Madeleine Ma, MS

Hospital admission represents a time period during which patients are at risk for poor clinical outcomes. Although some risk is directly generated by illness pathophysiology, some additive risk is generated by the emergency department (ED)inpatient service handover inherent in the admission process.[1] Increased risk of suboptimal outcomes can result from ED overcrowding, which has been associated with increased mortality, difficulty in patient disposition, and delays in provision of care.[2] Inpatient bed occupancy, as well as availability and organization of accepting inpatient service healthcare staff, can affect ED overcrowding as well.[3, 4]

The overwhelming majority of hospitalist groups accept a significant portion of their admissions via the ED.[5] Hospitalist services must balance their daily group workload between ongoing care and discharge of inpatients and the activity of admitting new patients to their service. Two major models of admission processing exist for hospitalist groups to accomplish these competing tasks. One model, called the general model, employs the use of individual hospitalists to simultaneously perform admission activity as well as ongoing ward‐based care for inpatients during their workday. In the general model, a hospitalist who admits patients on their first hospital day will generally continue to see them on their second hospital day. The other model, called the admitter‐rounder model, divides the hospitalist daily group workflow between hospitalists who are assigned to perform only admission activity (admitters), and hospitalists who are assigned to perform only ongoing care for patients who are already admitted (rounders). In the admitter‐rounder model, the admitter on a patient's first hospital day will generally not serve as the patient's rounder on subsequent hospital days.

Limited evidence exists to guide hospitalist groups on which model their service design should adopt. Conflicting evidence exists as to whether the fragmentation of care generated by an admitter‐rounder admission model is beneficial or harmful.[6, 7, 8, 9] Increased availability of attending inpatient physicians during the EDinpatient admission process has been associated with improved hospital mortality and decreased readmissions in hospital settings outside the United States, where attending availability may otherwise be limited.[10, 11, 12] Separation of admission and rounding activity within a hospitalist workforce may allow each group of hospitalists to provide more timely and effective care related to their respective tasks. Our division implemented a change from a general model to an admitter‐rounder model of care on January 2, 2012. We hypothesized that changing from a general admission model to an admitter‐rounder model of care would be associated with a decreased rate of transfer to the intensive care unit (ICU) 24 hours after floor arrival and shortened ED length of stay (LOS), due to improved availability of hospitalists during the admission process. Due to the introduction of discontinuity, we hypothesized that adoption of the admitter‐rounder model would be associated with a prolongation of hospital LOS and no overall effect on 30 day postdischarge readmission rate. We sought to examine the relationship between our division's service design change and our hypothesized variables of interest.

METHODS

RESULTS

DISCUSSION

Our observations were revealing for a statistically nonsignificant trend toward increased ICU transfers 24 hours after floor arrival after adoption of the admitter‐rounder model by the hospital medicine service. Despite prior publication of early transfer to the ICU being associated with adverse outcomes, including increased inpatient mortality, we observed no difference in mortality in our study groups.[13] We suspect that earlier transfer to the ICU in our study cohort may instead represent a protective action taken more frequently by admitting hospitalists in the admitter‐rounder model in response to provider discontinuity risks embedded in the admission process. Requests for transfer to the ICU at our institution require approval by the ICU team, and requests from attending hospitalists may be responded to differently from requests enacted by teaching team members, which as a factor also may account for some of the adjusted differences in transfer incidence. Taken together, increased availability of hospitalists during the admission process may result in earlier implementation of an overall lower threshold for implementation of ICU transfer. Our conclusion is limited by our study cohort's overall inpatient mortality rate, which is sufficiently low to preclude further assessment of the relationship of adverse outcomes with ICU transfer rate in our study groups. Therefore, clinical significance of our primary outcome findings, as well as the workload factors that impact ICU transfers initiated by hospitalist and teaching services, require further examination.

Despite a hypothesized increase in hospital LOS caused by additional discontinuity of hospitalist care in the admitter‐rounder model, adoption of the admitter‐rounder model was not associated with an increased hospital LOS. We suspect this finding may represent the presence of action(s) proximal to the admission process, on the part of either admission and/or rounding hospitalists, which decrease hospital LOS to a degree offsetting the expected LOS increase generated by provider discontinuity. Examples of such actions include more efficient testing or consultation, or improved detection of diagnostic errors.

Adoption of the admitter‐rounder model by the hospital medicine service was also associated with decreased hospital readmission rates compared to the time‐matched teaching service. We suspect that assignment of daily discharge and admission service activity to separate hospitalists in the admitter‐rounder model may allow more opportunity for rounder hospitalists to engage in activity protective against readmissions, such as greater direct engagement with postdischarge resources, or improved hospitalist availability for multidisciplinary inpatient efforts focused on discharge planning.

