target other high-risk prescription drugs, such as opioids, particularly if alternative treatment options can be provided.

Acknowledgments
Dr. Hauser thanks Cathy, Anika, Katia, and Max Hauser, and Alba and Kevin Quinlan, for their support. In memory of Jirina Hauser, who died on Mother’s Day, May 14, 2017, at the age of 100.

Benzodiazepines (BZDs) are among the most commonly prescribed medications. A recent study found that in 2008, more than 5% of Americans used a BZD, and the percentage was almost 9% among Americans aged ≥ 65 years.^1,2 Among veterans, BZD use is even higher, in part because of the high prevalence of posttraumatic stress disorder (PTSD). One study found that more than 30% of veterans with PTSD received at least 1 BZD prescription.³ The risks associated with BZD treatment for PTSD are compounded by concurrent use of other sedatives and opioids prescribed for co-occurring chronic pain and insomnia.³

Older adults metabolize long-acting BZDs more slowly and generally have an increased sensitivity to the adverse effects (AEs) of all BZDs.⁴ In older adults, BZD use has been associated with cognitive decline, dementia, falls and consequent fractures, and adverse respiratory outcomes.^5-12 The risk of most but not all of these AEs was increased with higher BZD dose or long-term BZD use, which this quality improvement project (QIP) defines as having at least a 60-day supply of BZD prescriptions dispensed within the past year.

Long-term BZD use increases with age. One study found that, among patients receiving a BZD, the rate of long-term BZD use was more than double in older adults (31.4%) than it was in adults aged between 18 and 35 years (14.7%).² For these reasons, the 2012 Beers criteria of the American Geriatrics Society recommend avoiding all types of BZDs in the treatment of insomnia, agitation, or delirium in patients aged > 65 years.¹³ Despite this recommendation, the prevalence of BZD use in older adults remains high.¹⁴

Some innovative approaches have been developed to address the inappropriate use, including overuse and misuse, of BZDs in older adults.¹⁵ In one approach, direct-to-consumer (DTC) information is used to empower patients to collaborate with their physician to manage their health. Results from several studies suggest that providing older patients with information on BZD risks and benefits increases patient–physician interaction and thereby decreases inappropriate BZD use and improves health outcomes.^4,16,17 One study found that perceptions of BZD risks increased 1 week after exposure to a DTC educational brochure (EB), with intention to discuss BZD discontinuation with their physician higher for patients who received the EB than it was for those who did not (83.1% vs 44.3%; P < .0001).¹⁶ The EMPOWER (Eliminating Medications Through Patient Ownership of End Results) cluster randomized controlled trial assessed the effectiveness of a DTC EB focused on BZD risks in older adults.¹⁷ In that seminal study, patients who received a DTC EB were more likely than were comparison patients to discontinue BZD within 6 months (27% vs 5%; risk difference, 23%; 95% CI, 14%-32%).

The Veterans Integrated Systems Network (VISN) 22 Academic Detailing Program is a pharmacy educational outreach program that uses unbiased clinical guidelines to promote physicians’ safety initiatives and align prescribing behavior with best practices.^18-20 With BZD use among older veterans remaining high, the VISN 22 program initiated a clinical QIP modeled on the EMPOWER trial. Veterans in VISN 22 received the DTC EB, which included information on BZD risks and encouraged them to discuss their BZD treatment with their health care provider. VISN 22 was the first VISN in the VHA to implement the EMPOWER protocol.

As this was a QIP, all eligible veterans in VISN 22 were mailed the DTC EB, thus making it difficult to estimate the impact of the EB on BZD discontinuation in this VISN. Therefore, DTC EB efficacy was estimated by comparing BZD discontinuation between VISN 22 and VISN 21, an adjacent VISN that did not mail the DTC EB. To reduce selection bias associated with different controls in the 2 VISNs, the authors performed propensity score matching (PSM) to balance the covariates and provide an unbiased estimate of the mean treatment effect of the DTC EB in VISN 22 veterans who were included in the initial descriptive QIP and received the EB; these veterans were compared with VISN 21 veterans who did not receive the EB.

Methods

Two QIPs were undertaken to determine the impact of DTC EB on BZD use in older veterans in the VHA.

Quality Improvement Project 1

Design. A retrospective cohort analysis was performed. The VISN 22 catchment area, which encompasses VA facilities and clinics in southern California and southern Nevada, serves about 500,000 veterans, a substantial proportion of whom are aged ≥ 65 years. Among these older veterans are active long-term BZD users, who were defined as having ≥ 60-day supply of BZD prescriptions dispensed within the past year. Each active long-term user with a BZD prescription released within 200 days before the index date (the date the user was to meet with the prescribing physician) was mailed an EB 2 to 8 weeks in advance of the visit. Excluded from analysis were veterans with a schizophrenia, spinal cord injury, or seizure disorder diagnosis recorded in both their inpatient and outpatient medical records; veterans seen by Palliative Care within the past year; and veterans who died before analysis was completed.

Education Brochure. The EB for VISN 22 (Figure 1, see

Patients. The sample consisted of all veterans identified as meeting the inclusion criteria and being enrolled in VISN 22. The EB was mailed once to veterans on a rolling basis from December 2014 to February 2016. Change in BZD use was analyzed only after 9 to 24 months had passed since the index appointment with the prescribing physician. This period included 12 weeks for BZD taper and then 6 months after taper.

Analysis. For each veteran, monthly mean lorazepam equivalent (LE) was calculated using as many as 12 fills before the index date. Average daily dose of LE was calculated by dividing the sum of LE from all included prescriptions by total number of days between the first fill and the index date. The BZD prescription fills were evaluated after the index date. Veterans who received at least 1 prescription after the index date but then had no BZD prescription activity in VA clinics for 3 consecutive months during the 9-month observation period were recorded as having tapered and then discontinued BZD. Veterans who had no BZD prescription activity in VA clinics after the index date and during the 9-month observation period were recorded as having discontinued BZD without tapering. For veterans who had BZD prescription activity in VA clinics after the index date and during the 9-month observation period, mean LE was calculated by dividing the total LE for BZD prescriptions after the index date by number of days from the first fill after the index date to the date of analysis.

Quality Improvement Project 2

Design. A retrospective cohort analysis using PSM was performed on a subgroup of the QIP-1 sample to evaluate the impact of EB on BZD prescribing in the VA during 2 periods: 6 to 9 months and 6 to 12 months after the index date. A secondary outcome was discontinuation 1 to 12 months after the index date. Veterans in the analysis were active long-term BZD users, had at least 1 BZD prescription released within 200 days before the index date, were aged ≥ 65 years, and had an appointment scheduled with their BZD prescriber within 2 to 8 weeks (Figure 2). 

Patients. VISN 22 implemented QIP-2, a real-world application of a modified EMPOWER program, by identifying eligible veterans on a rolling basis from December 2014 to August 2015. All veterans who were identified and sent an EB during this period were included in the case group. The index date was defined as the first of the month the EB was mailed. Veterans with a pending appointment were chosen because the lead time would allow them to receive the EB and prepare to discuss it with the physician during the visit.

A comparator group was drawn from the adjacent VISN 21 catchment area, which encompasses VA facilities and clinics in Hawaii, northern California, and northern Nevada. During the observation period, VISN 21 did not mail any EBs specifically addressing BZD risks. Veterans in the comparator group had an appointment scheduled with their BZD prescribing physician within 4 weeks, were aged ≥ 65 years on the index date (first of the month before the next appointment, coinciding with the date EBs were sent to VISN 22 veterans), were active long-term BZD users, and had at least 1 BZD prescription released within 200 days before the index date. All patients were followed for up to 12 months after the index date, with BZD discontinuation recorded 9 and 12 months after the index date.

Propensity Score Matching

Propensity score (PS) was estimated with logistic regression analysis with treatment as the dependent variable and baseline characteristics as the independent variables.^21,22 One-to-one matching on the PS was performed using the nearest neighbor approach without replacements. Independent variables related to outcome but unrelated to EB exposure were selected for PS development.²² These variables included year of birth; male sex; Hispanic ethnicity; annual income; service connection status; region; body mass index; Charlson Comorbidity Index category; total baseline BZD dose; and diagnosis of AIDS, nonmetastatic cancer, metastatic cancer, chronic obstructive pulmonary disease (COPD), congestive heart failure (CHF), dementia, diabetes mellitus (DM), DM with complications, gastroesophageal reflux disease (GERD), general anxiety disorder (GAD), hemiparaplegia, liver disease (mild), liver disease (moderate to severe), myocardial infarction (MI), Parkinson disease, peptic ulcer disease (PUD), psychosis, renal disease, rheumatoid arthritis (RA), or substance use disorder (SUD).

The EMPOWER cluster randomized controlled trial (RCT) demonstrated the effectiveness of EB exposure in a Canadian population of elderly patients who were long-term BZD users.¹⁷ Randomized controlled trials are the gold standard for clinical trials because they can establish causal inference.^23-25 Given ethical and practical concerns, however, RCTs cannot be applied to all clinical scenarios. Although EMPOWER is reported to be an effective tool in reducing BZD use in older adults, its application in a real-world, large, integrated health care system remains untested. Observational studies are often conducted as an alternative to RCTs but are subject to selection bias because of their lack of randomization.²⁶ Therefore, robust research methods are needed to generate unbiased estimates of the impact of an intervention on an outcome. Propensity score matching simulates an RCT by balancing the covariates across treatment groups.^21,22,27 Observed patient characteristics are used to estimate PS, the probability that treatment will be received. Logistic or probit regression is used to balance the potential confounding covariates between the treatment groups.Once PSs are known, mean treatment effect can be estimated without the mean model.²⁸ In other words, PSM methods can be used to generate an unbiased estimate of the treatment.

Propensity Score Analysis

Baseline characteristics were compared using Student t test (continuous variables) and χ² test (discrete variables). Results are presented as means and standard deviations (continuous variables) and frequency and percentage (discrete variables).

The main outcome was BZD discontinuation 9 and 12 months after the index date. A postindex lag of 6 months was used to capture any tapering (Figure 2). Discontinuation, defined as 3 consecutive months of no BZD prescription on hand, was measured for 2 periods: 6 to 9 months and 6 to 12 months after the index date. A secondary outcome was discontinuation 1 to 12 months after the index date. An estimate was made of the difference in the proportions of BZD discontinuers who received the EB and BZD discontinuers who did not receive the EB, where mean treatment (risk difference) was presented as the absolute risk difference with a 95% CI. Standard errors and 95% CIs for the risk differences were generated with biased-corrected CIs from 1,000 bootstrap samples.

Sensitivity Analyses

Naïve multivariate logistic regression analysis was performed to evaluate the association between EB exposure and BZD discontinuation while controlling for potential confounders. Results are presented as odds ratios (ORs) and 95% CIs. Confounders identified were the same covariates used to generate the PSs.

Several analyses were performed to test the sensitivity of the methods applied using PSM by changing caliber size while maintaining the nearest neighbor approach without replacement. Linear regression analysis was performed with robust standard errors to estimate the risk difference of BZD discontinuation between EB-exposed and EB-unexposed veterans.

Statistical significance was set at P < .05. All statistical analyses were performed with Stata/SE Version 13 (College Station, TX).

Results

Quality Improvement Project 1

On a rolling basis from December 2014 to February 2016, the EB was mailed once to 3,896 VISN 22 veterans 2 to 8 weeks before a clinic appointment with their BZD prescribing physician. 

Quality Improvement Project 2

Of all the VISN 22 and VISN 21 veterans, 24,420 met the inclusion and exclusion criteria. Of these 24,420 veterans, 2,020 (8.3%) were in VISN 22 and received the EB between December 2014 and August 2015 (QIP-1), and 22,400 (91.7%) were in VISN 21 and did not receive the EB.

Naïve Results Before PS Matching. In the naïve analyses, a larger proportion of EB-exposed vs unexposed veterans discontinued BZD; in addition, reductions were 6.6%, 7.4%, and 9.5% larger for 6 to 9 months, 6 to 12 months, and 1 to 12 months after the index date, respectively (P < .0001 for all comparisons; Table 2).

Propensity Score Matching. Before matching, there were significant differences in baseline characteristics of veterans who met the inclusion and exclusion criteria, with few exceptions (eAppendices 2 and 3, ).

Propensity Score Matching Results. Inspection of PSs revealed good coverage across treatment groups on a histogram plot and a kernel density plot (eAppendices 5 and 6).

Discussion

This QIP was the first to evaluate the impact of an EMPOWER-modeled DTC EB in a large, integrated health care system in the U.S. It was also the first to demonstrate potential benefits of a DTC EB designed for older veterans who are long-term BZD users. In this QIP, which mailed the EB to 3,896 veterans, 1,847 (47.4%) decreased their BZD dose, 458 (11.7%) tapered and then discontinued BZD, and 455 (11.7%) immediately discontinued BZD. The total percentage of veterans who discontinued BZD (23.4%; 913/3,896) was similar to the 27% reported in the EMPOWER trial.¹⁷ However, the risk difference between the 1,316 EB-exposed VISN 22 veterans (QIP-1) and the 1,316 EB-unexposed VISN 21 veterans in this QIP was significantly lower than the 23% risk difference in EMPOWER (though it still demonstrated a significantly larger reduction for EB-exposed veterans).¹⁷

Given this inclusion of all qualifying veterans from the catchment area studied in this QIP, and given the ethical and practical concerns, an RCT was not possible. Therefore, PSM methods were used to balance the covariates across treatment groups and thereby simulate an RCT.^21,22,27 With use of the PSM approach, findings from the descriptive analysis were confirmed and potential selection bias reduced.

Study Limitations

The less robust risk difference found in this QIP has several possible explanations. The authors’ use of a DTC EB coincided with a national VA effort to reduce older veterans’ use of BZDs and other inappropriate medications. For instance, during the study period, academic detailing was being implemented to reduce use of BZDs, particularly in combination with opioids, across VHA facilities and clinics. (Academic detailing is a pharmacy educational outreach program that uses unbiased clinical guidelines to promote physicians’ safety initiatives and align prescribing behavior with best practices.^18-20) However, QIP-2 results and PS analysis of a subgroup of the original sample suggest that EB-exposed veterans were significantly more likely than were their unexposed counterparts were to discontinue BZD. To an extent, this analysis controlled for these other efforts to reduce BZD use in VHA clinics and can be considered a study strength.

Another limitation is the study design, which lacked a control group and did not consider the possibility that some facility or clinic physicians might influence others. Although the region variable was controlled for in PSM, the authors did not capture facility characteristics, including frequency of prescribing BZD and use of a protocol for enforcing the Beers criteria. Such confounders might have influenced outcomes. Unlike the EMPOWER trial,¹⁷ this QIP did not assess or exclude cognitively impaired veterans. It is reasonable to assume that these veterans might not understand some EB messages and consequently might fail to engage their physicians. Failure to initiate discussion with a physician would attenuate the impact of the EB.

Study Strengths

A strength of this QIP was its use of a DTC EB in a large, regional sample of older veterans in a real-world clinical setting. In addition, the study group (EB-exposed veterans) and the comparator group (EB-unexposed veterans) were from similar geographic areas (primarily California and Nevada).

Conclusion

Results of this study suggest that a DTC EB, designed to reduce BZD use among older veterans, was effective in helping patients lower their BZD dose and discontinue BZD. The likelihood of discontinuing BZD 9 and 12 months after the index date was significantly higher for veterans who received an EB modeled on the EMPOWER educational brochure than for a comparator group of veterans who did not receive the EB and were receiving care during the same observation period. In the future, it would be beneficial to use a design that controls for physician exposure to academic detailing focused on BZD reduction and that accounts for the cluster effects of facility practice. Despite these limitations, this QIP is the first real-world empirical example of using an EMPOWER-modeled DTC EB to decrease BZD use among older veterans. Furthermore, these results suggest that a DTC EB can be used to target other high-risk prescription drugs, such as opioids, particularly if alternative treatment options can be provided.

Acknowledgments
Dr. Hauser thanks Cathy, Anika, Katia, and Max Hauser, and Alba and Kevin Quinlan, for their support. In memory of Jirina Hauser, who died on Mother’s Day, May 14, 2017, at the age of 100.

Hospital readmissions for acute myocardial infarction (AMI), heart failure, and pneumonia have become central to quality-measurement efforts by the Centers for Medicare & Medicaid Services (CMS), which seek to improve hospital care transitions through public reporting and payment programs.¹ Most current measures are limited to readmissions that require inpatient hospitalization and do not capture return visits to the emergency department (ED) that do not result in readmission but rather ED discharge. These visits may reflect important needs for acute, unscheduled care during the vulnerable posthospitalization period.^2-5 While previous research has suggested that nearly 10% of patients may return to the ED following hospital discharge without readmission, the characteristics of these visits among Medicare beneficiaries and the implications for national care-coordination quality-measurement initiatives have not been explored.^6,7

As the locus of acute outpatient care and the primary portal of hospital admissions and readmissions, ED visits following hospital discharge may convey meaningful information about posthospitalization care transitions.^8,9 In addition, recent reviews and perspectives have highlighted the role of ED care-coordination services as interventions to reduce inpatient hospitalizations and improve care transitions,^10,11 yet no empirical studies have evaluated the relationship between these unique care-coordination opportunities in the ED and care-coordination outcomes, such as hospital readmissions. As policymakers seek to develop accountability measures that capture the totality of acute, unscheduled visits following hospital discharge, describing the relationship between ED visits and readmissions will be essential to providers for benchmarking and to policymakers and payers seeking to reduce the total cost of care.^12,13

Accordingly, we sought to characterize the frequency, diagnoses, and hospital-level variation in treat-and-discharge ED visitation following hospital discharge for 3 conditions for which hospital readmission is publicly reported by the CMS: AMI, heart failure, and pneumonia. We also sought to evaluate the relationship between hospital-level ED visitation following hospital discharge and publicly reported, risk-standardized readmission rates (RSRRs).

Study Design

This study was a cross-sectional analysis of Medicare beneficiaries discharged alive following hospitalization for AMI, heart failure, and pneumonia between July 2011 and June 2012.

Selection of Participants

We used Medicare Standard Analytic Files to identify inpatient hospitalizations for each disease cohort based on principal discharge diagnoses. Each condition-specific cohort was constructed to be consistent with the CMS’s readmission measures using International Classification of Diseases, 9th Revision-Clinical Modification codes to identify AMI, heart failure, and pneumonia discharges.¹ We included only patients who were enrolled in fee-for-service (FFS) Medicare parts A and B for 12 months prior to their index hospitalization to maximize the capture of diagnoses for risk adjustment. Each cohort included only patients who were discharged alive while maintaining FFS coverage for at least 30 days following hospital discharge to minimize bias in outcome ascertainment. We excluded patients who were discharged against medical advice. All contiguous admissions that were identified in a transfer chain were considered to be a single admission. Hospitals with fewer than 25 condition-specific index hospital admissions were excluded from this analysis for consistency with publicly reported measures.¹

Measurements

We measured postdischarge, treat-and release ED visits that occurred at any hospital within 30 days of hospital discharge from the index hospitalization. ED visits were identified as a hospital outpatient claim for ED services using hospital outpatient revenue center codes 0450, 0451, 0452, 0456, and 0981. This definition is consistent with those of previous studies.^3,14 We defined postdischarge ED visits as treat-and-discharge visits or visits that did not result in inpatient readmission or observation stays. Similar to readmission measures, only 1 postdischarge ED visit was counted toward the hospital-level outcome in patients with multiple ED visits within the 30 days following hospital discharge. We defined readmission as the first unplanned, inpatient hospitalization occurring at any hospital within the 30-day period following discharge. Any subsequent inpatient admission following the 30-day period was considered a distinct index admission if it met the inclusion criteria. Consistent with CMS methods, unplanned, inpatient readmissions are from any source and are not limited to patients who were first evaluated in the ED.