Adoption of the admitter‐rounder model was found to be associated with a median 29‐minute increase in ED LOS compared to the time‐matched teaching service. As a floor team member's physical presence in the ED was not required for ED‐floor transfer during the study period, increased physical availability of admitting hospitalists in the admitter‐rounder model may allow for increased opportunity for a hospitalist to disrupt ED‐specific workflows related to patient transfer (eg, disruption of transportation service activity by an earlier bedside visit from the admitting hospitalist). Hospitalists in the general model were allowed to leave after performing their daily duties, whereas admitting hospitalists in the admitter‐rounder model were assigned to stay for a timed shift, regardless of the completion of admissions; the difference in duty assignment may be associated with different hospitalist behaviors during the admission process. Improved ease for ED staff to contact hospitalist staff in the admitter‐rounder model may have led ED staff to prioritize other tasks more demanding of their continuous engagement at the expense of initiating admissions, thereby paradoxically delaying admissions to hospital medicine.

Other studies exist that attempt to describe changes in admission service structure, particularly with regard to housestaff admission activity in relation to changes in resident work hours. Many of these studies vary with regard to implementation of separate physician teams for day and night coverage, or are focused on a specific medical condition, thereby limiting their applicability to a hospital medicine service free of work‐hour restrictions and engaged in care of a wide variety of medical conditions.[18, 19, 20] In contrast, our study is an attempt to examine, in isolation, outcomes associated with adoption of an admitter‐rounder model of care as a specific discontinuity risk during the admission process, within the context of a stable system of night coverage in place for all medical teams engaged in admission activity of undifferentiated medical patients.

Limitations of our study include the inability to ascertain causality of observed outcomes, due to our observational study design. Our study was of a single hospital, which may limit applicability of our results to other hospital environments. However, the admission models examined in our study are common among hospital medicine groups. Clinically relevant outcome metrics, such as mortality and unexpected transfer to the operating room, were measured but of too low incidence to allow for further meaningful analysis. The clinical consequences and workflow practices that correlate with our study's findings likely require case review and time‐motion analyses, respectively, to further delineate the relevance of our findings; these analyses were outside of the scope of our study, and further investigation is required. In summary, our observations suggest that adoption by hospitalist services of an admitter‐rounder model of care for admissions is associated with a decreased rate of hospital readmission 30 days after discharge, with no effect on median hospital LOS, a statistically nonsignificant trend toward more ICU transfers in the first 24 hours of a patient's hospital stay, and a slight increase in median ED LOS.

Hospital admission represents a time period during which patients are at risk for poor clinical outcomes. Although some risk is directly generated by illness pathophysiology, some additive risk is generated by the emergency department (ED)inpatient service handover inherent in the admission process.[1] Increased risk of suboptimal outcomes can result from ED overcrowding, which has been associated with increased mortality, difficulty in patient disposition, and delays in provision of care.[2] Inpatient bed occupancy, as well as availability and organization of accepting inpatient service healthcare staff, can affect ED overcrowding as well.[3, 4]

The overwhelming majority of hospitalist groups accept a significant portion of their admissions via the ED.[5] Hospitalist services must balance their daily group workload between ongoing care and discharge of inpatients and the activity of admitting new patients to their service. Two major models of admission processing exist for hospitalist groups to accomplish these competing tasks. One model, called the general model, employs the use of individual hospitalists to simultaneously perform admission activity as well as ongoing ward‐based care for inpatients during their workday. In the general model, a hospitalist who admits patients on their first hospital day will generally continue to see them on their second hospital day. The other model, called the admitter‐rounder model, divides the hospitalist daily group workflow between hospitalists who are assigned to perform only admission activity (admitters), and hospitalists who are assigned to perform only ongoing care for patients who are already admitted (rounders). In the admitter‐rounder model, the admitter on a patient's first hospital day will generally not serve as the patient's rounder on subsequent hospital days.

Limited evidence exists to guide hospitalist groups on which model their service design should adopt. Conflicting evidence exists as to whether the fragmentation of care generated by an admitter‐rounder admission model is beneficial or harmful.[6, 7, 8, 9] Increased availability of attending inpatient physicians during the EDinpatient admission process has been associated with improved hospital mortality and decreased readmissions in hospital settings outside the United States, where attending availability may otherwise be limited.[10, 11, 12] Separation of admission and rounding activity within a hospitalist workforce may allow each group of hospitalists to provide more timely and effective care related to their respective tasks. Our division implemented a change from a general model to an admitter‐rounder model of care on January 2, 2012. We hypothesized that changing from a general admission model to an admitter‐rounder model of care would be associated with a decreased rate of transfer to the intensive care unit (ICU) 24 hours after floor arrival and shortened ED length of stay (LOS), due to improved availability of hospitalists during the admission process. Due to the introduction of discontinuity, we hypothesized that adoption of the admitter‐rounder model would be associated with a prolongation of hospital LOS and no overall effect on 30 day postdischarge readmission rate. We sought to examine the relationship between our division's service design change and our hypothesized variables of interest.

METHODS

RESULTS

DISCUSSION

Our observations were revealing for a statistically nonsignificant trend toward increased ICU transfers 24 hours after floor arrival after adoption of the admitter‐rounder model by the hospital medicine service. Despite prior publication of early transfer to the ICU being associated with adverse outcomes, including increased inpatient mortality, we observed no difference in mortality in our study groups.[13] We suspect that earlier transfer to the ICU in our study cohort may instead represent a protective action taken more frequently by admitting hospitalists in the admitter‐rounder model in response to provider discontinuity risks embedded in the admission process. Requests for transfer to the ICU at our institution require approval by the ICU team, and requests from attending hospitalists may be responded to differently from requests enacted by teaching team members, which as a factor also may account for some of the adjusted differences in transfer incidence. Taken together, increased availability of hospitalists during the admission process may result in earlier implementation of an overall lower threshold for implementation of ICU transfer. Our conclusion is limited by our study cohort's overall inpatient mortality rate, which is sufficiently low to preclude further assessment of the relationship of adverse outcomes with ICU transfer rate in our study groups. Therefore, clinical significance of our primary outcome findings, as well as the workload factors that impact ICU transfers initiated by hospitalist and teaching services, require further examination.

Despite a hypothesized increase in hospital LOS caused by additional discontinuity of hospitalist care in the admitter‐rounder model, adoption of the admitter‐rounder model was not associated with an increased hospital LOS. We suspect this finding may represent the presence of action(s) proximal to the admission process, on the part of either admission and/or rounding hospitalists, which decrease hospital LOS to a degree offsetting the expected LOS increase generated by provider discontinuity. Examples of such actions include more efficient testing or consultation, or improved detection of diagnostic errors.

Adoption of the admitter‐rounder model by the hospital medicine service was also associated with decreased hospital readmission rates compared to the time‐matched teaching service. We suspect that assignment of daily discharge and admission service activity to separate hospitalists in the admitter‐rounder model may allow more opportunity for rounder hospitalists to engage in activity protective against readmissions, such as greater direct engagement with postdischarge resources, or improved hospitalist availability for multidisciplinary inpatient efforts focused on discharge planning.

Adoption of the admitter‐rounder model was found to be associated with a median 29‐minute increase in ED LOS compared to the time‐matched teaching service. As a floor team member's physical presence in the ED was not required for ED‐floor transfer during the study period, increased physical availability of admitting hospitalists in the admitter‐rounder model may allow for increased opportunity for a hospitalist to disrupt ED‐specific workflows related to patient transfer (eg, disruption of transportation service activity by an earlier bedside visit from the admitting hospitalist). Hospitalists in the general model were allowed to leave after performing their daily duties, whereas admitting hospitalists in the admitter‐rounder model were assigned to stay for a timed shift, regardless of the completion of admissions; the difference in duty assignment may be associated with different hospitalist behaviors during the admission process. Improved ease for ED staff to contact hospitalist staff in the admitter‐rounder model may have led ED staff to prioritize other tasks more demanding of their continuous engagement at the expense of initiating admissions, thereby paradoxically delaying admissions to hospital medicine.

Other studies exist that attempt to describe changes in admission service structure, particularly with regard to housestaff admission activity in relation to changes in resident work hours. Many of these studies vary with regard to implementation of separate physician teams for day and night coverage, or are focused on a specific medical condition, thereby limiting their applicability to a hospital medicine service free of work‐hour restrictions and engaged in care of a wide variety of medical conditions.[18, 19, 20] In contrast, our study is an attempt to examine, in isolation, outcomes associated with adoption of an admitter‐rounder model of care as a specific discontinuity risk during the admission process, within the context of a stable system of night coverage in place for all medical teams engaged in admission activity of undifferentiated medical patients.