Outcomes

We describe hospital-level, postdischarge ED visitation as the risk-standardized postdischarge ED visit rate. The general construct of this measure is consistent with those of prior studies that define postdischarge ED visitation as the proportion of index admissions followed by a treat-and-discharge ED visit without hospital readmission^2,3; however, this outcome also incorporates a risk-standardization model with covariates that are identical to the risk-standardization approach that is used for readmission measurement.

We describe hospital-level readmission by calculating RSRRs consistent with CMS readmission measures, which are endorsed by the National Quality Forum and used for public reporting.^15-17 Detailed technical documentation, including the SAS code used to replicate hospital-level measures of readmission, are available publicly through the CMS QualityNet portal.¹⁸

We calculated risk-standardized postdischarge ED visit rates and RSRRs as the ratio of the predicted number of postdischarge ED visits or readmissions for a hospital given its observed case mix to the expected number of postdischarge ED visits or readmissions based on the nation’s performance with that hospital’s case mix, respectively. This approach estimates a distinct risk-standardized postdischarge ED visit rate and RSRR for each hospital using hierarchical generalized linear models (HGLMs) and using a logit link with a first-level adjustment for age, sex, 29 clinical covariates for AMI, 35 clinical covariates for heart failure, and 38 clinical covariates for pneumonia. Each clinical covariate is identified based on inpatient and outpatient claims during the 12 months prior to the index hospitalization. The second level of the HGLM includes a random hospital-level intercept. This approach to measuring hospital readmissions accounts for the correlated nature of observed readmission rates within a hospital and reflects the assumption that after adjustment for patient characteristics and sampling variability, the remaining variation in postdischarge ED visit rates or readmission rates reflects hospital quality.

Analysis

In order to characterize treat-and-discharge postdischarge ED visits, we first described the clinical conditions that were evaluated during the first postdischarge ED visit. Based on the principal discharge diagnosis, ED visits were grouped into clinically meaningful categories using the Agency for Healthcare Research and Quality Clinical Classifications Software (CCS).¹⁹ We also report hospital-level variation in risk-standardized postdischarge ED visit rates for AMI, heart failure, and pneumonia.

Next, we examined the relationship between hospital characteristics and risk-standardized postdischarge ED visit rates. We linked hospital characteristics from the American Hospital Association (AHA) Annual Survey to the study dataset, including the following: safety-net status, teaching status, and urban or rural status. Consistent with prior work, hospital safety-net status was defined as a hospital Medicaid caseload greater than 1 standard deviation above the mean Medicaid caseload in the hospital’s state. Approximately 94% of the hospitals included in the 3 condition cohorts in the dataset had complete data in the 2011 AHA Annual Survey to be included in this analysis.

We evaluated the relationship between postdischarge ED visit rates and hospital readmission rates in 2 ways. First, we calculated Spearman rank correlation coefficients between hospital-level, risk-standardized postdischarge ED visit rates and RSRRs. Second, we calculated hospital-level variation in RSRRs based on the strata of risk-standardized postdischarge ED visit rates. Given the normal distribution of postdischarge ED visit rates, we grouped hospitals by quartile of postdischarge ED visit rates and 1 group for hospitals with no postdischarge ED visits.

Based on preliminary analyses indicating a relationship between hospital size, measured by condition-specific index hospitalization volume, and postdischarge treat-and-discharge ED visit rates, all descriptive statistics and correlations reported are weighted by the volume of condition-specific index hospitalizations. The study was approved by the Yale University Human Research Protection Program. All analyses were conducted using SAS 9.1 (SAS Institute Inc, Cary, NC). The analytic plan and results reported in this work are in compliance with the Strengthening the Reporting of Observational Studies in Epidemiology checklist.²⁰

During the 1-year study period, we included a total of 157,035 patients who were hospitalized at 1656 hospitals for AMI, 391,209 at 3044 hospitals for heart failure, and 342,376 at 3484 hospitals for pneumonia. Details of study cohort creation are available in supplementary Table 1. After hospitalization for AMI, 14,714 patients experienced a postdischarge ED visit (8.4%) and 27,214 an inpatient readmissions (17.3%) within 30 days of discharge; 31,621 (7.6%) and 88,106 (22.5%) patients after hospitalization for heart failure and 26,681 (7.4%) and 59,352 (17.3%) patients after hospitalization for pneumonia experienced a postdischarge ED visit and an inpatient readmission within 30 days of discharge, respectively.

Postdischarge ED visits were for a wide variety of conditions, with the top 10 CCS categories comprising 44% of postdischarge ED visits following AMI hospitalizations, 44% of following heart failure hospitalizations, and 41% following pneumonia hospitalizations (supplementary Table 2). The first postdischarge ED visit was rarely for the same condition as the index hospitalization in the AMI cohort (224 visits; 1.5%) as well as the pneumonia cohort (1401 visits; 5.3%). Among patients who were originally admitted for heart failure, 10.6% of the first postdischarge ED visits were also for congestive heart failure. However, the first postdischarge ED visit was commonly for associated conditions, such as coronary artery disease in the case of AMI or chronic obstructive pulmonary disease in the case of pneumonia, albeit these related conditions did not comprise the majority of postdischarge ED visitation.

We found wide hospital-level variation in postdischarge ED visit rates for each condition: AMI (median: 8.3%; 5th and 95th percentile: 2.8%-14.3%), heart failure (median: 7.3%; 5th and 95th percentile: 3.0%-13.3%), and pneumonia (median: 7.1%; 5th and 95th percentile: 2.4%-13.2%; supplementary Table 3). The variation persisted after accounting for hospital case mix, as evidenced in the supplementary Figure, which describes hospital variation in risk-standardized postdischarge ED visit rates. This variation was statistically significant (P < .001), as demonstrated by the isolated relationship between the random effect and the outcome (AMI: random effect estimate 0.0849 [95% confidence interval (CI), 0.0832 to 0.0866]; heart failure: random effect estimate 0.0796 [95% CI, 0.0784 to 0.0809]; pneumonia: random effect estimate 0.0753 [95% CI, 0.0741 to 0.0764]).

Across all 3 conditions, hospitals located in rural areas had significantly higher risk-standardized postdischarge ED visit rates than hospitals located in urban areas (10.1% vs 8.6% for AMI, 8.4% vs 7.5% for heart failure, and 8.0% vs 7.4% for pneumonia). In comparison to teaching hospitals, nonteaching hospitals had significantly higher risk-standardized postdischarge ED visit rates following hospital discharge for pneumonia (7.6% vs 7.1%). Safety-net hospitals also had higher risk-standardized postdischarge ED visitation rates following discharge for heart failure (8.4% vs 7.7%) and pneumonia (7.7% vs 7.3%). Risk-standardized postdischarge ED visit rates were higher in publicly owned hospitals than in nonprofit or privately owned hospitals for heart failure (8.0% vs 7.5% in nonprofit hospitals or 7.5% in private hospitals) and pneumonia (7.7% vs 7.4% in nonprofit hospitals and 7.3% in private hospitals; Table).

Across a national cohort of Medicare beneficiaries, we found frequent treat-and-discharge ED utilization following hospital discharge for AMI, heart failure, and pneumonia, suggesting that publicly reported readmission measures are capturing only a portion of postdischarge acute-care use. Our findings confirm prior work describing a 30-day postdischarge ED visit rate of 8% to 9% among Medicare beneficiaries for all hospitalizations in several states.^3,6While many of the first postdischarge ED visits were for conditions related to the index hospitalization, the majority represent acute, unscheduled visits for different diagnoses. These findings are consistent with prior work studying inpatient readmissions and observation readmissions that find similar heterogeneity in the clinical reasons for hospital return.^21,22

We also described substantial hospital-level variation in risk-standardized ED postdischarge rates. Prior work by Vashi et al.³ demonstrated substantial variation in observed postdischarge ED visit rates and inpatient readmissions following hospital discharge between clinical conditions in a population-level study. Our work extends upon this by demonstrating hospital-level variation for 3 conditions of high volume and substantial policy importance after accounting for differences in hospital case mix. Interestingly, our work also found similar rates of postdischarge ED treat-and-discharge visitation as recent work by Sabbatini et al.²³ analyzing an all-payer, adult population with any clinical condition. Taken together, these studies show the substantial volume of postdischarge acute-care utilization in the ED not captured by existing readmission measures.

We found several hospital characteristics of importance in describing variation in postdischarge ED visitation rates. Notably, hospitals located in rural areas and safety-net hospitals demonstrated higher postdischarge ED visitation rates. This may reflect a higher use of the ED as an acute, unscheduled care access point in rural communities without access to alternative acute diagnostic and treatment services.²⁴ Similarly, safety-net hospitals may be more likely to provide unscheduled care for patients with poor access to primary care in the ED setting. Yet, consistent with prior work, our results also indicate that these differences do not result in different readmission rates.²⁵ Regarding hospital teaching status, unlike prior work suggesting that teaching hospitals care for more safety-net Medicare beneficiaries,²⁶ our work found opposite patterns of postdischarge ED visitation between hospital teaching and safety-net status following pneumonia hospitalization. This may reflect differences in the organization of acute care as patients with limited access to unscheduled primary and specialty care in safety-net communities utilize the ED, whereas patients in teaching-hospital communities may be able to access hospital-based clinics for care.

Contrary to the expectations of many clinicians and policymakers, we found an inverse relationship between postdischarge ED visit rates and readmission rates. While the cross-sectional design of our study cannot provide a causal explanation, these findings merit policy attention and future exploration of several hypotheses. One possible explanation for this finding is that hospitals with high postdischarge ED visit rates provide care in communities in which acute, unscheduled care is consolidated to the ED setting and thereby permits the ED to serve a gatekeeper function for scarce inpatient resources. This hypothesis may also be supported by recent interventions demonstrating that the use of ED care coordination and geriatric ED services at higher-volume EDs can reduce hospitalizations. Also, hospitals with greater ED capacity may have easier ED access and may be able to see patients earlier in their disease courses post discharge or more frequently in the ED for follow-up, therefore increasing ED visits but avoiding rehospitalization. Another possible explanation is that hospitals with lower postdischarge ED visit rates may also have a lower propensity to admit patients. Because our definition of postdischarge ED visitation did not include ED visits that result in hospitalization, hospitals with a lower propensity to admit from the ED may therefore appear to have higher ED visit rates. This explanation may be further supported by our finding that many postdischarge ED visits are for conditions that are associated with discretionary hospitalization in the ED.²⁷ A third explanation for this finding may be that poor access to outpatient care outside the hospital setting results in higher postdischarge ED visit rates without increasing the acuity of these revisits or increasing readmission rates²⁸; however, given the validated, risk-standardized approach to readmission measurement, this is unlikely. This is also unlikely given recent work by Sabbatini et al.²³ demonstrating substantial acuity among patients who return to the ED following hospital discharge. Future work should seek to evaluate the relationship between the availability of ED care-coordination services and the specific ED, hospital, and community care-coordination activities undertaken in the ED following hospital discharge to reduce readmission rates.

This work should be interpreted within the confines of its design. First, it is possible that some of the variation detected in postdischarge ED visit rates is mediated by hospital-level variation in postdischarge observation visits that are not captured in this outcome. However, in previous work, we have demonstrated that almost one-third of hospitals have no postdischarge observation stays and that most postdischarge observation stays are for more than 24 hours, which is unlikely to reflect the intensity of care of postdischarge ED visits.²⁷ Second, our analyses were limited to Medicare FFS beneficiaries, which may limit the generalizability of this work to other patient populations. However, this dataset did include a national cohort of Medicare beneficiaries that is identical to those included in publicly reported CMS readmission measures; therefore, these results have substantial policy relevance. Third, this work was limited to 3 conditions of high illness severity of policy focus, and future work applying similar analyses to less severe conditions may find different degrees of hospital-level variation in postdischarge outcomes that are amenable to quality improvement. Finally, we assessed the rate of treat-and-discharge ED visits only after hospital discharge; this understates the frequency of ED visits since repeat ED visits and ED visits resulting in rehospitalization are not included. However, our definition was designed to mirror the definition used to assess hospital readmissions for policy purposes and is a conservative approach.

In summary, ED visits following hospital discharge are common, as Medicare beneficiaries have 1 treat-and-discharge ED visit for every 2 readmissions within 30 days of hospital discharge. Postdischarge ED visits occur for a wide variety of conditions, with wide risk-standardized, hospital-level variation. Hospitals with the highest risk-standardized postdischarge ED visitation rates demonstrated lower RSRRs, suggesting that policymakers and researchers should further examine the role of the hospital-based ED in providing access to acute care and supporting care transitions for the vulnerable Medicare population.

Disclosure

 Dr. Venkatesh received contract support from the CMS, an agency of the U.S. Department of Health & Human Services, and grant support from the Emergency Medicine Foundation’s Health Policy Research Scholar Award during the conduct of the study; and Dr. Wang, Mr. Wang, Ms. Altaf, Dr. Bernheim, and Dr. Horwitz received contract support from the CMS, an agency of the U.S. Department of Health & Human Services, during the conduct of the study.

Hospital readmissions for acute myocardial infarction (AMI), heart failure, and pneumonia have become central to quality-measurement efforts by the Centers for Medicare & Medicaid Services (CMS), which seek to improve hospital care transitions through public reporting and payment programs.¹ Most current measures are limited to readmissions that require inpatient hospitalization and do not capture return visits to the emergency department (ED) that do not result in readmission but rather ED discharge. These visits may reflect important needs for acute, unscheduled care during the vulnerable posthospitalization period.^2-5 While previous research has suggested that nearly 10% of patients may return to the ED following hospital discharge without readmission, the characteristics of these visits among Medicare beneficiaries and the implications for national care-coordination quality-measurement initiatives have not been explored.^6,7

As the locus of acute outpatient care and the primary portal of hospital admissions and readmissions, ED visits following hospital discharge may convey meaningful information about posthospitalization care transitions.^8,9 In addition, recent reviews and perspectives have highlighted the role of ED care-coordination services as interventions to reduce inpatient hospitalizations and improve care transitions,^10,11 yet no empirical studies have evaluated the relationship between these unique care-coordination opportunities in the ED and care-coordination outcomes, such as hospital readmissions. As policymakers seek to develop accountability measures that capture the totality of acute, unscheduled visits following hospital discharge, describing the relationship between ED visits and readmissions will be essential to providers for benchmarking and to policymakers and payers seeking to reduce the total cost of care.^12,13

Accordingly, we sought to characterize the frequency, diagnoses, and hospital-level variation in treat-and-discharge ED visitation following hospital discharge for 3 conditions for which hospital readmission is publicly reported by the CMS: AMI, heart failure, and pneumonia. We also sought to evaluate the relationship between hospital-level ED visitation following hospital discharge and publicly reported, risk-standardized readmission rates (RSRRs).

Study Design

This study was a cross-sectional analysis of Medicare beneficiaries discharged alive following hospitalization for AMI, heart failure, and pneumonia between July 2011 and June 2012.

Selection of Participants

We used Medicare Standard Analytic Files to identify inpatient hospitalizations for each disease cohort based on principal discharge diagnoses. Each condition-specific cohort was constructed to be consistent with the CMS’s readmission measures using International Classification of Diseases, 9th Revision-Clinical Modification codes to identify AMI, heart failure, and pneumonia discharges.¹ We included only patients who were enrolled in fee-for-service (FFS) Medicare parts A and B for 12 months prior to their index hospitalization to maximize the capture of diagnoses for risk adjustment. Each cohort included only patients who were discharged alive while maintaining FFS coverage for at least 30 days following hospital discharge to minimize bias in outcome ascertainment. We excluded patients who were discharged against medical advice. All contiguous admissions that were identified in a transfer chain were considered to be a single admission. Hospitals with fewer than 25 condition-specific index hospital admissions were excluded from this analysis for consistency with publicly reported measures.¹

Measurements

We measured postdischarge, treat-and release ED visits that occurred at any hospital within 30 days of hospital discharge from the index hospitalization. ED visits were identified as a hospital outpatient claim for ED services using hospital outpatient revenue center codes 0450, 0451, 0452, 0456, and 0981. This definition is consistent with those of previous studies.^3,14 We defined postdischarge ED visits as treat-and-discharge visits or visits that did not result in inpatient readmission or observation stays. Similar to readmission measures, only 1 postdischarge ED visit was counted toward the hospital-level outcome in patients with multiple ED visits within the 30 days following hospital discharge. We defined readmission as the first unplanned, inpatient hospitalization occurring at any hospital within the 30-day period following discharge. Any subsequent inpatient admission following the 30-day period was considered a distinct index admission if it met the inclusion criteria. Consistent with CMS methods, unplanned, inpatient readmissions are from any source and are not limited to patients who were first evaluated in the ED.

Outcomes

We describe hospital-level, postdischarge ED visitation as the risk-standardized postdischarge ED visit rate. The general construct of this measure is consistent with those of prior studies that define postdischarge ED visitation as the proportion of index admissions followed by a treat-and-discharge ED visit without hospital readmission^2,3; however, this outcome also incorporates a risk-standardization model with covariates that are identical to the risk-standardization approach that is used for readmission measurement.

We describe hospital-level readmission by calculating RSRRs consistent with CMS readmission measures, which are endorsed by the National Quality Forum and used for public reporting.^15-17 Detailed technical documentation, including the SAS code used to replicate hospital-level measures of readmission, are available publicly through the CMS QualityNet portal.¹⁸

We calculated risk-standardized postdischarge ED visit rates and RSRRs as the ratio of the predicted number of postdischarge ED visits or readmissions for a hospital given its observed case mix to the expected number of postdischarge ED visits or readmissions based on the nation’s performance with that hospital’s case mix, respectively. This approach estimates a distinct risk-standardized postdischarge ED visit rate and RSRR for each hospital using hierarchical generalized linear models (HGLMs) and using a logit link with a first-level adjustment for age, sex, 29 clinical covariates for AMI, 35 clinical covariates for heart failure, and 38 clinical covariates for pneumonia. Each clinical covariate is identified based on inpatient and outpatient claims during the 12 months prior to the index hospitalization. The second level of the HGLM includes a random hospital-level intercept. This approach to measuring hospital readmissions accounts for the correlated nature of observed readmission rates within a hospital and reflects the assumption that after adjustment for patient characteristics and sampling variability, the remaining variation in postdischarge ED visit rates or readmission rates reflects hospital quality.