Limitations of our study include the inability to ascertain causality of observed outcomes, due to our observational study design. Our study was of a single hospital, which may limit applicability of our results to other hospital environments. However, the admission models examined in our study are common among hospital medicine groups. Clinically relevant outcome metrics, such as mortality and unexpected transfer to the operating room, were measured but of too low incidence to allow for further meaningful analysis. The clinical consequences and workflow practices that correlate with our study's findings likely require case review and time‐motion analyses, respectively, to further delineate the relevance of our findings; these analyses were outside of the scope of our study, and further investigation is required. In summary, our observations suggest that adoption by hospitalist services of an admitter‐rounder model of care for admissions is associated with a decreased rate of hospital readmission 30 days after discharge, with no effect on median hospital LOS, a statistically nonsignificant trend toward more ICU transfers in the first 24 hours of a patient's hospital stay, and a slight increase in median ED LOS.

Hospital admission represents a time period during which patients are at risk for poor clinical outcomes. Although some risk is directly generated by illness pathophysiology, some additive risk is generated by the emergency department (ED)inpatient service handover inherent in the admission process.[1] Increased risk of suboptimal outcomes can result from ED overcrowding, which has been associated with increased mortality, difficulty in patient disposition, and delays in provision of care.[2] Inpatient bed occupancy, as well as availability and organization of accepting inpatient service healthcare staff, can affect ED overcrowding as well.[3, 4]

The overwhelming majority of hospitalist groups accept a significant portion of their admissions via the ED.[5] Hospitalist services must balance their daily group workload between ongoing care and discharge of inpatients and the activity of admitting new patients to their service. Two major models of admission processing exist for hospitalist groups to accomplish these competing tasks. One model, called the general model, employs the use of individual hospitalists to simultaneously perform admission activity as well as ongoing ward‐based care for inpatients during their workday. In the general model, a hospitalist who admits patients on their first hospital day will generally continue to see them on their second hospital day. The other model, called the admitter‐rounder model, divides the hospitalist daily group workflow between hospitalists who are assigned to perform only admission activity (admitters), and hospitalists who are assigned to perform only ongoing care for patients who are already admitted (rounders). In the admitter‐rounder model, the admitter on a patient's first hospital day will generally not serve as the patient's rounder on subsequent hospital days.

Limited evidence exists to guide hospitalist groups on which model their service design should adopt. Conflicting evidence exists as to whether the fragmentation of care generated by an admitter‐rounder admission model is beneficial or harmful.[6, 7, 8, 9] Increased availability of attending inpatient physicians during the EDinpatient admission process has been associated with improved hospital mortality and decreased readmissions in hospital settings outside the United States, where attending availability may otherwise be limited.[10, 11, 12] Separation of admission and rounding activity within a hospitalist workforce may allow each group of hospitalists to provide more timely and effective care related to their respective tasks. Our division implemented a change from a general model to an admitter‐rounder model of care on January 2, 2012. We hypothesized that changing from a general admission model to an admitter‐rounder model of care would be associated with a decreased rate of transfer to the intensive care unit (ICU) 24 hours after floor arrival and shortened ED length of stay (LOS), due to improved availability of hospitalists during the admission process. Due to the introduction of discontinuity, we hypothesized that adoption of the admitter‐rounder model would be associated with a prolongation of hospital LOS and no overall effect on 30 day postdischarge readmission rate. We sought to examine the relationship between our division's service design change and our hypothesized variables of interest.

METHODS

RESULTS

DISCUSSION

Our observations were revealing for a statistically nonsignificant trend toward increased ICU transfers 24 hours after floor arrival after adoption of the admitter‐rounder model by the hospital medicine service. Despite prior publication of early transfer to the ICU being associated with adverse outcomes, including increased inpatient mortality, we observed no difference in mortality in our study groups.[13] We suspect that earlier transfer to the ICU in our study cohort may instead represent a protective action taken more frequently by admitting hospitalists in the admitter‐rounder model in response to provider discontinuity risks embedded in the admission process. Requests for transfer to the ICU at our institution require approval by the ICU team, and requests from attending hospitalists may be responded to differently from requests enacted by teaching team members, which as a factor also may account for some of the adjusted differences in transfer incidence. Taken together, increased availability of hospitalists during the admission process may result in earlier implementation of an overall lower threshold for implementation of ICU transfer. Our conclusion is limited by our study cohort's overall inpatient mortality rate, which is sufficiently low to preclude further assessment of the relationship of adverse outcomes with ICU transfer rate in our study groups. Therefore, clinical significance of our primary outcome findings, as well as the workload factors that impact ICU transfers initiated by hospitalist and teaching services, require further examination.