Analysis

In order to characterize treat-and-discharge postdischarge ED visits, we first described the clinical conditions that were evaluated during the first postdischarge ED visit. Based on the principal discharge diagnosis, ED visits were grouped into clinically meaningful categories using the Agency for Healthcare Research and Quality Clinical Classifications Software (CCS).¹⁹ We also report hospital-level variation in risk-standardized postdischarge ED visit rates for AMI, heart failure, and pneumonia.

Next, we examined the relationship between hospital characteristics and risk-standardized postdischarge ED visit rates. We linked hospital characteristics from the American Hospital Association (AHA) Annual Survey to the study dataset, including the following: safety-net status, teaching status, and urban or rural status. Consistent with prior work, hospital safety-net status was defined as a hospital Medicaid caseload greater than 1 standard deviation above the mean Medicaid caseload in the hospital’s state. Approximately 94% of the hospitals included in the 3 condition cohorts in the dataset had complete data in the 2011 AHA Annual Survey to be included in this analysis.

We evaluated the relationship between postdischarge ED visit rates and hospital readmission rates in 2 ways. First, we calculated Spearman rank correlation coefficients between hospital-level, risk-standardized postdischarge ED visit rates and RSRRs. Second, we calculated hospital-level variation in RSRRs based on the strata of risk-standardized postdischarge ED visit rates. Given the normal distribution of postdischarge ED visit rates, we grouped hospitals by quartile of postdischarge ED visit rates and 1 group for hospitals with no postdischarge ED visits.

Based on preliminary analyses indicating a relationship between hospital size, measured by condition-specific index hospitalization volume, and postdischarge treat-and-discharge ED visit rates, all descriptive statistics and correlations reported are weighted by the volume of condition-specific index hospitalizations. The study was approved by the Yale University Human Research Protection Program. All analyses were conducted using SAS 9.1 (SAS Institute Inc, Cary, NC). The analytic plan and results reported in this work are in compliance with the Strengthening the Reporting of Observational Studies in Epidemiology checklist.²⁰

During the 1-year study period, we included a total of 157,035 patients who were hospitalized at 1656 hospitals for AMI, 391,209 at 3044 hospitals for heart failure, and 342,376 at 3484 hospitals for pneumonia. Details of study cohort creation are available in supplementary Table 1. After hospitalization for AMI, 14,714 patients experienced a postdischarge ED visit (8.4%) and 27,214 an inpatient readmissions (17.3%) within 30 days of discharge; 31,621 (7.6%) and 88,106 (22.5%) patients after hospitalization for heart failure and 26,681 (7.4%) and 59,352 (17.3%) patients after hospitalization for pneumonia experienced a postdischarge ED visit and an inpatient readmission within 30 days of discharge, respectively.

Postdischarge ED visits were for a wide variety of conditions, with the top 10 CCS categories comprising 44% of postdischarge ED visits following AMI hospitalizations, 44% of following heart failure hospitalizations, and 41% following pneumonia hospitalizations (supplementary Table 2). The first postdischarge ED visit was rarely for the same condition as the index hospitalization in the AMI cohort (224 visits; 1.5%) as well as the pneumonia cohort (1401 visits; 5.3%). Among patients who were originally admitted for heart failure, 10.6% of the first postdischarge ED visits were also for congestive heart failure. However, the first postdischarge ED visit was commonly for associated conditions, such as coronary artery disease in the case of AMI or chronic obstructive pulmonary disease in the case of pneumonia, albeit these related conditions did not comprise the majority of postdischarge ED visitation.

We found wide hospital-level variation in postdischarge ED visit rates for each condition: AMI (median: 8.3%; 5th and 95th percentile: 2.8%-14.3%), heart failure (median: 7.3%; 5th and 95th percentile: 3.0%-13.3%), and pneumonia (median: 7.1%; 5th and 95th percentile: 2.4%-13.2%; supplementary Table 3). The variation persisted after accounting for hospital case mix, as evidenced in the supplementary Figure, which describes hospital variation in risk-standardized postdischarge ED visit rates. This variation was statistically significant (P < .001), as demonstrated by the isolated relationship between the random effect and the outcome (AMI: random effect estimate 0.0849 [95% confidence interval (CI), 0.0832 to 0.0866]; heart failure: random effect estimate 0.0796 [95% CI, 0.0784 to 0.0809]; pneumonia: random effect estimate 0.0753 [95% CI, 0.0741 to 0.0764]).

Across all 3 conditions, hospitals located in rural areas had significantly higher risk-standardized postdischarge ED visit rates than hospitals located in urban areas (10.1% vs 8.6% for AMI, 8.4% vs 7.5% for heart failure, and 8.0% vs 7.4% for pneumonia). In comparison to teaching hospitals, nonteaching hospitals had significantly higher risk-standardized postdischarge ED visit rates following hospital discharge for pneumonia (7.6% vs 7.1%). Safety-net hospitals also had higher risk-standardized postdischarge ED visitation rates following discharge for heart failure (8.4% vs 7.7%) and pneumonia (7.7% vs 7.3%). Risk-standardized postdischarge ED visit rates were higher in publicly owned hospitals than in nonprofit or privately owned hospitals for heart failure (8.0% vs 7.5% in nonprofit hospitals or 7.5% in private hospitals) and pneumonia (7.7% vs 7.4% in nonprofit hospitals and 7.3% in private hospitals; Table).

Across a national cohort of Medicare beneficiaries, we found frequent treat-and-discharge ED utilization following hospital discharge for AMI, heart failure, and pneumonia, suggesting that publicly reported readmission measures are capturing only a portion of postdischarge acute-care use. Our findings confirm prior work describing a 30-day postdischarge ED visit rate of 8% to 9% among Medicare beneficiaries for all hospitalizations in several states.^3,6While many of the first postdischarge ED visits were for conditions related to the index hospitalization, the majority represent acute, unscheduled visits for different diagnoses. These findings are consistent with prior work studying inpatient readmissions and observation readmissions that find similar heterogeneity in the clinical reasons for hospital return.^21,22

We also described substantial hospital-level variation in risk-standardized ED postdischarge rates. Prior work by Vashi et al.³ demonstrated substantial variation in observed postdischarge ED visit rates and inpatient readmissions following hospital discharge between clinical conditions in a population-level study. Our work extends upon this by demonstrating hospital-level variation for 3 conditions of high volume and substantial policy importance after accounting for differences in hospital case mix. Interestingly, our work also found similar rates of postdischarge ED treat-and-discharge visitation as recent work by Sabbatini et al.²³ analyzing an all-payer, adult population with any clinical condition. Taken together, these studies show the substantial volume of postdischarge acute-care utilization in the ED not captured by existing readmission measures.

We found several hospital characteristics of importance in describing variation in postdischarge ED visitation rates. Notably, hospitals located in rural areas and safety-net hospitals demonstrated higher postdischarge ED visitation rates. This may reflect a higher use of the ED as an acute, unscheduled care access point in rural communities without access to alternative acute diagnostic and treatment services.²⁴ Similarly, safety-net hospitals may be more likely to provide unscheduled care for patients with poor access to primary care in the ED setting. Yet, consistent with prior work, our results also indicate that these differences do not result in different readmission rates.²⁵ Regarding hospital teaching status, unlike prior work suggesting that teaching hospitals care for more safety-net Medicare beneficiaries,²⁶ our work found opposite patterns of postdischarge ED visitation between hospital teaching and safety-net status following pneumonia hospitalization. This may reflect differences in the organization of acute care as patients with limited access to unscheduled primary and specialty care in safety-net communities utilize the ED, whereas patients in teaching-hospital communities may be able to access hospital-based clinics for care.

Contrary to the expectations of many clinicians and policymakers, we found an inverse relationship between postdischarge ED visit rates and readmission rates. While the cross-sectional design of our study cannot provide a causal explanation, these findings merit policy attention and future exploration of several hypotheses. One possible explanation for this finding is that hospitals with high postdischarge ED visit rates provide care in communities in which acute, unscheduled care is consolidated to the ED setting and thereby permits the ED to serve a gatekeeper function for scarce inpatient resources. This hypothesis may also be supported by recent interventions demonstrating that the use of ED care coordination and geriatric ED services at higher-volume EDs can reduce hospitalizations. Also, hospitals with greater ED capacity may have easier ED access and may be able to see patients earlier in their disease courses post discharge or more frequently in the ED for follow-up, therefore increasing ED visits but avoiding rehospitalization. Another possible explanation is that hospitals with lower postdischarge ED visit rates may also have a lower propensity to admit patients. Because our definition of postdischarge ED visitation did not include ED visits that result in hospitalization, hospitals with a lower propensity to admit from the ED may therefore appear to have higher ED visit rates. This explanation may be further supported by our finding that many postdischarge ED visits are for conditions that are associated with discretionary hospitalization in the ED.²⁷ A third explanation for this finding may be that poor access to outpatient care outside the hospital setting results in higher postdischarge ED visit rates without increasing the acuity of these revisits or increasing readmission rates²⁸; however, given the validated, risk-standardized approach to readmission measurement, this is unlikely. This is also unlikely given recent work by Sabbatini et al.²³ demonstrating substantial acuity among patients who return to the ED following hospital discharge. Future work should seek to evaluate the relationship between the availability of ED care-coordination services and the specific ED, hospital, and community care-coordination activities undertaken in the ED following hospital discharge to reduce readmission rates.

This work should be interpreted within the confines of its design. First, it is possible that some of the variation detected in postdischarge ED visit rates is mediated by hospital-level variation in postdischarge observation visits that are not captured in this outcome. However, in previous work, we have demonstrated that almost one-third of hospitals have no postdischarge observation stays and that most postdischarge observation stays are for more than 24 hours, which is unlikely to reflect the intensity of care of postdischarge ED visits.²⁷ Second, our analyses were limited to Medicare FFS beneficiaries, which may limit the generalizability of this work to other patient populations. However, this dataset did include a national cohort of Medicare beneficiaries that is identical to those included in publicly reported CMS readmission measures; therefore, these results have substantial policy relevance. Third, this work was limited to 3 conditions of high illness severity of policy focus, and future work applying similar analyses to less severe conditions may find different degrees of hospital-level variation in postdischarge outcomes that are amenable to quality improvement. Finally, we assessed the rate of treat-and-discharge ED visits only after hospital discharge; this understates the frequency of ED visits since repeat ED visits and ED visits resulting in rehospitalization are not included. However, our definition was designed to mirror the definition used to assess hospital readmissions for policy purposes and is a conservative approach.

In summary, ED visits following hospital discharge are common, as Medicare beneficiaries have 1 treat-and-discharge ED visit for every 2 readmissions within 30 days of hospital discharge. Postdischarge ED visits occur for a wide variety of conditions, with wide risk-standardized, hospital-level variation. Hospitals with the highest risk-standardized postdischarge ED visitation rates demonstrated lower RSRRs, suggesting that policymakers and researchers should further examine the role of the hospital-based ED in providing access to acute care and supporting care transitions for the vulnerable Medicare population.

Disclosure

 Dr. Venkatesh received contract support from the CMS, an agency of the U.S. Department of Health & Human Services, and grant support from the Emergency Medicine Foundation’s Health Policy Research Scholar Award during the conduct of the study; and Dr. Wang, Mr. Wang, Ms. Altaf, Dr. Bernheim, and Dr. Horwitz received contract support from the CMS, an agency of the U.S. Department of Health & Human Services, during the conduct of the study.

Hospital readmissions for acute myocardial infarction (AMI), heart failure, and pneumonia have become central to quality-measurement efforts by the Centers for Medicare & Medicaid Services (CMS), which seek to improve hospital care transitions through public reporting and payment programs.¹ Most current measures are limited to readmissions that require inpatient hospitalization and do not capture return visits to the emergency department (ED) that do not result in readmission but rather ED discharge. These visits may reflect important needs for acute, unscheduled care during the vulnerable posthospitalization period.^2-5 While previous research has suggested that nearly 10% of patients may return to the ED following hospital discharge without readmission, the characteristics of these visits among Medicare beneficiaries and the implications for national care-coordination quality-measurement initiatives have not been explored.^6,7

As the locus of acute outpatient care and the primary portal of hospital admissions and readmissions, ED visits following hospital discharge may convey meaningful information about posthospitalization care transitions.^8,9 In addition, recent reviews and perspectives have highlighted the role of ED care-coordination services as interventions to reduce inpatient hospitalizations and improve care transitions,^10,11 yet no empirical studies have evaluated the relationship between these unique care-coordination opportunities in the ED and care-coordination outcomes, such as hospital readmissions. As policymakers seek to develop accountability measures that capture the totality of acute, unscheduled visits following hospital discharge, describing the relationship between ED visits and readmissions will be essential to providers for benchmarking and to policymakers and payers seeking to reduce the total cost of care.^12,13

Accordingly, we sought to characterize the frequency, diagnoses, and hospital-level variation in treat-and-discharge ED visitation following hospital discharge for 3 conditions for which hospital readmission is publicly reported by the CMS: AMI, heart failure, and pneumonia. We also sought to evaluate the relationship between hospital-level ED visitation following hospital discharge and publicly reported, risk-standardized readmission rates (RSRRs).

Study Design

This study was a cross-sectional analysis of Medicare beneficiaries discharged alive following hospitalization for AMI, heart failure, and pneumonia between July 2011 and June 2012.

Selection of Participants

We used Medicare Standard Analytic Files to identify inpatient hospitalizations for each disease cohort based on principal discharge diagnoses. Each condition-specific cohort was constructed to be consistent with the CMS’s readmission measures using International Classification of Diseases, 9th Revision-Clinical Modification codes to identify AMI, heart failure, and pneumonia discharges.¹ We included only patients who were enrolled in fee-for-service (FFS) Medicare parts A and B for 12 months prior to their index hospitalization to maximize the capture of diagnoses for risk adjustment. Each cohort included only patients who were discharged alive while maintaining FFS coverage for at least 30 days following hospital discharge to minimize bias in outcome ascertainment. We excluded patients who were discharged against medical advice. All contiguous admissions that were identified in a transfer chain were considered to be a single admission. Hospitals with fewer than 25 condition-specific index hospital admissions were excluded from this analysis for consistency with publicly reported measures.¹

Measurements

We measured postdischarge, treat-and release ED visits that occurred at any hospital within 30 days of hospital discharge from the index hospitalization. ED visits were identified as a hospital outpatient claim for ED services using hospital outpatient revenue center codes 0450, 0451, 0452, 0456, and 0981. This definition is consistent with those of previous studies.^3,14 We defined postdischarge ED visits as treat-and-discharge visits or visits that did not result in inpatient readmission or observation stays. Similar to readmission measures, only 1 postdischarge ED visit was counted toward the hospital-level outcome in patients with multiple ED visits within the 30 days following hospital discharge. We defined readmission as the first unplanned, inpatient hospitalization occurring at any hospital within the 30-day period following discharge. Any subsequent inpatient admission following the 30-day period was considered a distinct index admission if it met the inclusion criteria. Consistent with CMS methods, unplanned, inpatient readmissions are from any source and are not limited to patients who were first evaluated in the ED.

Outcomes

We describe hospital-level, postdischarge ED visitation as the risk-standardized postdischarge ED visit rate. The general construct of this measure is consistent with those of prior studies that define postdischarge ED visitation as the proportion of index admissions followed by a treat-and-discharge ED visit without hospital readmission^2,3; however, this outcome also incorporates a risk-standardization model with covariates that are identical to the risk-standardization approach that is used for readmission measurement.

We describe hospital-level readmission by calculating RSRRs consistent with CMS readmission measures, which are endorsed by the National Quality Forum and used for public reporting.^15-17 Detailed technical documentation, including the SAS code used to replicate hospital-level measures of readmission, are available publicly through the CMS QualityNet portal.¹⁸

We calculated risk-standardized postdischarge ED visit rates and RSRRs as the ratio of the predicted number of postdischarge ED visits or readmissions for a hospital given its observed case mix to the expected number of postdischarge ED visits or readmissions based on the nation’s performance with that hospital’s case mix, respectively. This approach estimates a distinct risk-standardized postdischarge ED visit rate and RSRR for each hospital using hierarchical generalized linear models (HGLMs) and using a logit link with a first-level adjustment for age, sex, 29 clinical covariates for AMI, 35 clinical covariates for heart failure, and 38 clinical covariates for pneumonia. Each clinical covariate is identified based on inpatient and outpatient claims during the 12 months prior to the index hospitalization. The second level of the HGLM includes a random hospital-level intercept. This approach to measuring hospital readmissions accounts for the correlated nature of observed readmission rates within a hospital and reflects the assumption that after adjustment for patient characteristics and sampling variability, the remaining variation in postdischarge ED visit rates or readmission rates reflects hospital quality.

Analysis

In order to characterize treat-and-discharge postdischarge ED visits, we first described the clinical conditions that were evaluated during the first postdischarge ED visit. Based on the principal discharge diagnosis, ED visits were grouped into clinically meaningful categories using the Agency for Healthcare Research and Quality Clinical Classifications Software (CCS).¹⁹ We also report hospital-level variation in risk-standardized postdischarge ED visit rates for AMI, heart failure, and pneumonia.

Next, we examined the relationship between hospital characteristics and risk-standardized postdischarge ED visit rates. We linked hospital characteristics from the American Hospital Association (AHA) Annual Survey to the study dataset, including the following: safety-net status, teaching status, and urban or rural status. Consistent with prior work, hospital safety-net status was defined as a hospital Medicaid caseload greater than 1 standard deviation above the mean Medicaid caseload in the hospital’s state. Approximately 94% of the hospitals included in the 3 condition cohorts in the dataset had complete data in the 2011 AHA Annual Survey to be included in this analysis.

We evaluated the relationship between postdischarge ED visit rates and hospital readmission rates in 2 ways. First, we calculated Spearman rank correlation coefficients between hospital-level, risk-standardized postdischarge ED visit rates and RSRRs. Second, we calculated hospital-level variation in RSRRs based on the strata of risk-standardized postdischarge ED visit rates. Given the normal distribution of postdischarge ED visit rates, we grouped hospitals by quartile of postdischarge ED visit rates and 1 group for hospitals with no postdischarge ED visits.

Based on preliminary analyses indicating a relationship between hospital size, measured by condition-specific index hospitalization volume, and postdischarge treat-and-discharge ED visit rates, all descriptive statistics and correlations reported are weighted by the volume of condition-specific index hospitalizations. The study was approved by the Yale University Human Research Protection Program. All analyses were conducted using SAS 9.1 (SAS Institute Inc, Cary, NC). The analytic plan and results reported in this work are in compliance with the Strengthening the Reporting of Observational Studies in Epidemiology checklist.²⁰

During the 1-year study period, we included a total of 157,035 patients who were hospitalized at 1656 hospitals for AMI, 391,209 at 3044 hospitals for heart failure, and 342,376 at 3484 hospitals for pneumonia. Details of study cohort creation are available in supplementary Table 1. After hospitalization for AMI, 14,714 patients experienced a postdischarge ED visit (8.4%) and 27,214 an inpatient readmissions (17.3%) within 30 days of discharge; 31,621 (7.6%) and 88,106 (22.5%) patients after hospitalization for heart failure and 26,681 (7.4%) and 59,352 (17.3%) patients after hospitalization for pneumonia experienced a postdischarge ED visit and an inpatient readmission within 30 days of discharge, respectively.