Despite a hypothesized increase in hospital LOS caused by additional discontinuity of hospitalist care in the admitter‐rounder model, adoption of the admitter‐rounder model was not associated with an increased hospital LOS. We suspect this finding may represent the presence of action(s) proximal to the admission process, on the part of either admission and/or rounding hospitalists, which decrease hospital LOS to a degree offsetting the expected LOS increase generated by provider discontinuity. Examples of such actions include more efficient testing or consultation, or improved detection of diagnostic errors.

Adoption of the admitter‐rounder model by the hospital medicine service was also associated with decreased hospital readmission rates compared to the time‐matched teaching service. We suspect that assignment of daily discharge and admission service activity to separate hospitalists in the admitter‐rounder model may allow more opportunity for rounder hospitalists to engage in activity protective against readmissions, such as greater direct engagement with postdischarge resources, or improved hospitalist availability for multidisciplinary inpatient efforts focused on discharge planning.

Adoption of the admitter‐rounder model was found to be associated with a median 29‐minute increase in ED LOS compared to the time‐matched teaching service. As a floor team member's physical presence in the ED was not required for ED‐floor transfer during the study period, increased physical availability of admitting hospitalists in the admitter‐rounder model may allow for increased opportunity for a hospitalist to disrupt ED‐specific workflows related to patient transfer (eg, disruption of transportation service activity by an earlier bedside visit from the admitting hospitalist). Hospitalists in the general model were allowed to leave after performing their daily duties, whereas admitting hospitalists in the admitter‐rounder model were assigned to stay for a timed shift, regardless of the completion of admissions; the difference in duty assignment may be associated with different hospitalist behaviors during the admission process. Improved ease for ED staff to contact hospitalist staff in the admitter‐rounder model may have led ED staff to prioritize other tasks more demanding of their continuous engagement at the expense of initiating admissions, thereby paradoxically delaying admissions to hospital medicine.

Other studies exist that attempt to describe changes in admission service structure, particularly with regard to housestaff admission activity in relation to changes in resident work hours. Many of these studies vary with regard to implementation of separate physician teams for day and night coverage, or are focused on a specific medical condition, thereby limiting their applicability to a hospital medicine service free of work‐hour restrictions and engaged in care of a wide variety of medical conditions.[18, 19, 20] In contrast, our study is an attempt to examine, in isolation, outcomes associated with adoption of an admitter‐rounder model of care as a specific discontinuity risk during the admission process, within the context of a stable system of night coverage in place for all medical teams engaged in admission activity of undifferentiated medical patients.

Limitations of our study include the inability to ascertain causality of observed outcomes, due to our observational study design. Our study was of a single hospital, which may limit applicability of our results to other hospital environments. However, the admission models examined in our study are common among hospital medicine groups. Clinically relevant outcome metrics, such as mortality and unexpected transfer to the operating room, were measured but of too low incidence to allow for further meaningful analysis. The clinical consequences and workflow practices that correlate with our study's findings likely require case review and time‐motion analyses, respectively, to further delineate the relevance of our findings; these analyses were outside of the scope of our study, and further investigation is required. In summary, our observations suggest that adoption by hospitalist services of an admitter‐rounder model of care for admissions is associated with a decreased rate of hospital readmission 30 days after discharge, with no effect on median hospital LOS, a statistically nonsignificant trend toward more ICU transfers in the first 24 hours of a patient's hospital stay, and a slight increase in median ED LOS.

We greatly appreciate the thoughtful points made by Dr. Kerman regarding our recently published study evaluating the association of hospitalist continuity on adverse events (AEs).[1] We agree that a 7‐on/7‐off staffing model may limit discontinuity relative to models using shorter rotations lengths. Many hospital medicine programs use a 7‐on/7‐off model to optimize continuity. Longer rotation lengths are uncommon, as they may lead to fatigue and negatively affect physician work‐life balance. Shorter rotation lengths do exist, and we acknowledge that a study in a setting with greater fragmentation may have detected an effect.