Postdischarge ED visits were for a wide variety of conditions, with the top 10 CCS categories comprising 44% of postdischarge ED visits following AMI hospitalizations, 44% of following heart failure hospitalizations, and 41% following pneumonia hospitalizations (supplementary Table 2). The first postdischarge ED visit was rarely for the same condition as the index hospitalization in the AMI cohort (224 visits; 1.5%) as well as the pneumonia cohort (1401 visits; 5.3%). Among patients who were originally admitted for heart failure, 10.6% of the first postdischarge ED visits were also for congestive heart failure. However, the first postdischarge ED visit was commonly for associated conditions, such as coronary artery disease in the case of AMI or chronic obstructive pulmonary disease in the case of pneumonia, albeit these related conditions did not comprise the majority of postdischarge ED visitation.

We found wide hospital-level variation in postdischarge ED visit rates for each condition: AMI (median: 8.3%; 5th and 95th percentile: 2.8%-14.3%), heart failure (median: 7.3%; 5th and 95th percentile: 3.0%-13.3%), and pneumonia (median: 7.1%; 5th and 95th percentile: 2.4%-13.2%; supplementary Table 3). The variation persisted after accounting for hospital case mix, as evidenced in the supplementary Figure, which describes hospital variation in risk-standardized postdischarge ED visit rates. This variation was statistically significant (P < .001), as demonstrated by the isolated relationship between the random effect and the outcome (AMI: random effect estimate 0.0849 [95% confidence interval (CI), 0.0832 to 0.0866]; heart failure: random effect estimate 0.0796 [95% CI, 0.0784 to 0.0809]; pneumonia: random effect estimate 0.0753 [95% CI, 0.0741 to 0.0764]).

Across all 3 conditions, hospitals located in rural areas had significantly higher risk-standardized postdischarge ED visit rates than hospitals located in urban areas (10.1% vs 8.6% for AMI, 8.4% vs 7.5% for heart failure, and 8.0% vs 7.4% for pneumonia). In comparison to teaching hospitals, nonteaching hospitals had significantly higher risk-standardized postdischarge ED visit rates following hospital discharge for pneumonia (7.6% vs 7.1%). Safety-net hospitals also had higher risk-standardized postdischarge ED visitation rates following discharge for heart failure (8.4% vs 7.7%) and pneumonia (7.7% vs 7.3%). Risk-standardized postdischarge ED visit rates were higher in publicly owned hospitals than in nonprofit or privately owned hospitals for heart failure (8.0% vs 7.5% in nonprofit hospitals or 7.5% in private hospitals) and pneumonia (7.7% vs 7.4% in nonprofit hospitals and 7.3% in private hospitals; Table).

Across a national cohort of Medicare beneficiaries, we found frequent treat-and-discharge ED utilization following hospital discharge for AMI, heart failure, and pneumonia, suggesting that publicly reported readmission measures are capturing only a portion of postdischarge acute-care use. Our findings confirm prior work describing a 30-day postdischarge ED visit rate of 8% to 9% among Medicare beneficiaries for all hospitalizations in several states.^3,6While many of the first postdischarge ED visits were for conditions related to the index hospitalization, the majority represent acute, unscheduled visits for different diagnoses. These findings are consistent with prior work studying inpatient readmissions and observation readmissions that find similar heterogeneity in the clinical reasons for hospital return.^21,22

We also described substantial hospital-level variation in risk-standardized ED postdischarge rates. Prior work by Vashi et al.³ demonstrated substantial variation in observed postdischarge ED visit rates and inpatient readmissions following hospital discharge between clinical conditions in a population-level study. Our work extends upon this by demonstrating hospital-level variation for 3 conditions of high volume and substantial policy importance after accounting for differences in hospital case mix. Interestingly, our work also found similar rates of postdischarge ED treat-and-discharge visitation as recent work by Sabbatini et al.²³ analyzing an all-payer, adult population with any clinical condition. Taken together, these studies show the substantial volume of postdischarge acute-care utilization in the ED not captured by existing readmission measures.

We found several hospital characteristics of importance in describing variation in postdischarge ED visitation rates. Notably, hospitals located in rural areas and safety-net hospitals demonstrated higher postdischarge ED visitation rates. This may reflect a higher use of the ED as an acute, unscheduled care access point in rural communities without access to alternative acute diagnostic and treatment services.²⁴ Similarly, safety-net hospitals may be more likely to provide unscheduled care for patients with poor access to primary care in the ED setting. Yet, consistent with prior work, our results also indicate that these differences do not result in different readmission rates.²⁵ Regarding hospital teaching status, unlike prior work suggesting that teaching hospitals care for more safety-net Medicare beneficiaries,²⁶ our work found opposite patterns of postdischarge ED visitation between hospital teaching and safety-net status following pneumonia hospitalization. This may reflect differences in the organization of acute care as patients with limited access to unscheduled primary and specialty care in safety-net communities utilize the ED, whereas patients in teaching-hospital communities may be able to access hospital-based clinics for care.

Contrary to the expectations of many clinicians and policymakers, we found an inverse relationship between postdischarge ED visit rates and readmission rates. While the cross-sectional design of our study cannot provide a causal explanation, these findings merit policy attention and future exploration of several hypotheses. One possible explanation for this finding is that hospitals with high postdischarge ED visit rates provide care in communities in which acute, unscheduled care is consolidated to the ED setting and thereby permits the ED to serve a gatekeeper function for scarce inpatient resources. This hypothesis may also be supported by recent interventions demonstrating that the use of ED care coordination and geriatric ED services at higher-volume EDs can reduce hospitalizations. Also, hospitals with greater ED capacity may have easier ED access and may be able to see patients earlier in their disease courses post discharge or more frequently in the ED for follow-up, therefore increasing ED visits but avoiding rehospitalization. Another possible explanation is that hospitals with lower postdischarge ED visit rates may also have a lower propensity to admit patients. Because our definition of postdischarge ED visitation did not include ED visits that result in hospitalization, hospitals with a lower propensity to admit from the ED may therefore appear to have higher ED visit rates. This explanation may be further supported by our finding that many postdischarge ED visits are for conditions that are associated with discretionary hospitalization in the ED.²⁷ A third explanation for this finding may be that poor access to outpatient care outside the hospital setting results in higher postdischarge ED visit rates without increasing the acuity of these revisits or increasing readmission rates²⁸; however, given the validated, risk-standardized approach to readmission measurement, this is unlikely. This is also unlikely given recent work by Sabbatini et al.²³ demonstrating substantial acuity among patients who return to the ED following hospital discharge. Future work should seek to evaluate the relationship between the availability of ED care-coordination services and the specific ED, hospital, and community care-coordination activities undertaken in the ED following hospital discharge to reduce readmission rates.

This work should be interpreted within the confines of its design. First, it is possible that some of the variation detected in postdischarge ED visit rates is mediated by hospital-level variation in postdischarge observation visits that are not captured in this outcome. However, in previous work, we have demonstrated that almost one-third of hospitals have no postdischarge observation stays and that most postdischarge observation stays are for more than 24 hours, which is unlikely to reflect the intensity of care of postdischarge ED visits.²⁷ Second, our analyses were limited to Medicare FFS beneficiaries, which may limit the generalizability of this work to other patient populations. However, this dataset did include a national cohort of Medicare beneficiaries that is identical to those included in publicly reported CMS readmission measures; therefore, these results have substantial policy relevance. Third, this work was limited to 3 conditions of high illness severity of policy focus, and future work applying similar analyses to less severe conditions may find different degrees of hospital-level variation in postdischarge outcomes that are amenable to quality improvement. Finally, we assessed the rate of treat-and-discharge ED visits only after hospital discharge; this understates the frequency of ED visits since repeat ED visits and ED visits resulting in rehospitalization are not included. However, our definition was designed to mirror the definition used to assess hospital readmissions for policy purposes and is a conservative approach.

In summary, ED visits following hospital discharge are common, as Medicare beneficiaries have 1 treat-and-discharge ED visit for every 2 readmissions within 30 days of hospital discharge. Postdischarge ED visits occur for a wide variety of conditions, with wide risk-standardized, hospital-level variation. Hospitals with the highest risk-standardized postdischarge ED visitation rates demonstrated lower RSRRs, suggesting that policymakers and researchers should further examine the role of the hospital-based ED in providing access to acute care and supporting care transitions for the vulnerable Medicare population.

Disclosure

 Dr. Venkatesh received contract support from the CMS, an agency of the U.S. Department of Health & Human Services, and grant support from the Emergency Medicine Foundation’s Health Policy Research Scholar Award during the conduct of the study; and Dr. Wang, Mr. Wang, Ms. Altaf, Dr. Bernheim, and Dr. Horwitz received contract support from the CMS, an agency of the U.S. Department of Health & Human Services, during the conduct of the study.

A central tenet of modern medicine is that patients must provide fully informed consent to receive or refuse medical care offered by their clinical teams.^1–4 If a patient is unable to make and communicate a choice or clearly indicate an understanding of the information presented, then he or she is considered to lack the capacity to make medical decisions and the medical team must seek consent from the patient’s surrogate decision-maker.^2-7 Every U.S. state recognizes a patient’s healthcare proxy (HCP) and a court-appointed guardian as a legally recognized surrogate.^8,9 Most of the states also have statutes or regulations establishing a hierarchy of legally recognized surrogate decision-makers in the absence of a HCP or a court-appointed guardian, such as spouses, adult children, parents, siblings, and grandparents.^8,10 In states that do not have such a statute, hospitals develop their own institutional policies for surrogate decision-making.

However, there are important limitations on the authority of these surrogate decision-makers.¹⁰ For instance, patients may not have a family member or a friend to serve as a surrogate decision-maker, often family members cannot override a patient’s objection, even when that patient lacks decision-making capacity, and certain decisions require a guardian or a HCP.^8-10 In these circumstances, the hospital must petition a court to appoint a guardian as a legally recognized surrogate decision-maker. This can be an involved family member, if one exists, or an independent, typically volunteer, guardian.¹¹ The process of guardian appointment is complex^7,11 and can range from a few days to more than a month, largely dependent on court dates and finding a volunteer guardian. Much of the process occurs during the patient’s hospital stay. This prolongation of hospitalization would be expected to increase health care costs and iatrogenic complications,^12–14 but data quantifying these for patients requiring guardianship are lacking. The goal of this study was to describe the characteristics of patients who undergo the process of guardianship and measure the associated burdens. These burdens include the financial costs to the medical system, the prolonged length of stay beyond medical necessity, and the costs to the patient in the form of hospital-acquired complications. Investigating the burden of guardianship is an important first step in uncovering opportunities to improve the process. We hypothesized that patients requiring guardianship would have lengths of stay and healthcare costs that were at least as large as those for patients whose conditions required similar durations of hospitalization prior to medical clearance, in part due to iatrogenic complications that would accrue while awaiting guardian appointment.

Setting

We conducted a retrospective matched cohort study of adult inpatients at Beth Israel Deaconess Medical Center (BIDMC), a 651-bed academic, tertiary care facility in Boston, MA. The study was approved by the BIDMC Institutional Review Board as a nonhuman subject research consistent with hospital operations.

Population

For this matched cohort study, we identified case patients as those hospitalized for any reason for whom guardianship proceedings were initiated and obtained; only the first hospitalization during which the guardianship was pursued was used. Cases were identified by obtaining the data of all patients for whom the BIDMC general counsel completed the process of guardianship between October 2014 and September 2015. At BIDMC, all the guardianship proceedings are referred to the general counsel.

To determine the postclearance experience for referred patients compared with that for other patients with similar lengths of stay up to those of the referred patients’ point of clearance, we identified up to three matched controls for each case (Supplemental Figure 1). Medical clearance was defined as the date when the patient was medically stable to be discharged from the hospital, and it was determined in an iterative manner. We identified controls as hospitalized patients admitted for any cause and matched to the cases requiring guardianship on discharging service and length of stay prior to clearance. Specifically, we identified patients on the same service as the case whose length of stay was at least as long as the length of stay of the case patient until medical clearance, as defined below. We then determined the total and the excess length of stay, defined as the duration beyond clearance for each case referred for guardianship; for controls, the ‘excess’ length of stay was the number of hospitalized days beyond the corresponding time that a matched case had been provided clearance. To account for seasonal influences and the training level of house officers, we selected the three controls whose discharge date was closest (before or after) to the discharge date of their matched case.

From legal team files, we identified 61 patients hospitalized at BIDMC for whom new guardianship was pursued to completion. Of these 61 patients, 10 could not be matched to an appropriate control and were included in descriptive analyses but not in comparisons with controls.

Covariates and Outcomes

We collected the details regarding age, gender, primary language, highest level of education, marital status, insurance status, race, date of admission, date of discharge, discharge disposition, principal diagnosis, case mix index (CMI), and discharging service from our administrative and billing data. Outcomes of interest included length of stay and total hospital charges that were collected from the same databases. We used hospital charges, rather than payments, to ensure uniformity across payers.

Chart Review

Unique to cases, a team of two medical residents (JP, RP) and a hospitalist (DR) determined the date of medical clearance and hospital-associated complications by a chart review. The date of medical clearance was then used to calculate excess length of stay, ie, the duration of stay beyond the date of medical clearance, by subtracting the time to medical clearance from the total inpatient length of stay.

We developed a novel algorithm to determine the date of medical clearance consistently (Figure 1). We first determined whether the discharge summary indicated a clear date of medical readiness for discharge. If the discharge summary was unclear, then a case management or a social work note was used. The date of medical clearance determined by the case management or the social work note was then confirmed with clinical data. The date was confirmed if there were no significant laboratory orders and major medication changes or procedures for 24 h from the date identified. If notes were also inconclusive, then the medical clearance was determined by a review of provider order entry. Medical readiness for discharge was then defined as the first day when there were no laboratory orders for 48 h and no significant medication changes, imaging studies, or microbiologic orders.

• Catheter-associated urinary tract infection (CAUTI): active Foley catheter order and positive urine culture that resulted in antibiotic administration.
• Hospital-acquired pneumonia (HAP): chest X-ray or computed tomography (CT) scan showing a consolidation that resulted in antibiotic administration.
• Venous thromboembolism (VTE): positive venous ultrasound or CT angiography of the chest for deep venous thrombosis (DVT) or pulmonary embolism (PE).
• Decubitus ulcer: new wound care consultation for sacral decubitus ulceration.
• Clostridium difficile (C. diff) infection: positive stool polymerase chain reaction that resulted in antibiotic administration.

The algorithm for identifying the date of clearance and the presence of complications was piloted independently by three investigators (RP, JP, DR) using a single chart review and was redesigned until a consensus was obtained. The same three investigators then independently reviewed three additional charts, including all notes, laboratory results, imaging results, and orders, with complete agreement for both date of clearance and presence of complications. Two investigators (RP, JP) then individually reviewed the remaining 57 charts. Of these, 10 were selected a priori for review by both investigators for interrater reliability, with a mean difference of 0.5 days in the estimated time to clearance and complete concordance in complications. In addition, a third investigator (DR) independently reread 5 of the 57 reviewed charts, with complete concordance in both time to clearance and presence of complications with the original readings.

Statistical Analysis

SAS 9.3 was used for all analyses (SAS Institute Inc., Cary, NC, USA).

We first examined the demographic and clinical characteristics of all 61 patients who underwent guardianship proceedings. Second, we described the primary outcomes of interest–length of stay, costs, and likelihood of complications–in this series of patients with associated 95% confidence intervals.

Third, we examined the associations between guardianship and length of stay and healthcare costs using generalized estimating equations with clustering by matched set and compound symmetry. For length of stay, we specifically assessed excess length of stay (the matching variable) to avoid immortal time bias; we also examined the total length of stay. For all regression analyses, we adjusted for the following covariates: age, gender, education, marital status, race/ethnicity, CMI, insurance status, discharging service, and principal diagnosis. To maximize normality of residuals, costs were log-transformed; length of stay beyond clearance was log-transformed after addition of 1. For both outcomes, we back-transformed the regression coefficients and presented percent change between case and control patients. All reported tests are two-sided.

A total of 61 guardianship cases and 118 controls were included in the analysis.

General Characteristics

The characteristics of all cases prior to matching are included in Table 1. The department of internal medicine discharged the largest proportion of cases, followed by neurosurgery and neurology departments. More than 65% of cases were insured by Medicare or Medicaid. Three-quarters of cases were discharged from the hospital to another medical facility, with about half discharged to a skilled nursing facility (SNF) or a rehabilitation center and one-quarter to a long-term acute care hospital (LTACH).

The median length of stay for patients requiring guardianship was 28 (range, 23-36) days, and the median total charges were $171,083 ($106,897-$245,281), with a total cost approximating $10.9 million for these patients. Regarding hospital-acquired complications, 10 (16%; 95% confidence interval, 8%–28%) unique cases suffered from a complication, with HAP being the most frequently (n = 5) occurring complication.

Comparison with Matched Controls

No statistically significant differences were observed between cases and controls in terms of age, primary language, highest level of education, ethnicity, insurance status, or discharging service as shown in Table 2; discharging service was a matched variable and comparable by design. However, cases tended to be less likely to be married and had a higher CMI.

When compared with control patients in terms of similar services who stayed for at least as long as their duration to clearance, the cases had significantly longer lengths of stay compared to those of controls (29 total days compared to 18 days, P < .001; Figure 2). In addition, cases incurred significantly higher median total charges ($168,666) compared to those of controls ($104,190; P = .02).

After accounting for potential confounders, including age, gender, language, education, marital status, discharging service, ethnicity, insurance status, CMI, and principal diagnosis, guardianship was associated with 58% higher excess length of stay (P = .04, 95% CI [2%-145%]). Furthermore, guardianship was associated with 23% higher total charges (P = .02, 95% CI [4%-46%]) and 37% longer total length of stay (P = .002, 95% CI [12%-67%]).

To our knowledge, this is among the first studies to investigate healthcare costs and harm to the patient in the form of hospital-associated complications as a result of guardianship proceedings. Other studies^15,16 have also demonstrated excessive length of stay attributed to nonclinical factors such as guardianship, though they did not quantify the excess stay or compare guardianship cases with a matched control. One study¹⁷ demonstrated total charges of $150,000 per patient requiring guardianship, which are similar to our results. However, Chen et al. also observed an average of 27.8 medically unnecessary days, which are 16 more days than those in our study sample. This may reflect the difference in how excess days were determined, namely, statistical process control analysis in the previous study compared with a manual chart review in our study. To our knowledge, no other study has compared guardianship cases with matched controls to compare their experiences to patients with similarly prolonged stays prior to clearance.

After matching by service and the length of stay until medical clearance in each guardianship case, the subsequent length of stay was higher among cases than among controls, even after adjustment for differences in CMI and diagnosis. This suggests that the process of obtaining guardianship results in a particularly prolonged length of stay, which is presumably attributable to factors other than medical complexity or ongoing illness.

It is probable that at least two interrelated mechanisms are responsible for the particularly high costs and the long stay of patients who require guardianship. First, the process of obtaining guardianship is itself protracted in several cases, necessitating long-term admissions well beyond the point of medical stability. Second, our results suggest that longer hospital stays are apt to grow further in a feed-forward cycle due to hospital-acquired complications that develop after the date of medical clearance. Indeed, in our series, 16% of patients sustained a complication that is readily attributable to hospital care after their date of clearance, and these types of complications are likely to lengthen the stay even further.

We compared cases referred for guardianship to control patients on the same services, at similar time points, whose length of stay was at least as long as the point of medical clearance as their corresponding case patient. Because cases were hospitalized with active medical needs to at least the point of clearance, we anticipated that costs might well be lower among cases, who had no medical necessity for hospitalization at the point of clearance, compared with controls who remained hospitalized presumably for active medical needs. Counter to this hypothesis, and accounting for potentially confounding variables, undergoing a guardianship proceeding was associated with nearly 25% higher costs of patient care. This may ultimately represent a substantial burden on the healthcare system. For example, in just 1 year in our hospital, the total hospital charges reached almost $11 million for the 61 patients who underwent guardianship proceedings. Considering that 65% of the patients requiring guardianship had Medicaid or Medicare coverage, there are significant financial implications for the hospital systems and to the public.

Limitations of our study relate to its retrospective nature at a single center. Investigating guardianship cases at a single center and with a small sample size of 61 patients limits generalizability. Nevertheless, we still had enough power to detect significant differences compared with matched controls, and this study remains the largest investigation into the cost associated with guardianship to date and the only study comparing guardianship cases with matched controls. Furthermore, we did not complete chart reviews of controls, which limits direct comparisons of complications and precluded our matching on variables that required detailed review.

The retrospective design may include confounders unaccounted for in our statistical design, though we attempted to match cases with controls to account for some of these potential differences and included a broad set of covariates that included measures of comorbidity and diagnosis. To this point, we included only CMI and principal diagnosis as the measures of severity, and adjustment for CMI, which includes features of the index hospitalization itself, may represent overadjustment. However, this type of overadjustment would tend to bias toward the null hypothesis.

Investigators only completed chart reviews for cases, which limits our ability to contrast the rate of hospital-associated complications for cases with that of controls. However, the rates of CAUTI and HAP complications among our cases were notably higher than national inpatient estimates, ie, 5% and 8% compared to 0.2%¹⁸ and 0.5%-1%,¹⁹ respectively. Furthermore, we demonstrated higher total costs and total lengths of stay among guardianship patients, analyses for which the attributed date of clearance for controls was not required, and the rate of complications among the case patients was sizable despite their being formally medically cleared. In other words, regardless of whether a complication rate of 16% is “typical” for inpatients hospitalized for these durations, this suggests that persistent hospitalization after clearance does not carry a benign prognosis.

In addition, to estimate healthcare costs, we relied on total hospital charges, which are readily available and reflect, at least in part, payer costs but do not reflect true costs to the medical center. Nonetheless, charges approximately reflect costs–with some variation across cost centers–and hence provide a useful metric for comparing cases and controls. To provide context, for academic medical centers such as ours, costs are typically about half of charges.

Finally, each state has different statutes for surrogate decision-making. The results of this study reflect the Massachusetts’ experience, with no public guardianship program or hierarchy statute. That being said, while this presumably causes the need for more guardianships in Massachusetts, the mechanisms for guardianship are broadly similar nationwide and are likely to result in excessive length of stay and cost similar to those in our population, as demonstrated in studies from other states.^7,15–17

Implications

At a time where medical systems are searching for opportunities to reduce the length of stay, prevent unnecessary hospitalization, and improve the quality of care, reevaluating the guardianship process is ripe with opportunity. In this single academic center, the process of guardianship was associated with 58% excess length of stay and 23% higher total hospital charges. Furthermore, one in six patients requiring guardianship suffered from hospital-associated complications.

This matched cohort study adds quantitative data demonstrating substantial burdens to the healthcare system as a result of the guardianship process and can be used as an impetus for hospital administration and legal systems to expedite the process. Potential improvements include increasing HCP form completions (which would eliminate the need to pursue guardianship for most of such patients), identifying patients who lack a legally recognized surrogate decision-maker earlier in their hospital stay (ideally upon admission), and providing resources to assist clinical teams in the completion of affidavits necessary to support the appointment of a guardian, so that paperwork can be filed with courts sooner. Further research that provides more generalizable prospective data could potentially improve the guardianship process and reduce its burden on hospitals and patients even further.

Acknowledgments

The authors express their tremendous thanks to Gail Piatkowski for her invaluable assistance in collecting administrative and billing data.

Disclosures 

The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the article

A central tenet of modern medicine is that patients must provide fully informed consent to receive or refuse medical care offered by their clinical teams.^1–4 If a patient is unable to make and communicate a choice or clearly indicate an understanding of the information presented, then he or she is considered to lack the capacity to make medical decisions and the medical team must seek consent from the patient’s surrogate decision-maker.^2-7 Every U.S. state recognizes a patient’s healthcare proxy (HCP) and a court-appointed guardian as a legally recognized surrogate.^8,9 Most of the states also have statutes or regulations establishing a hierarchy of legally recognized surrogate decision-makers in the absence of a HCP or a court-appointed guardian, such as spouses, adult children, parents, siblings, and grandparents.^8,10 In states that do not have such a statute, hospitals develop their own institutional policies for surrogate decision-making.

However, there are important limitations on the authority of these surrogate decision-makers.¹⁰ For instance, patients may not have a family member or a friend to serve as a surrogate decision-maker, often family members cannot override a patient’s objection, even when that patient lacks decision-making capacity, and certain decisions require a guardian or a HCP.^8-10 In these circumstances, the hospital must petition a court to appoint a guardian as a legally recognized surrogate decision-maker. This can be an involved family member, if one exists, or an independent, typically volunteer, guardian.¹¹ The process of guardian appointment is complex^7,11 and can range from a few days to more than a month, largely dependent on court dates and finding a volunteer guardian. Much of the process occurs during the patient’s hospital stay. This prolongation of hospitalization would be expected to increase health care costs and iatrogenic complications,^12–14 but data quantifying these for patients requiring guardianship are lacking. The goal of this study was to describe the characteristics of patients who undergo the process of guardianship and measure the associated burdens. These burdens include the financial costs to the medical system, the prolonged length of stay beyond medical necessity, and the costs to the patient in the form of hospital-acquired complications. Investigating the burden of guardianship is an important first step in uncovering opportunities to improve the process. We hypothesized that patients requiring guardianship would have lengths of stay and healthcare costs that were at least as large as those for patients whose conditions required similar durations of hospitalization prior to medical clearance, in part due to iatrogenic complications that would accrue while awaiting guardian appointment.

Setting

We conducted a retrospective matched cohort study of adult inpatients at Beth Israel Deaconess Medical Center (BIDMC), a 651-bed academic, tertiary care facility in Boston, MA. The study was approved by the BIDMC Institutional Review Board as a nonhuman subject research consistent with hospital operations.

Population

For this matched cohort study, we identified case patients as those hospitalized for any reason for whom guardianship proceedings were initiated and obtained; only the first hospitalization during which the guardianship was pursued was used. Cases were identified by obtaining the data of all patients for whom the BIDMC general counsel completed the process of guardianship between October 2014 and September 2015. At BIDMC, all the guardianship proceedings are referred to the general counsel.

To determine the postclearance experience for referred patients compared with that for other patients with similar lengths of stay up to those of the referred patients’ point of clearance, we identified up to three matched controls for each case (Supplemental Figure 1). Medical clearance was defined as the date when the patient was medically stable to be discharged from the hospital, and it was determined in an iterative manner. We identified controls as hospitalized patients admitted for any cause and matched to the cases requiring guardianship on discharging service and length of stay prior to clearance. Specifically, we identified patients on the same service as the case whose length of stay was at least as long as the length of stay of the case patient until medical clearance, as defined below. We then determined the total and the excess length of stay, defined as the duration beyond clearance for each case referred for guardianship; for controls, the ‘excess’ length of stay was the number of hospitalized days beyond the corresponding time that a matched case had been provided clearance. To account for seasonal influences and the training level of house officers, we selected the three controls whose discharge date was closest (before or after) to the discharge date of their matched case.

From legal team files, we identified 61 patients hospitalized at BIDMC for whom new guardianship was pursued to completion. Of these 61 patients, 10 could not be matched to an appropriate control and were included in descriptive analyses but not in comparisons with controls.

Covariates and Outcomes

We collected the details regarding age, gender, primary language, highest level of education, marital status, insurance status, race, date of admission, date of discharge, discharge disposition, principal diagnosis, case mix index (CMI), and discharging service from our administrative and billing data. Outcomes of interest included length of stay and total hospital charges that were collected from the same databases. We used hospital charges, rather than payments, to ensure uniformity across payers.

Chart Review

Unique to cases, a team of two medical residents (JP, RP) and a hospitalist (DR) determined the date of medical clearance and hospital-associated complications by a chart review. The date of medical clearance was then used to calculate excess length of stay, ie, the duration of stay beyond the date of medical clearance, by subtracting the time to medical clearance from the total inpatient length of stay.

We developed a novel algorithm to determine the date of medical clearance consistently (Figure 1). We first determined whether the discharge summary indicated a clear date of medical readiness for discharge. If the discharge summary was unclear, then a case management or a social work note was used. The date of medical clearance determined by the case management or the social work note was then confirmed with clinical data. The date was confirmed if there were no significant laboratory orders and major medication changes or procedures for 24 h from the date identified. If notes were also inconclusive, then the medical clearance was determined by a review of provider order entry. Medical readiness for discharge was then defined as the first day when there were no laboratory orders for 48 h and no significant medication changes, imaging studies, or microbiologic orders.

• Catheter-associated urinary tract infection (CAUTI): active Foley catheter order and positive urine culture that resulted in antibiotic administration.
• Hospital-acquired pneumonia (HAP): chest X-ray or computed tomography (CT) scan showing a consolidation that resulted in antibiotic administration.
• Venous thromboembolism (VTE): positive venous ultrasound or CT angiography of the chest for deep venous thrombosis (DVT) or pulmonary embolism (PE).
• Decubitus ulcer: new wound care consultation for sacral decubitus ulceration.
• Clostridium difficile (C. diff) infection: positive stool polymerase chain reaction that resulted in antibiotic administration.

The algorithm for identifying the date of clearance and the presence of complications was piloted independently by three investigators (RP, JP, DR) using a single chart review and was redesigned until a consensus was obtained. The same three investigators then independently reviewed three additional charts, including all notes, laboratory results, imaging results, and orders, with complete agreement for both date of clearance and presence of complications. Two investigators (RP, JP) then individually reviewed the remaining 57 charts. Of these, 10 were selected a priori for review by both investigators for interrater reliability, with a mean difference of 0.5 days in the estimated time to clearance and complete concordance in complications. In addition, a third investigator (DR) independently reread 5 of the 57 reviewed charts, with complete concordance in both time to clearance and presence of complications with the original readings.

Statistical Analysis

SAS 9.3 was used for all analyses (SAS Institute Inc., Cary, NC, USA).

We first examined the demographic and clinical characteristics of all 61 patients who underwent guardianship proceedings. Second, we described the primary outcomes of interest–length of stay, costs, and likelihood of complications–in this series of patients with associated 95% confidence intervals.

Third, we examined the associations between guardianship and length of stay and healthcare costs using generalized estimating equations with clustering by matched set and compound symmetry. For length of stay, we specifically assessed excess length of stay (the matching variable) to avoid immortal time bias; we also examined the total length of stay. For all regression analyses, we adjusted for the following covariates: age, gender, education, marital status, race/ethnicity, CMI, insurance status, discharging service, and principal diagnosis. To maximize normality of residuals, costs were log-transformed; length of stay beyond clearance was log-transformed after addition of 1. For both outcomes, we back-transformed the regression coefficients and presented percent change between case and control patients. All reported tests are two-sided.

A total of 61 guardianship cases and 118 controls were included in the analysis.

General Characteristics

The characteristics of all cases prior to matching are included in Table 1. The department of internal medicine discharged the largest proportion of cases, followed by neurosurgery and neurology departments. More than 65% of cases were insured by Medicare or Medicaid. Three-quarters of cases were discharged from the hospital to another medical facility, with about half discharged to a skilled nursing facility (SNF) or a rehabilitation center and one-quarter to a long-term acute care hospital (LTACH).

The median length of stay for patients requiring guardianship was 28 (range, 23-36) days, and the median total charges were $171,083 ($106,897-$245,281), with a total cost approximating $10.9 million for these patients. Regarding hospital-acquired complications, 10 (16%; 95% confidence interval, 8%–28%) unique cases suffered from a complication, with HAP being the most frequently (n = 5) occurring complication.

Comparison with Matched Controls

No statistically significant differences were observed between cases and controls in terms of age, primary language, highest level of education, ethnicity, insurance status, or discharging service as shown in Table 2; discharging service was a matched variable and comparable by design. However, cases tended to be less likely to be married and had a higher CMI.

When compared with control patients in terms of similar services who stayed for at least as long as their duration to clearance, the cases had significantly longer lengths of stay compared to those of controls (29 total days compared to 18 days, P < .001; Figure 2). In addition, cases incurred significantly higher median total charges ($168,666) compared to those of controls ($104,190; P = .02).

After accounting for potential confounders, including age, gender, language, education, marital status, discharging service, ethnicity, insurance status, CMI, and principal diagnosis, guardianship was associated with 58% higher excess length of stay (P = .04, 95% CI [2%-145%]). Furthermore, guardianship was associated with 23% higher total charges (P = .02, 95% CI [4%-46%]) and 37% longer total length of stay (P = .002, 95% CI [12%-67%]).

To our knowledge, this is among the first studies to investigate healthcare costs and harm to the patient in the form of hospital-associated complications as a result of guardianship proceedings. Other studies^15,16 have also demonstrated excessive length of stay attributed to nonclinical factors such as guardianship, though they did not quantify the excess stay or compare guardianship cases with a matched control. One study¹⁷ demonstrated total charges of $150,000 per patient requiring guardianship, which are similar to our results. However, Chen et al. also observed an average of 27.8 medically unnecessary days, which are 16 more days than those in our study sample. This may reflect the difference in how excess days were determined, namely, statistical process control analysis in the previous study compared with a manual chart review in our study. To our knowledge, no other study has compared guardianship cases with matched controls to compare their experiences to patients with similarly prolonged stays prior to clearance.

After matching by service and the length of stay until medical clearance in each guardianship case, the subsequent length of stay was higher among cases than among controls, even after adjustment for differences in CMI and diagnosis. This suggests that the process of obtaining guardianship results in a particularly prolonged length of stay, which is presumably attributable to factors other than medical complexity or ongoing illness.

It is probable that at least two interrelated mechanisms are responsible for the particularly high costs and the long stay of patients who require guardianship. First, the process of obtaining guardianship is itself protracted in several cases, necessitating long-term admissions well beyond the point of medical stability. Second, our results suggest that longer hospital stays are apt to grow further in a feed-forward cycle due to hospital-acquired complications that develop after the date of medical clearance. Indeed, in our series, 16% of patients sustained a complication that is readily attributable to hospital care after their date of clearance, and these types of complications are likely to lengthen the stay even further.

We compared cases referred for guardianship to control patients on the same services, at similar time points, whose length of stay was at least as long as the point of medical clearance as their corresponding case patient. Because cases were hospitalized with active medical needs to at least the point of clearance, we anticipated that costs might well be lower among cases, who had no medical necessity for hospitalization at the point of clearance, compared with controls who remained hospitalized presumably for active medical needs. Counter to this hypothesis, and accounting for potentially confounding variables, undergoing a guardianship proceeding was associated with nearly 25% higher costs of patient care. This may ultimately represent a substantial burden on the healthcare system. For example, in just 1 year in our hospital, the total hospital charges reached almost $11 million for the 61 patients who underwent guardianship proceedings. Considering that 65% of the patients requiring guardianship had Medicaid or Medicare coverage, there are significant financial implications for the hospital systems and to the public.

Limitations of our study relate to its retrospective nature at a single center. Investigating guardianship cases at a single center and with a small sample size of 61 patients limits generalizability. Nevertheless, we still had enough power to detect significant differences compared with matched controls, and this study remains the largest investigation into the cost associated with guardianship to date and the only study comparing guardianship cases with matched controls. Furthermore, we did not complete chart reviews of controls, which limits direct comparisons of complications and precluded our matching on variables that required detailed review.

The retrospective design may include confounders unaccounted for in our statistical design, though we attempted to match cases with controls to account for some of these potential differences and included a broad set of covariates that included measures of comorbidity and diagnosis. To this point, we included only CMI and principal diagnosis as the measures of severity, and adjustment for CMI, which includes features of the index hospitalization itself, may represent overadjustment. However, this type of overadjustment would tend to bias toward the null hypothesis.

Investigators only completed chart reviews for cases, which limits our ability to contrast the rate of hospital-associated complications for cases with that of controls. However, the rates of CAUTI and HAP complications among our cases were notably higher than national inpatient estimates, ie, 5% and 8% compared to 0.2%¹⁸ and 0.5%-1%,¹⁹ respectively. Furthermore, we demonstrated higher total costs and total lengths of stay among guardianship patients, analyses for which the attributed date of clearance for controls was not required, and the rate of complications among the case patients was sizable despite their being formally medically cleared. In other words, regardless of whether a complication rate of 16% is “typical” for inpatients hospitalized for these durations, this suggests that persistent hospitalization after clearance does not carry a benign prognosis.

In addition, to estimate healthcare costs, we relied on total hospital charges, which are readily available and reflect, at least in part, payer costs but do not reflect true costs to the medical center. Nonetheless, charges approximately reflect costs–with some variation across cost centers–and hence provide a useful metric for comparing cases and controls. To provide context, for academic medical centers such as ours, costs are typically about half of charges.

Finally, each state has different statutes for surrogate decision-making. The results of this study reflect the Massachusetts’ experience, with no public guardianship program or hierarchy statute. That being said, while this presumably causes the need for more guardianships in Massachusetts, the mechanisms for guardianship are broadly similar nationwide and are likely to result in excessive length of stay and cost similar to those in our population, as demonstrated in studies from other states.^7,15–17

Implications

At a time where medical systems are searching for opportunities to reduce the length of stay, prevent unnecessary hospitalization, and improve the quality of care, reevaluating the guardianship process is ripe with opportunity. In this single academic center, the process of guardianship was associated with 58% excess length of stay and 23% higher total hospital charges. Furthermore, one in six patients requiring guardianship suffered from hospital-associated complications.

This matched cohort study adds quantitative data demonstrating substantial burdens to the healthcare system as a result of the guardianship process and can be used as an impetus for hospital administration and legal systems to expedite the process. Potential improvements include increasing HCP form completions (which would eliminate the need to pursue guardianship for most of such patients), identifying patients who lack a legally recognized surrogate decision-maker earlier in their hospital stay (ideally upon admission), and providing resources to assist clinical teams in the completion of affidavits necessary to support the appointment of a guardian, so that paperwork can be filed with courts sooner. Further research that provides more generalizable prospective data could potentially improve the guardianship process and reduce its burden on hospitals and patients even further.

Acknowledgments

The authors express their tremendous thanks to Gail Piatkowski for her invaluable assistance in collecting administrative and billing data.

Disclosures 

The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the article

A central tenet of modern medicine is that patients must provide fully informed consent to receive or refuse medical care offered by their clinical teams.^1–4 If a patient is unable to make and communicate a choice or clearly indicate an understanding of the information presented, then he or she is considered to lack the capacity to make medical decisions and the medical team must seek consent from the patient’s surrogate decision-maker.^2-7 Every U.S. state recognizes a patient’s healthcare proxy (HCP) and a court-appointed guardian as a legally recognized surrogate.^8,9 Most of the states also have statutes or regulations establishing a hierarchy of legally recognized surrogate decision-makers in the absence of a HCP or a court-appointed guardian, such as spouses, adult children, parents, siblings, and grandparents.^8,10 In states that do not have such a statute, hospitals develop their own institutional policies for surrogate decision-making.

However, there are important limitations on the authority of these surrogate decision-makers.¹⁰ For instance, patients may not have a family member or a friend to serve as a surrogate decision-maker, often family members cannot override a patient’s objection, even when that patient lacks decision-making capacity, and certain decisions require a guardian or a HCP.^8-10 In these circumstances, the hospital must petition a court to appoint a guardian as a legally recognized surrogate decision-maker. This can be an involved family member, if one exists, or an independent, typically volunteer, guardian.¹¹ The process of guardian appointment is complex^7,11 and can range from a few days to more than a month, largely dependent on court dates and finding a volunteer guardian. Much of the process occurs during the patient’s hospital stay. This prolongation of hospitalization would be expected to increase health care costs and iatrogenic complications,^12–14 but data quantifying these for patients requiring guardianship are lacking. The goal of this study was to describe the characteristics of patients who undergo the process of guardianship and measure the associated burdens. These burdens include the financial costs to the medical system, the prolonged length of stay beyond medical necessity, and the costs to the patient in the form of hospital-acquired complications. Investigating the burden of guardianship is an important first step in uncovering opportunities to improve the process. We hypothesized that patients requiring guardianship would have lengths of stay and healthcare costs that were at least as large as those for patients whose conditions required similar durations of hospitalization prior to medical clearance, in part due to iatrogenic complications that would accrue while awaiting guardian appointment.

Setting

We conducted a retrospective matched cohort study of adult inpatients at Beth Israel Deaconess Medical Center (BIDMC), a 651-bed academic, tertiary care facility in Boston, MA. The study was approved by the BIDMC Institutional Review Board as a nonhuman subject research consistent with hospital operations.

Population

For this matched cohort study, we identified case patients as those hospitalized for any reason for whom guardianship proceedings were initiated and obtained; only the first hospitalization during which the guardianship was pursued was used. Cases were identified by obtaining the data of all patients for whom the BIDMC general counsel completed the process of guardianship between October 2014 and September 2015. At BIDMC, all the guardianship proceedings are referred to the general counsel.

To determine the postclearance experience for referred patients compared with that for other patients with similar lengths of stay up to those of the referred patients’ point of clearance, we identified up to three matched controls for each case (Supplemental Figure 1). Medical clearance was defined as the date when the patient was medically stable to be discharged from the hospital, and it was determined in an iterative manner. We identified controls as hospitalized patients admitted for any cause and matched to the cases requiring guardianship on discharging service and length of stay prior to clearance. Specifically, we identified patients on the same service as the case whose length of stay was at least as long as the length of stay of the case patient until medical clearance, as defined below. We then determined the total and the excess length of stay, defined as the duration beyond clearance for each case referred for guardianship; for controls, the ‘excess’ length of stay was the number of hospitalized days beyond the corresponding time that a matched case had been provided clearance. To account for seasonal influences and the training level of house officers, we selected the three controls whose discharge date was closest (before or after) to the discharge date of their matched case.

From legal team files, we identified 61 patients hospitalized at BIDMC for whom new guardianship was pursued to completion. Of these 61 patients, 10 could not be matched to an appropriate control and were included in descriptive analyses but not in comparisons with controls.

Covariates and Outcomes

We collected the details regarding age, gender, primary language, highest level of education, marital status, insurance status, race, date of admission, date of discharge, discharge disposition, principal diagnosis, case mix index (CMI), and discharging service from our administrative and billing data. Outcomes of interest included length of stay and total hospital charges that were collected from the same databases. We used hospital charges, rather than payments, to ensure uniformity across payers.

Chart Review

Unique to cases, a team of two medical residents (JP, RP) and a hospitalist (DR) determined the date of medical clearance and hospital-associated complications by a chart review. The date of medical clearance was then used to calculate excess length of stay, ie, the duration of stay beyond the date of medical clearance, by subtracting the time to medical clearance from the total inpatient length of stay.

We developed a novel algorithm to determine the date of medical clearance consistently (Figure 1). We first determined whether the discharge summary indicated a clear date of medical readiness for discharge. If the discharge summary was unclear, then a case management or a social work note was used. The date of medical clearance determined by the case management or the social work note was then confirmed with clinical data. The date was confirmed if there were no significant laboratory orders and major medication changes or procedures for 24 h from the date identified. If notes were also inconclusive, then the medical clearance was determined by a review of provider order entry. Medical readiness for discharge was then defined as the first day when there were no laboratory orders for 48 h and no significant medication changes, imaging studies, or microbiologic orders.

• Catheter-associated urinary tract infection (CAUTI): active Foley catheter order and positive urine culture that resulted in antibiotic administration.
• Hospital-acquired pneumonia (HAP): chest X-ray or computed tomography (CT) scan showing a consolidation that resulted in antibiotic administration.
• Venous thromboembolism (VTE): positive venous ultrasound or CT angiography of the chest for deep venous thrombosis (DVT) or pulmonary embolism (PE).
• Decubitus ulcer: new wound care consultation for sacral decubitus ulceration.
• Clostridium difficile (C. diff) infection: positive stool polymerase chain reaction that resulted in antibiotic administration.

The algorithm for identifying the date of clearance and the presence of complications was piloted independently by three investigators (RP, JP, DR) using a single chart review and was redesigned until a consensus was obtained. The same three investigators then independently reviewed three additional charts, including all notes, laboratory results, imaging results, and orders, with complete agreement for both date of clearance and presence of complications. Two investigators (RP, JP) then individually reviewed the remaining 57 charts. Of these, 10 were selected a priori for review by both investigators for interrater reliability, with a mean difference of 0.5 days in the estimated time to clearance and complete concordance in complications. In addition, a third investigator (DR) independently reread 5 of the 57 reviewed charts, with complete concordance in both time to clearance and presence of complications with the original readings.

Statistical Analysis

SAS 9.3 was used for all analyses (SAS Institute Inc., Cary, NC, USA).

We first examined the demographic and clinical characteristics of all 61 patients who underwent guardianship proceedings. Second, we described the primary outcomes of interest–length of stay, costs, and likelihood of complications–in this series of patients with associated 95% confidence intervals.

Third, we examined the associations between guardianship and length of stay and healthcare costs using generalized estimating equations with clustering by matched set and compound symmetry. For length of stay, we specifically assessed excess length of stay (the matching variable) to avoid immortal time bias; we also examined the total length of stay. For all regression analyses, we adjusted for the following covariates: age, gender, education, marital status, race/ethnicity, CMI, insurance status, discharging service, and principal diagnosis. To maximize normality of residuals, costs were log-transformed; length of stay beyond clearance was log-transformed after addition of 1. For both outcomes, we back-transformed the regression coefficients and presented percent change between case and control patients. All reported tests are two-sided.

A total of 61 guardianship cases and 118 controls were included in the analysis.

General Characteristics

The characteristics of all cases prior to matching are included in Table 1. The department of internal medicine discharged the largest proportion of cases, followed by neurosurgery and neurology departments. More than 65% of cases were insured by Medicare or Medicaid. Three-quarters of cases were discharged from the hospital to another medical facility, with about half discharged to a skilled nursing facility (SNF) or a rehabilitation center and one-quarter to a long-term acute care hospital (LTACH).

The median length of stay for patients requiring guardianship was 28 (range, 23-36) days, and the median total charges were $171,083 ($106,897-$245,281), with a total cost approximating $10.9 million for these patients. Regarding hospital-acquired complications, 10 (16%; 95% confidence interval, 8%–28%) unique cases suffered from a complication, with HAP being the most frequently (n = 5) occurring complication.

Comparison with Matched Controls

No statistically significant differences were observed between cases and controls in terms of age, primary language, highest level of education, ethnicity, insurance status, or discharging service as shown in Table 2; discharging service was a matched variable and comparable by design. However, cases tended to be less likely to be married and had a higher CMI.

When compared with control patients in terms of similar services who stayed for at least as long as their duration to clearance, the cases had significantly longer lengths of stay compared to those of controls (29 total days compared to 18 days, P < .001; Figure 2). In addition, cases incurred significantly higher median total charges ($168,666) compared to those of controls ($104,190; P = .02).

After accounting for potential confounders, including age, gender, language, education, marital status, discharging service, ethnicity, insurance status, CMI, and principal diagnosis, guardianship was associated with 58% higher excess length of stay (P = .04, 95% CI [2%-145%]). Furthermore, guardianship was associated with 23% higher total charges (P = .02, 95% CI [4%-46%]) and 37% longer total length of stay (P = .002, 95% CI [12%-67%]).

To our knowledge, this is among the first studies to investigate healthcare costs and harm to the patient in the form of hospital-associated complications as a result of guardianship proceedings. Other studies^15,16 have also demonstrated excessive length of stay attributed to nonclinical factors such as guardianship, though they did not quantify the excess stay or compare guardianship cases with a matched control. One study¹⁷ demonstrated total charges of $150,000 per patient requiring guardianship, which are similar to our results. However, Chen et al. also observed an average of 27.8 medically unnecessary days, which are 16 more days than those in our study sample. This may reflect the difference in how excess days were determined, namely, statistical process control analysis in the previous study compared with a manual chart review in our study. To our knowledge, no other study has compared guardianship cases with matched controls to compare their experiences to patients with similarly prolonged stays prior to clearance.

After matching by service and the length of stay until medical clearance in each guardianship case, the subsequent length of stay was higher among cases than among controls, even after adjustment for differences in CMI and diagnosis. This suggests that the process of obtaining guardianship results in a particularly prolonged length of stay, which is presumably attributable to factors other than medical complexity or ongoing illness.

It is probable that at least two interrelated mechanisms are responsible for the particularly high costs and the long stay of patients who require guardianship. First, the process of obtaining guardianship is itself protracted in several cases, necessitating long-term admissions well beyond the point of medical stability. Second, our results suggest that longer hospital stays are apt to grow further in a feed-forward cycle due to hospital-acquired complications that develop after the date of medical clearance. Indeed, in our series, 16% of patients sustained a complication that is readily attributable to hospital care after their date of clearance, and these types of complications are likely to lengthen the stay even further.

We compared cases referred for guardianship to control patients on the same services, at similar time points, whose length of stay was at least as long as the point of medical clearance as their corresponding case patient. Because cases were hospitalized with active medical needs to at least the point of clearance, we anticipated that costs might well be lower among cases, who had no medical necessity for hospitalization at the point of clearance, compared with controls who remained hospitalized presumably for active medical needs. Counter to this hypothesis, and accounting for potentially confounding variables, undergoing a guardianship proceeding was associated with nearly 25% higher costs of patient care. This may ultimately represent a substantial burden on the healthcare system. For example, in just 1 year in our hospital, the total hospital charges reached almost $11 million for the 61 patients who underwent guardianship proceedings. Considering that 65% of the patients requiring guardianship had Medicaid or Medicare coverage, there are significant financial implications for the hospital systems and to the public.

Limitations of our study relate to its retrospective nature at a single center. Investigating guardianship cases at a single center and with a small sample size of 61 patients limits generalizability. Nevertheless, we still had enough power to detect significant differences compared with matched controls, and this study remains the largest investigation into the cost associated with guardianship to date and the only study comparing guardianship cases with matched controls. Furthermore, we did not complete chart reviews of controls, which limits direct comparisons of complications and precluded our matching on variables that required detailed review.

The retrospective design may include confounders unaccounted for in our statistical design, though we attempted to match cases with controls to account for some of these potential differences and included a broad set of covariates that included measures of comorbidity and diagnosis. To this point, we included only CMI and principal diagnosis as the measures of severity, and adjustment for CMI, which includes features of the index hospitalization itself, may represent overadjustment. However, this type of overadjustment would tend to bias toward the null hypothesis.

Investigators only completed chart reviews for cases, which limits our ability to contrast the rate of hospital-associated complications for cases with that of controls. However, the rates of CAUTI and HAP complications among our cases were notably higher than national inpatient estimates, ie, 5% and 8% compared to 0.2%¹⁸ and 0.5%-1%,¹⁹ respectively. Furthermore, we demonstrated higher total costs and total lengths of stay among guardianship patients, analyses for which the attributed date of clearance for controls was not required, and the rate of complications among the case patients was sizable despite their being formally medically cleared. In other words, regardless of whether a complication rate of 16% is “typical” for inpatients hospitalized for these durations, this suggests that persistent hospitalization after clearance does not carry a benign prognosis.

In addition, to estimate healthcare costs, we relied on total hospital charges, which are readily available and reflect, at least in part, payer costs but do not reflect true costs to the medical center. Nonetheless, charges approximately reflect costs–with some variation across cost centers–and hence provide a useful metric for comparing cases and controls. To provide context, for academic medical centers such as ours, costs are typically about half of charges.

Finally, each state has different statutes for surrogate decision-making. The results of this study reflect the Massachusetts’ experience, with no public guardianship program or hierarchy statute. That being said, while this presumably causes the need for more guardianships in Massachusetts, the mechanisms for guardianship are broadly similar nationwide and are likely to result in excessive length of stay and cost similar to those in our population, as demonstrated in studies from other states.^7,15–17

Implications

At a time where medical systems are searching for opportunities to reduce the length of stay, prevent unnecessary hospitalization, and improve the quality of care, reevaluating the guardianship process is ripe with opportunity. In this single academic center, the process of guardianship was associated with 58% excess length of stay and 23% higher total hospital charges. Furthermore, one in six patients requiring guardianship suffered from hospital-associated complications.

This matched cohort study adds quantitative data demonstrating substantial burdens to the healthcare system as a result of the guardianship process and can be used as an impetus for hospital administration and legal systems to expedite the process. Potential improvements include increasing HCP form completions (which would eliminate the need to pursue guardianship for most of such patients), identifying patients who lack a legally recognized surrogate decision-maker earlier in their hospital stay (ideally upon admission), and providing resources to assist clinical teams in the completion of affidavits necessary to support the appointment of a guardian, so that paperwork can be filed with courts sooner. Further research that provides more generalizable prospective data could potentially improve the guardianship process and reduce its burden on hospitals and patients even further.

Acknowledgments

The authors express their tremendous thanks to Gail Piatkowski for her invaluable assistance in collecting administrative and billing data.

Disclosures 

The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the article

Hospitals are increasingly assessed comparatively in terms of costs and quality for benchmarking purposes. These comparisons can be used by patients and families to determine where to seek care, to report compliance and grant certifications by oversight organizations (eg, Leapfrog, Magnet, Joint Commission), and by payers, to determine reimbursement models and/or to assess financial penalty or bonuses for underperforming or overperforming hospitals. As these efforts can cause substantial reputational and financial consequences for hospitals, these metrics must be contextualized within the population of patients that each hospital serves.

In adult Medicare patient populations, methods have been developed to assess the relative severity of a hospital’s full complement of patients.^1,2 These methods assume a relationship between severity and hospital resource intensity (ie, cost) and typically assume the form of relative weights (RWs), which are developed for clinically similar groups of patients (eg, Medicare Diagnosis Related Groups; MS-DRG) from a reference population. A RW for each MS-DRG is calculated as the average cost of patients within the group divided by the average cost for all patients in the reference population. These weights are then applied to a hospital’s discharges over a specific time period and averaged to obtain a hospital-level case-mix index (CMI). A value of 1 indicates that a hospital serves a mix of patients with similar severity (or resource intensity) to that of an “average” hospital discharge in the reference population, whereas a value of 1.2 indicates that a hospital serves a population of patients with 20% more severity than that of an “average” hospital discharge. Since 1983, the Centers for Medicare and Medicaid Services (CMS) has used RWs in their inpatient prospective payment system.³

Similar pediatric methods are less developed and necessitate special consideration as the use of existing weights may be inappropriate for a pediatric population. First, MS-DRGs were developed primarily for the Medicare population and lack sufficient granularity for pediatric populations, specifically newborns. Second, a severity stratification which incorporates important patient characteristics, such as age in pediatrics, does not exist in the MS-DRG system . Finally, although the reference populations that are used to develop MS-DRG weights do not explicitly exclude children, children typically account for approximately 15% of hospitalizations (6% excluding neonatal/maternal) and possibly feature different utilization patterns than adults with similar conditions. Thus, weights developed from a combined pediatric/adult reference population primarily reflect an adult population.

With valid pediatric RWs, stakeholders can assess a hospital’s severity mix of patients in a comparable fashion and contextualize outcome metrics. Additionally, these same weights can be used to estimate expected costs for hospitalizations or for risk adjusting various outcomes at the discharge- or hospital-level. Thus, we sought to develop hospitalization resource intensity scores for kids (H-RISK) using pediatric-specific weights and compare hospital-level CMIs across various hospital types and locations as an example of the application of this novel methodology.

Dataset

Data for this analysis were obtained from the 2012 Healthcare Cost and Utilization Project (HCUP) Kids’ Inpatient Database (KID).⁴ KID is the largest publicly available all-payer inpatient administrative database in the United States and is sponsored by the Agency for Healthcare Research and Quality as part of the HCUP. The 2012 KID included a sample of approximately 3.2 million discharge records of children <21 years old from 44 states and 4,179 community, nonrehabilitation hospitals weighted for national estimates.

Hospital discharge costs were estimated from charges using cost-to-charge ratios (CCR) provided by HCUP as a supplement to the 2012 KID.⁵ Cost estimates associated with a specific discharge were estimated by multiplying the total charges reported in the data by the appropriate hospital-specific CCR and then adjusted for price factors beyond a hospital’s control using the area wage index also provided by HCUP as a supplement.

H-RISK and Case-Mix Index Calculations

We calculated H-RISK as pediatric-specific RWs based on version 30 of 3M’s All Patient Refined DRG (APR-DRG; 3M Health Information Systems, Salt Lake City, Utah) system as a measure of resource intensity. The APR-DRG system classifies hospital discharges into over 300 base DRGs based on demographic, diagnostic, and therapeutic characteristics. Each APR-DRG is further sub-divided into 4 subclasses of severity of illness (SOI; eg, minor, moderate, major, and extreme) to indicate the intensity of resource utilization during hospitalization. However, SOI levels for differing APR-DRGs are not comparable.

For every APR-DRG SOI combinations available in the 2012 KID, calculation of RW was based on the ratio of the mean cost for patients assigned to a particular APR-DRG SOI compared with the mean cost for all patients in the database. Inpatient costs less than $0.50 were set to missing and removed from analysis. Mortalities and discharges with missing CCR and wage index values were also excluded from analysis. We required that estimates for RWs be based on a reasonable set of data (ie, 10 or more discharges) for each APR-DRG SOI, and that estimates across the 4 SOI levels within an APR-DRG be monotonically nondecreasing (ie, as SOI level increases, weights must either be the same or increasing). Winsorized means were used as point estimates for mean cost in both the numerator and denominator of RW computation. Winsorizing refers to an analytic transformation by which the influence of outliers (eg, values beyond a certain threshold) is mitigated by replacing the value of outliers with the value of the threshold. We used the 5^th and 95^th percentiles as thresholds for Winsorizing our point estimates.

Winsorized point estimates failing to meet the minimum sample size of 10 or nondecreasing monotonicity requirement were modified by one of the two following methods:

• Cost data were modeled using a generalized linear model assuming an exponential distribution. Covariates in the model included APR-DRG and SOI within APR-DRG as a continuous variable. Where applicable, Winsorized estimates of the mean were replaced with modeled estimates.
• Data from an APR-DRG SOI in question were combined with other SOIs within the same APR-DRG with the closest Winsorized mean value. Once data were combined, a common Winsorized value was re-computed and values across SOIs were checked to ensure that nondecreasing monotonicity was maintained. In some APR-DRGs with sparse data, this involved combining pairs of severity levels; in others, it involved combining three or four severity levels together.

For APR-DRGs in which no discharges at any SOI were recorded in the 2012 KID, we used the Winsorized mean of all encounters with a common major diagnostic category (MDC) as the missing APR-DRG as point estimate for all 4 SOI levels.

To calculate the CMI for a set of discharges (eg, discharges at a hospital in a year), RWs were assigned to each discharge based on APR-DRG SOI designation. Consequently, all discharges from a specific APR-DRG SOI were assigned the same RW. Once RWs were assigned, CMI was calculated as the mean RW across all discharges. To compare hospital types based on acute-care hospital stays which are usually considered with the realm of pediatric care, we excluded RWs for normal newborns, defined as APR-DRG 626 (neonate birthweight of 2000–2499 g, normal newborn or neonate with other problems) and 640 (neonate birthweight >2499 g, normal newborn or neonate with other problems), and maternal hospitalizations, defined as APR-DRG 540 (cesarean delivery) and 560 (vaginal delivery), from our CMI calculations.

Statistical Methodology

Categorical variables were summarized using frequencies and percentages; continuous variables were summarized using medians and interquartile ranges. Differences between hospital

types (eg, rural, urban nonteaching, urban teaching, and

free-standing) were assessed using a Chi-square test for association for categorical variables. Differences in continuous variables including comparisons of neonatal (MDC 15) and nonneonatal discharges, and medical versus procedural discharges as defined by the APR-DRG grouper were assessed using a Kruskal–Wallis test. All analyses were performed using SAS, Version 9.4 (SAS Institute, Cary, North Carolina); P values <.05 were considered statistically significant.

This study was considered nonhuman subjects research by the Institutional Review Board of Vanderbilt University Medical Center.

Patient Population

Table 1 summarizes the patient characteristics for all 4 hospital types. All comparisons of patient characteristics across the four hospital types are significant (P < .001). Of the 6,675,222 weighted discharges in HCUP KID 2012, almost two-thirds were less than one year old (4,269,984). Three-quarters of those infant discharges (3,733,760) were in-hospital births. The South was the Census region with the most number of discharges (38.8%), and over half of discharges (53.2%) included patients who lived in metro areas with more than 1 million residents. Patients disproportionately originated from lower-income areas with 30.9% living in zip codes with median incomes in the first quartile.

H-RISK Generation

The weighted Winsorized mean cost of all discharges was $6,135 per discharge. The majority of cost-based H-RISK were higher than 1, with 1,038 (82.5%) of APR-DRG SOIs incurring an estimated cost higher than $6,135. Solid organ and bone marrow transplantations represented 4 of the 10 highest cost-based RWs for procedural APR-DRG SOIs (Table 3). Neonatal APR-DRG SOIs accounted for 8 of the 10 highest medical RWs. A list of all APR-DRG SOIs and H-RISK can be found in Appendix A.

Hospital-Level Case-Mix Index for Acute Hospitalizations

After excluding normal newborn and maternal hospitalizations, median CMI of the 3117 hospitals with at least 20 unweighted discharges was 1.0 (interquartile range [IQR]: 0.8, 1.7). CMI varied significantly across hospital types (P < .001). Free-standing children’s hospitals exhibited the highest cost-based CMI (median: 2.7, IQR: 2.2–3.1), followed by urban teaching hospitals (median: 1.8, IQR: 1.3–2.6), urban nonteaching hospitals (median: 1.1, IQR: 0.9–1.5), and rural hospitals (median: 0.9, IQR: 0.7–0.9).

Currently, no widely available measures can compare the relative intensity of hospital care specific for inpatient pediatric populations. To meet this important need, we have developed a methodology to determine valid pediatric RWs (H-RISK) which can be used to estimate the intensity of care for applications across entire hospital patient populations and specific subpopulations. H-RISK allow calculation of CMIs for risk adjustment of various outcomes at the discharge- or hospital-level and for comparisons among hospitals and populations. Using this methodology, we demonstrated that the CMI for free-standing children’s hospitals was significantly higher than those of rural, urban, nonteaching and urban teaching hospitals for all discharges and medical or procedural subgroups.

CMS has used RWs based on DRGs since the inception of the prospective payment system in 1983. The sequence of DRGs used by CMS has purposely focused on older adult Medicare population, and CMS itself recommends applying Medicare-focused DRGs (MS-DRGs being the current iteration) only for the >65 year population.⁶ Nevertheless, many payers, both government and commercial, utilize MS-DRGs and their RWs for payment purposes when reimbursing children’s hospitals. The validity of using weights developed using this grouper in hospitals treating large numbers of pediatric patients and childhood illnesses has been called into question, particularly when such weights are used in reimbursement of children’s hospitals.⁷

Several factors contribute to the validity of a model for developing RWs. First, the system used to describe patient hospitalizations and illnesses should be appropriate to the population in question. As described above, the original DRG system and its subsequent iterations were designed to describe hospitalizations for adults >65 years of age.^{8, 9} Over the years, CMS DRGs incorporated rudimentary categories for neonatal and obstetrical hospitalizations. Still, the current MS-DRGs lack sufficient focus on common inpatient pediatric conditions to adequately describe pediatric hospitalizations, particularly those in free-standing children’s hospitals delivering tertiary and quaternary care. Thus, a more appropriate classification schema for developing RWs specific for pediatric hospitalization should include patients across the entire age spectrum. APR-DRGs represent one such classification system.

Once an appropriate patient classification system is selected, then the population of hospitalized patients to be used as the reference group becomes important. For a system targeting a pediatric inpatient population, a hospital discharge database representing a broad sample of pediatric hospitalizations offers the best basis for developing a system of weights applicable to different types of hospitals providing care for children. For this purpose, we selected the 2012 KID database, a nationally representative dataset containing data on newborn and pediatric discharges from the majority of states within the US. This choice assured that the RWs developed were based on and applicable to pediatric hospitalizations across the entire spectrum of SOI and resource intensity.

A number of measures of hospital performance and quality have been developed and are used by various entities, including individual hospitals, CMS, Leapfrog, Magnet, Joint Commission, and payers, for purposes ranging from benchmarking for improvement to payment models to reimbursement penalties. However, SOI of a hospital’s patient population influences not only the intensity of care that a hospital provides but also presents a potential impact on process and outcome measures. Thus, fair and appropriate measures must consider differences in SOI when comparing hospital performances. Using the weights derived in this paper, these adjustments can be possibly made at either the discharge- or hospital-level, depending on the application, and may include comparisons by hospital location, ownership, payer mix, or socioeconomic strata.

It is also common for hospitals to quantitatively express the uniqueness of services that they deliver to payers or the general public. A hospital-level CMI (derived as the average discharge weight for patients within a hospital) is one way that hospitals may differentiate themselves. This can be accomplished by considering the ratio of one hospital’s CMI to another hospital’s (or an average of a group of hospitals) as an expression of the relative intensity of services. For example, if hospital x has a CMI of 2.3, and hospital y has a CMI of 1.4, the population of children hospitalized at hospital x was 64.3% (1–2.3/1.4) more resource intensive than the children seen at hospital y.

This study should be considered in terms of several limitations. We used costs as the basis for determining intensity of service. Thus, the difference in cost structure among children’s hospitals and between children’s hospitals and other hospital types in the KID could have affected the final calculated weights. Also, the RWs calculated in this study rely on hospital discharge data. Thus, complications which were not “present on admission” and occurred during a hospitalization could have reflected poor quality of care yet still increase resource intensity as measured by total costs. Future studies should examine the potential impact of using present-on-admission diagnoses only for the APR-DRG grouping on the values of RWs. Significant variation may have existed among hospitals in resource utilization, and some hospitals may have exhibited significant overutilization of resources for the same conditions. However, as we used Winsorized means, the impact of potential outliers should have been reduced. Some APR-DRG-SOI combinations were seen mainly at children’s hospitals. Thus, cost structure and resource utilization practices of this subset of hospitals would have been the only contributors to weights for these patients. Given that the 2012 KID contained a broad representation of pediatric hospitalizations, with age 0–20 years, newborns accounted for the majority of total cases in the database. While providing a full range of pediatric weights, inclusion of these patients lowered the overall average RW. For this reason, we excluded normal newborn categories and maternal categories from analysis of CMI across hospital types and focused on acute-care hospitalizations. Lastly, as with any study relying on administrative data, there is always the possibility of coding errors or data entry errors in the reference dataset.

H-RISK can be used to risk adjust measures to account for severity differences across populations. These weights can also be averaged across hospitals’ patient populations to compare relative resource intensities of the patients served.

Disclosures

The authors have nothing to disclose.

Hospitals are increasingly assessed comparatively in terms of costs and quality for benchmarking purposes. These comparisons can be used by patients and families to determine where to seek care, to report compliance and grant certifications by oversight organizations (eg, Leapfrog, Magnet, Joint Commission), and by payers, to determine reimbursement models and/or to assess financial penalty or bonuses for underperforming or overperforming hospitals. As these efforts can cause substantial reputational and financial consequences for hospitals, these metrics must be contextualized within the population of patients that each hospital serves.

In adult Medicare patient populations, methods have been developed to assess the relative severity of a hospital’s full complement of patients.^1,2 These methods assume a relationship between severity and hospital resource intensity (ie, cost) and typically assume the form of relative weights (RWs), which are developed for clinically similar groups of patients (eg, Medicare Diagnosis Related Groups; MS-DRG) from a reference population. A RW for each MS-DRG is calculated as the average cost of patients within the group divided by the average cost for all patients in the reference population. These weights are then applied to a hospital’s discharges over a specific time period and averaged to obtain a hospital-level case-mix index (CMI). A value of 1 indicates that a hospital serves a mix of patients with similar severity (or resource intensity) to that of an “average” hospital discharge in the reference population, whereas a value of 1.2 indicates that a hospital serves a population of patients with 20% more severity than that of an “average” hospital discharge. Since 1983, the Centers for Medicare and Medicaid Services (CMS) has used RWs in their inpatient prospective payment system.³

Similar pediatric methods are less developed and necessitate special consideration as the use of existing weights may be inappropriate for a pediatric population. First, MS-DRGs were developed primarily for the Medicare population and lack sufficient granularity for pediatric populations, specifically newborns. Second, a severity stratification which incorporates important patient characteristics, such as age in pediatrics, does not exist in the MS-DRG system . Finally, although the reference populations that are used to develop MS-DRG weights do not explicitly exclude children, children typically account for approximately 15% of hospitalizations (6% excluding neonatal/maternal) and possibly feature different utilization patterns than adults with similar conditions. Thus, weights developed from a combined pediatric/adult reference population primarily reflect an adult population.

With valid pediatric RWs, stakeholders can assess a hospital’s severity mix of patients in a comparable fashion and contextualize outcome metrics. Additionally, these same weights can be used to estimate expected costs for hospitalizations or for risk adjusting various outcomes at the discharge- or hospital-level. Thus, we sought to develop hospitalization resource intensity scores for kids (H-RISK) using pediatric-specific weights and compare hospital-level CMIs across various hospital types and locations as an example of the application of this novel methodology.

Dataset

Data for this analysis were obtained from the 2012 Healthcare Cost and Utilization Project (HCUP) Kids’ Inpatient Database (KID).⁴ KID is the largest publicly available all-payer inpatient administrative database in the United States and is sponsored by the Agency for Healthcare Research and Quality as part of the HCUP. The 2012 KID included a sample of approximately 3.2 million discharge records of children <21 years old from 44 states and 4,179 community, nonrehabilitation hospitals weighted for national estimates.

Hospital discharge costs were estimated from charges using cost-to-charge ratios (CCR) provided by HCUP as a supplement to the 2012 KID.⁵ Cost estimates associated with a specific discharge were estimated by multiplying the total charges reported in the data by the appropriate hospital-specific CCR and then adjusted for price factors beyond a hospital’s control using the area wage index also provided by HCUP as a supplement.

H-RISK and Case-Mix Index Calculations

We calculated H-RISK as pediatric-specific RWs based on version 30 of 3M’s All Patient Refined DRG (APR-DRG; 3M Health Information Systems, Salt Lake City, Utah) system as a measure of resource intensity. The APR-DRG system classifies hospital discharges into over 300 base DRGs based on demographic, diagnostic, and therapeutic characteristics. Each APR-DRG is further sub-divided into 4 subclasses of severity of illness (SOI; eg, minor, moderate, major, and extreme) to indicate the intensity of resource utilization during hospitalization. However, SOI levels for differing APR-DRGs are not comparable.

For every APR-DRG SOI combinations available in the 2012 KID, calculation of RW was based on the ratio of the mean cost for patients assigned to a particular APR-DRG SOI compared with the mean cost for all patients in the database. Inpatient costs less than $0.50 were set to missing and removed from analysis. Mortalities and discharges with missing CCR and wage index values were also excluded from analysis. We required that estimates for RWs be based on a reasonable set of data (ie, 10 or more discharges) for each APR-DRG SOI, and that estimates across the 4 SOI levels within an APR-DRG be monotonically nondecreasing (ie, as SOI level increases, weights must either be the same or increasing). Winsorized means were used as point estimates for mean cost in both the numerator and denominator of RW computation. Winsorizing refers to an analytic transformation by which the influence of outliers (eg, values beyond a certain threshold) is mitigated by replacing the value of outliers with the value of the threshold. We used the 5^th and 95^th percentiles as thresholds for Winsorizing our point estimates.

Winsorized point estimates failing to meet the minimum sample size of 10 or nondecreasing monotonicity requirement were modified by one of the two following methods:

• Cost data were modeled using a generalized linear model assuming an exponential distribution. Covariates in the model included APR-DRG and SOI within APR-DRG as a continuous variable. Where applicable, Winsorized estimates of the mean were replaced with modeled estimates.
• Data from an APR-DRG SOI in question were combined with other SOIs within the same APR-DRG with the closest Winsorized mean value. Once data were combined, a common Winsorized value was re-computed and values across SOIs were checked to ensure that nondecreasing monotonicity was maintained. In some APR-DRGs with sparse data, this involved combining pairs of severity levels; in others, it involved combining three or four severity levels together.

For APR-DRGs in which no discharges at any SOI were recorded in the 2012 KID, we used the Winsorized mean of all encounters with a common major diagnostic category (MDC) as the missing APR-DRG as point estimate for all 4 SOI levels.

To calculate the CMI for a set of discharges (eg, discharges at a hospital in a year), RWs were assigned to each discharge based on APR-DRG SOI designation. Consequently, all discharges from a specific APR-DRG SOI were assigned the same RW. Once RWs were assigned, CMI was calculated as the mean RW across all discharges. To compare hospital types based on acute-care hospital stays which are usually considered with the realm of pediatric care, we excluded RWs for normal newborns, defined as APR-DRG 626 (neonate birthweight of 2000–2499 g, normal newborn or neonate with other problems) and 640 (neonate birthweight >2499 g, normal newborn or neonate with other problems), and maternal hospitalizations, defined as APR-DRG 540 (cesarean delivery) and 560 (vaginal delivery), from our CMI calculations.

Statistical Methodology

Categorical variables were summarized using frequencies and percentages; continuous variables were summarized using medians and interquartile ranges. Differences between hospital

types (eg, rural, urban nonteaching, urban teaching, and

free-standing) were assessed using a Chi-square test for association for categorical variables. Differences in continuous variables including comparisons of neonatal (MDC 15) and nonneonatal discharges, and medical versus procedural discharges as defined by the APR-DRG grouper were assessed using a Kruskal–Wallis test. All analyses were performed using SAS, Version 9.4 (SAS Institute, Cary, North Carolina); P values <.05 were considered statistically significant.

This study was considered nonhuman subjects research by the Institutional Review Board of Vanderbilt University Medical Center.

Patient Population

Table 1 summarizes the patient characteristics for all 4 hospital types. All comparisons of patient characteristics across the four hospital types are significant (P < .001). Of the 6,675,222 weighted discharges in HCUP KID 2012, almost two-thirds were less than one year old (4,269,984). Three-quarters of those infant discharges (3,733,760) were in-hospital births. The South was the Census region with the most number of discharges (38.8%), and over half of discharges (53.2%) included patients who lived in metro areas with more than 1 million residents. Patients disproportionately originated from lower-income areas with 30.9% living in zip codes with median incomes in the first quartile.

H-RISK Generation

The weighted Winsorized mean cost of all discharges was $6,135 per discharge. The majority of cost-based H-RISK were higher than 1, with 1,038 (82.5%) of APR-DRG SOIs incurring an estimated cost higher than $6,135. Solid organ and bone marrow transplantations represented 4 of the 10 highest cost-based RWs for procedural APR-DRG SOIs (Table 3). Neonatal APR-DRG SOIs accounted for 8 of the 10 highest medical RWs. A list of all APR-DRG SOIs and H-RISK can be found in Appendix A.

Hospital-Level Case-Mix Index for Acute Hospitalizations

After excluding normal newborn and maternal hospitalizations, median CMI of the 3117 hospitals with at least 20 unweighted discharges was 1.0 (interquartile range [IQR]: 0.8, 1.7). CMI varied significantly across hospital types (P < .001). Free-standing children’s hospitals exhibited the highest cost-based CMI (median: 2.7, IQR: 2.2–3.1), followed by urban teaching hospitals (median: 1.8, IQR: 1.3–2.6), urban nonteaching hospitals (median: 1.1, IQR: 0.9–1.5), and rural hospitals (median: 0.9, IQR: 0.7–0.9).

Currently, no widely available measures can compare the relative intensity of hospital care specific for inpatient pediatric populations. To meet this important need, we have developed a methodology to determine valid pediatric RWs (H-RISK) which can be used to estimate the intensity of care for applications across entire hospital patient populations and specific subpopulations. H-RISK allow calculation of CMIs for risk adjustment of various outcomes at the discharge- or hospital-level and for comparisons among hospitals and populations. Using this methodology, we demonstrated that the CMI for free-standing children’s hospitals was significantly higher than those of rural, urban, nonteaching and urban teaching hospitals for all discharges and medical or procedural subgroups.

CMS has used RWs based on DRGs since the inception of the prospective payment system in 1983. The sequence of DRGs used by CMS has purposely focused on older adult Medicare population, and CMS itself recommends applying Medicare-focused DRGs (MS-DRGs being the current iteration) only for the >65 year population.⁶ Nevertheless, many payers, both government and commercial, utilize MS-DRGs and their RWs for payment purposes when reimbursing children’s hospitals. The validity of using weights developed using this grouper in hospitals treating large numbers of pediatric patients and childhood illnesses has been called into question, particularly when such weights are used in reimbursement of children’s hospitals.⁷

Several factors contribute to the validity of a model for developing RWs. First, the system used to describe patient hospitalizations and illnesses should be appropriate to the population in question. As described above, the original DRG system and its subsequent iterations were designed to describe hospitalizations for adults >65 years of age.^{8, 9} Over the years, CMS DRGs incorporated rudimentary categories for neonatal and obstetrical hospitalizations. Still, the current MS-DRGs lack sufficient focus on common inpatient pediatric conditions to adequately describe pediatric hospitalizations, particularly those in free-standing children’s hospitals delivering tertiary and quaternary care. Thus, a more appropriate classification schema for developing RWs specific for pediatric hospitalization should include patients across the entire age spectrum. APR-DRGs represent one such classification system.

Once an appropriate patient classification system is selected, then the population of hospitalized patients to be used as the reference group becomes important. For a system targeting a pediatric inpatient population, a hospital discharge database representing a broad sample of pediatric hospitalizations offers the best basis for developing a system of weights applicable to different types of hospitals providing care for children. For this purpose, we selected the 2012 KID database, a nationally representative dataset containing data on newborn and pediatric discharges from the majority of states within the US. This choice assured that the RWs developed were based on and applicable to pediatric hospitalizations across the entire spectrum of SOI and resource intensity.

A number of measures of hospital performance and quality have been developed and are used by various entities, including individual hospitals, CMS, Leapfrog, Magnet, Joint Commission, and payers, for purposes ranging from benchmarking for improvement to payment models to reimbursement penalties. However, SOI of a hospital’s patient population influences not only the intensity of care that a hospital provides but also presents a potential impact on process and outcome measures. Thus, fair and appropriate measures must consider differences in SOI when comparing hospital performances. Using the weights derived in this paper, these adjustments can be possibly made at either the discharge- or hospital-level, depending on the application, and may include comparisons by hospital location, ownership, payer mix, or socioeconomic strata.

It is also common for hospitals to quantitatively express the uniqueness of services that they deliver to payers or the general public. A hospital-level CMI (derived as the average discharge weight for patients within a hospital) is one way that hospitals may differentiate themselves. This can be accomplished by considering the ratio of one hospital’s CMI to another hospital’s (or an average of a group of hospitals) as an expression of the relative intensity of services. For example, if hospital x has a CMI of 2.3, and hospital y has a CMI of 1.4, the population of children hospitalized at hospital x was 64.3% (1–2.3/1.4) more resource intensive than the children seen at hospital y.

This study should be considered in terms of several limitations. We used costs as the basis for determining intensity of service. Thus, the difference in cost structure among children’s hospitals and between children’s hospitals and other hospital types in the KID could have affected the final calculated weights. Also, the RWs calculated in this study rely on hospital discharge data. Thus, complications which were not “present on admission” and occurred during a hospitalization could have reflected poor quality of care yet still increase resource intensity as measured by total costs. Future studies should examine the potential impact of using present-on-admission diagnoses only for the APR-DRG grouping on the values of RWs. Significant variation may have existed among hospitals in resource utilization, and some hospitals may have exhibited significant overutilization of resources for the same conditions. However, as we used Winsorized means, the impact of potential outliers should have been reduced. Some APR-DRG-SOI combinations were seen mainly at children’s hospitals. Thus, cost structure and resource utilization practices of this subset of hospitals would have been the only contributors to weights for these patients. Given that the 2012 KID contained a broad representation of pediatric hospitalizations, with age 0–20 years, newborns accounted for the majority of total cases in the database. While providing a full range of pediatric weights, inclusion of these patients lowered the overall average RW. For this reason, we excluded normal newborn categories and maternal categories from analysis of CMI across hospital types and focused on acute-care hospitalizations. Lastly, as with any study relying on administrative data, there is always the possibility of coding errors or data entry errors in the reference dataset.

H-RISK can be used to risk adjust measures to account for severity differences across populations. These weights can also be averaged across hospitals’ patient populations to compare relative resource intensities of the patients served.

Disclosures

The authors have nothing to disclose.

Hospitals are increasingly assessed comparatively in terms of costs and quality for benchmarking purposes. These comparisons can be used by patients and families to determine where to seek care, to report compliance and grant certifications by oversight organizations (eg, Leapfrog, Magnet, Joint Commission), and by payers, to determine reimbursement models and/or to assess financial penalty or bonuses for underperforming or overperforming hospitals. As these efforts can cause substantial reputational and financial consequences for hospitals, these metrics must be contextualized within the population of patients that each hospital serves.

In adult Medicare patient populations, methods have been developed to assess the relative severity of a hospital’s full complement of patients.^1,2 These methods assume a relationship between severity and hospital resource intensity (ie, cost) and typically assume the form of relative weights (RWs), which are developed for clinically similar groups of patients (eg, Medicare Diagnosis Related Groups; MS-DRG) from a reference population. A RW for each MS-DRG is calculated as the average cost of patients within the group divided by the average cost for all patients in the reference population. These weights are then applied to a hospital’s discharges over a specific time period and averaged to obtain a hospital-level case-mix index (CMI). A value of 1 indicates that a hospital serves a mix of patients with similar severity (or resource intensity) to that of an “average” hospital discharge in the reference population, whereas a value of 1.2 indicates that a hospital serves a population of patients with 20% more severity than that of an “average” hospital discharge. Since 1983, the Centers for Medicare and Medicaid Services (CMS) has used RWs in their inpatient prospective payment system.³

Similar pediatric methods are less developed and necessitate special consideration as the use of existing weights may be inappropriate for a pediatric population. First, MS-DRGs were developed primarily for the Medicare population and lack sufficient granularity for pediatric populations, specifically newborns. Second, a severity stratification which incorporates important patient characteristics, such as age in pediatrics, does not exist in the MS-DRG system . Finally, although the reference populations that are used to develop MS-DRG weights do not explicitly exclude children, children typically account for approximately 15% of hospitalizations (6% excluding neonatal/maternal) and possibly feature different utilization patterns than adults with similar conditions. Thus, weights developed from a combined pediatric/adult reference population primarily reflect an adult population.

With valid pediatric RWs, stakeholders can assess a hospital’s severity mix of patients in a comparable fashion and contextualize outcome metrics. Additionally, these same weights can be used to estimate expected costs for hospitalizations or for risk adjusting various outcomes at the discharge- or hospital-level. Thus, we sought to develop hospitalization resource intensity scores for kids (H-RISK) using pediatric-specific weights and compare hospital-level CMIs across various hospital types and locations as an example of the application of this novel methodology.

Dataset

Data for this analysis were obtained from the 2012 Healthcare Cost and Utilization Project (HCUP) Kids’ Inpatient Database (KID).⁴ KID is the largest publicly available all-payer inpatient administrative database in the United States and is sponsored by the Agency for Healthcare Research and Quality as part of the HCUP. The 2012 KID included a sample of approximately 3.2 million discharge records of children <21 years old from 44 states and 4,179 community, nonrehabilitation hospitals weighted for national estimates.

Hospital discharge costs were estimated from charges using cost-to-charge ratios (CCR) provided by HCUP as a supplement to the 2012 KID.⁵ Cost estimates associated with a specific discharge were estimated by multiplying the total charges reported in the data by the appropriate hospital-specific CCR and then adjusted for price factors beyond a hospital’s control using the area wage index also provided by HCUP as a supplement.

H-RISK and Case-Mix Index Calculations

We calculated H-RISK as pediatric-specific RWs based on version 30 of 3M’s All Patient Refined DRG (APR-DRG; 3M Health Information Systems, Salt Lake City, Utah) system as a measure of resource intensity. The APR-DRG system classifies hospital discharges into over 300 base DRGs based on demographic, diagnostic, and therapeutic characteristics. Each APR-DRG is further sub-divided into 4 subclasses of severity of illness (SOI; eg, minor, moderate, major, and extreme) to indicate the intensity of resource utilization during hospitalization. However, SOI levels for differing APR-DRGs are not comparable.

For every APR-DRG SOI combinations available in the 2012 KID, calculation of RW was based on the ratio of the mean cost for patients assigned to a particular APR-DRG SOI compared with the mean cost for all patients in the database. Inpatient costs less than $0.50 were set to missing and removed from analysis. Mortalities and discharges with missing CCR and wage index values were also excluded from analysis. We required that estimates for RWs be based on a reasonable set of data (ie, 10 or more discharges) for each APR-DRG SOI, and that estimates across the 4 SOI levels within an APR-DRG be monotonically nondecreasing (ie, as SOI level increases, weights must either be the same or increasing). Winsorized means were used as point estimates for mean cost in both the numerator and denominator of RW computation. Winsorizing refers to an analytic transformation by which the influence of outliers (eg, values beyond a certain threshold) is mitigated by replacing the value of outliers with the value of the threshold. We used the 5^th and 95^th percentiles as thresholds for Winsorizing our point estimates.

Winsorized point estimates failing to meet the minimum sample size of 10 or nondecreasing monotonicity requirement were modified by one of the two following methods:

• Cost data were modeled using a generalized linear model assuming an exponential distribution. Covariates in the model included APR-DRG and SOI within APR-DRG as a continuous variable. Where applicable, Winsorized estimates of the mean were replaced with modeled estimates.
• Data from an APR-DRG SOI in question were combined with other SOIs within the same APR-DRG with the closest Winsorized mean value. Once data were combined, a common Winsorized value was re-computed and values across SOIs were checked to ensure that nondecreasing monotonicity was maintained. In some APR-DRGs with sparse data, this involved combining pairs of severity levels; in others, it involved combining three or four severity levels together.

For APR-DRGs in which no discharges at any SOI were recorded in the 2012 KID, we used the Winsorized mean of all encounters with a common major diagnostic category (MDC) as the missing APR-DRG as point estimate for all 4 SOI levels.

To calculate the CMI for a set of discharges (eg, discharges at a hospital in a year), RWs were assigned to each discharge based on APR-DRG SOI designation. Consequently, all discharges from a specific APR-DRG SOI were assigned the same RW. Once RWs were assigned, CMI was calculated as the mean RW across all discharges. To compare hospital types based on acute-care hospital stays which are usually considered with the realm of pediatric care, we excluded RWs for normal newborns, defined as APR-DRG 626 (neonate birthweight of 2000–2499 g, normal newborn or neonate with other problems) and 640 (neonate birthweight >2499 g, normal newborn or neonate with other problems), and maternal hospitalizations, defined as APR-DRG 540 (cesarean delivery) and 560 (vaginal delivery), from our CMI calculations.

Statistical Methodology

Categorical variables were summarized using frequencies and percentages; continuous variables were summarized using medians and interquartile ranges. Differences between hospital

types (eg, rural, urban nonteaching, urban teaching, and

free-standing) were assessed using a Chi-square test for association for categorical variables. Differences in continuous variables including comparisons of neonatal (MDC 15) and nonneonatal discharges, and medical versus procedural discharges as defined by the APR-DRG grouper were assessed using a Kruskal–Wallis test. All analyses were performed using SAS, Version 9.4 (SAS Institute, Cary, North Carolina); P values <.05 were considered statistically significant.

This study was considered nonhuman subjects research by the Institutional Review Board of Vanderbilt University Medical Center.

Patient Population

Table 1 summarizes the patient characteristics for all 4 hospital types. All comparisons of patient characteristics across the four hospital types are significant (P < .001). Of the 6,675,222 weighted discharges in HCUP KID 2012, almost two-thirds were less than one year old (4,269,984). Three-quarters of those infant discharges (3,733,760) were in-hospital births. The South was the Census region with the most number of discharges (38.8%), and over half of discharges (53.2%) included patients who lived in metro areas with more than 1 million residents. Patients disproportionately originated from lower-income areas with 30.9% living in zip codes with median incomes in the first quartile.

H-RISK Generation

The weighted Winsorized mean cost of all discharges was $6,135 per discharge. The majority of cost-based H-RISK were higher than 1, with 1,038 (82.5%) of APR-DRG SOIs incurring an estimated cost higher than $6,135. Solid organ and bone marrow transplantations represented 4 of the 10 highest cost-based RWs for procedural APR-DRG SOIs (Table 3). Neonatal APR-DRG SOIs accounted for 8 of the 10 highest medical RWs. A list of all APR-DRG SOIs and H-RISK can be found in Appendix A.

Hospital-Level Case-Mix Index for Acute Hospitalizations

After excluding normal newborn and maternal hospitalizations, median CMI of the 3117 hospitals with at least 20 unweighted discharges was 1.0 (interquartile range [IQR]: 0.8, 1.7). CMI varied significantly across hospital types (P < .001). Free-standing children’s hospitals exhibited the highest cost-based CMI (median: 2.7, IQR: 2.2–3.1), followed by urban teaching hospitals (median: 1.8, IQR: 1.3–2.6), urban nonteaching hospitals (median: 1.1, IQR: 0.9–1.5), and rural hospitals (median: 0.9, IQR: 0.7–0.9).

Currently, no widely available measures can compare the relative intensity of hospital care specific for inpatient pediatric populations. To meet this important need, we have developed a methodology to determine valid pediatric RWs (H-RISK) which can be used to estimate the intensity of care for applications across entire hospital patient populations and specific subpopulations. H-RISK allow calculation of CMIs for risk adjustment of various outcomes at the discharge- or hospital-level and for comparisons among hospitals and populations. Using this methodology, we demonstrated that the CMI for free-standing children’s hospitals was significantly higher than those of rural, urban, nonteaching and urban teaching hospitals for all discharges and medical or procedural subgroups.

CMS has used RWs based on DRGs since the inception of the prospective payment system in 1983. The sequence of DRGs used by CMS has purposely focused on older adult Medicare population, and CMS itself recommends applying Medicare-focused DRGs (MS-DRGs being the current iteration) only for the >65 year population.⁶ Nevertheless, many payers, both government and commercial, utilize MS-DRGs and their RWs for payment purposes when reimbursing children’s hospitals. The validity of using weights developed using this grouper in hospitals treating large numbers of pediatric patients and childhood illnesses has been called into question, particularly when such weights are used in reimbursement of children’s hospitals.⁷

Several factors contribute to the validity of a model for developing RWs. First, the system used to describe patient hospitalizations and illnesses should be appropriate to the population in question. As described above, the original DRG system and its subsequent iterations were designed to describe hospitalizations for adults >65 years of age.^{8, 9} Over the years, CMS DRGs incorporated rudimentary categories for neonatal and obstetrical hospitalizations. Still, the current MS-DRGs lack sufficient focus on common inpatient pediatric conditions to adequately describe pediatric hospitalizations, particularly those in free-standing children’s hospitals delivering tertiary and quaternary care. Thus, a more appropriate classification schema for developing RWs specific for pediatric hospitalization should include patients across the entire age spectrum. APR-DRGs represent one such classification system.

Once an appropriate patient classification system is selected, then the population of hospitalized patients to be used as the reference group becomes important. For a system targeting a pediatric inpatient population, a hospital discharge database representing a broad sample of pediatric hospitalizations offers the best basis for developing a system of weights applicable to different types of hospitals providing care for children. For this purpose, we selected the 2012 KID database, a nationally representative dataset containing data on newborn and pediatric discharges from the majority of states within the US. This choice assured that the RWs developed were based on and applicable to pediatric hospitalizations across the entire spectrum of SOI and resource intensity.

A number of measures of hospital performance and quality have been developed and are used by various entities, including individual hospitals, CMS, Leapfrog, Magnet, Joint Commission, and payers, for purposes ranging from benchmarking for improvement to payment models to reimbursement penalties. However, SOI of a hospital’s patient population influences not only the intensity of care that a hospital provides but also presents a potential impact on process and outcome measures. Thus, fair and appropriate measures must consider differences in SOI when comparing hospital performances. Using the weights derived in this paper, these adjustments can be possibly made at either the discharge- or hospital-level, depending on the application, and may include comparisons by hospital location, ownership, payer mix, or socioeconomic strata.

It is also common for hospitals to quantitatively express the uniqueness of services that they deliver to payers or the general public. A hospital-level CMI (derived as the average discharge weight for patients within a hospital) is one way that hospitals may differentiate themselves. This can be accomplished by considering the ratio of one hospital’s CMI to another hospital’s (or an average of a group of hospitals) as an expression of the relative intensity of services. For example, if hospital x has a CMI of 2.3, and hospital y has a CMI of 1.4, the population of children hospitalized at hospital x was 64.3% (1–2.3/1.4) more resource intensive than the children seen at hospital y.

This study should be considered in terms of several limitations. We used costs as the basis for determining intensity of service. Thus, the difference in cost structure among children’s hospitals and between children’s hospitals and other hospital types in the KID could have affected the final calculated weights. Also, the RWs calculated in this study rely on hospital discharge data. Thus, complications which were not “present on admission” and occurred during a hospitalization could have reflected poor quality of care yet still increase resource intensity as measured by total costs. Future studies should examine the potential impact of using present-on-admission diagnoses only for the APR-DRG grouping on the values of RWs. Significant variation may have existed among hospitals in resource utilization, and some hospitals may have exhibited significant overutilization of resources for the same conditions. However, as we used Winsorized means, the impact of potential outliers should have been reduced. Some APR-DRG-SOI combinations were seen mainly at children’s hospitals. Thus, cost structure and resource utilization practices of this subset of hospitals would have been the only contributors to weights for these patients. Given that the 2012 KID contained a broad representation of pediatric hospitalizations, with age 0–20 years, newborns accounted for the majority of total cases in the database. While providing a full range of pediatric weights, inclusion of these patients lowered the overall average RW. For this reason, we excluded normal newborn categories and maternal categories from analysis of CMI across hospital types and focused on acute-care hospitalizations. Lastly, as with any study relying on administrative data, there is always the possibility of coding errors or data entry errors in the reference dataset.

H-RISK can be used to risk adjust measures to account for severity differences across populations. These weights can also be averaged across hospitals’ patient populations to compare relative resource intensities of the patients served.

Disclosures

The authors have nothing to disclose.