We respectfully disagree with Dr. Kerman's concern that our methods for AE detection and confirmation may have been insensitive. We did not rely on incident reports, as these systems suffer from under‐reporting and often represent only a fraction of true AEs. We used a modified version of the classic 2‐stage method to identify and confirm AEs.[2] In the first stage, we used computerized screens, based on criteria from the Harvard Medical Practice Study and Institute for Healthcare Improvement global trigger tool, to identify potential AEs.[3, 4, 5] A research nurse created narrative summaries of potential AEs. A physician researcher then reviewed the narrative summaries to confirm whether an AE was truly present. This time‐consuming method is much more sensitive and specific than other options for patient safety measurement, including administrative data analyses and incident reporting systems.[6, 7]

With respect to other outcomes that may be affected by hospitalist continuity, we recently published a separate study showing that lower inpatient physician continuity was significantly associated with modest increases in hospital costs.[8] We found no association between continuity and patient satisfaction, but were likely underpowered to detect one. Interestingly, some of the models in our study suggested a slightly reduced risk of readmission with lower continuity. We were surprised by this finding and hypothesized that countervailing forces may be at play during handoffs of care from 1 hospitalist to another. Transitions of care introduce the opportunity for critical information to be lost, but they also introduce the potential for patient reassessment. A hospitalist newly taking over care from another may not be anchored to the initial diagnostic impressions and management plan established by the first. Of course, the potential benefit of a reassessment could only occur if the new hospitalist has time to perform one. At extremely high patient volumes, this theoretical benefit is unlikely to exist.

We did not include length of stay (LOS) as an outcome because hospitalist continuity and LOS are interdependent. Although discontinuity may lead to longer LOS, longer LOS definitely increases the probability of discontinuity. Thus, we controlled for LOS in our statistical models to isolate the effect of continuity. The study by Epstein and colleagues did not take into account the interdependence between LOS and hospitalist continuity.[9] Observational studies are not ideal for determining the effect of continuity on LOS. The Combing Incentives and Continuity Leading to Efficiency (CICLE) study by Chandra and colleagues was a pre‐post evaluation of a hospitalist staffing model specifically designed to improve continuity.[10] In the CICLE model, physicians work in a 4‐day rotation. On day 1, physicians exclusively admit patients. On day 2, physicians care for patients admitted on day 1 and accept patients admitted overnight. On days 3 and 4, physicians continue to care for patients received on days 1 and 2, but receive no additional patients. The remaining patients are transitioned to the next physician entering the cycle at the end of day 4. Chandra and colleagues found a 7.5% reduction in LOS and an 8.5% reduction in charges. Interestingly, they also found a 13.5% increase in readmissions that did not achieve statistical significance (P=0.08). The CICLE study suggests continuity does affect LOS, but is limited in that it did not account for a potential preexisting trend toward lower LOS.

Dr. Kerman presents data showing that it takes longer for a physician to care for a patient who is new to him or her than for a patient who is previously known. This finding has face validity. However, as we have suggested, the extra time spent by the oncoming physician may have both advantages and disadvantages. The disadvantages include time‐consuming cognitive work for the physician and the potential for information loss affecting patient care. The potential advantage is a second physician reassessing the diagnosis and management decisions established by the first, potentially correcting errors and optimizing care.

Ultimately, more research is needed to illuminate the effect of hospitalist continuity on patient outcomes. For now, we feel that hospital medicine group leaders need not institute lengthy rotations or staffing models that prioritize continuity above all other factors, as continuity appears to have little impact on patient outcomes.

We greatly appreciate the thoughtful points made by Dr. Kerman regarding our recently published study evaluating the association of hospitalist continuity on adverse events (AEs).[1] We agree that a 7‐on/7‐off staffing model may limit discontinuity relative to models using shorter rotations lengths. Many hospital medicine programs use a 7‐on/7‐off model to optimize continuity. Longer rotation lengths are uncommon, as they may lead to fatigue and negatively affect physician work‐life balance. Shorter rotation lengths do exist, and we acknowledge that a study in a setting with greater fragmentation may have detected an effect.

We respectfully disagree with Dr. Kerman's concern that our methods for AE detection and confirmation may have been insensitive. We did not rely on incident reports, as these systems suffer from under‐reporting and often represent only a fraction of true AEs. We used a modified version of the classic 2‐stage method to identify and confirm AEs.[2] In the first stage, we used computerized screens, based on criteria from the Harvard Medical Practice Study and Institute for Healthcare Improvement global trigger tool, to identify potential AEs.[3, 4, 5] A research nurse created narrative summaries of potential AEs. A physician researcher then reviewed the narrative summaries to confirm whether an AE was truly present. This time‐consuming method is much more sensitive and specific than other options for patient safety measurement, including administrative data analyses and incident reporting systems.[6, 7]

With respect to other outcomes that may be affected by hospitalist continuity, we recently published a separate study showing that lower inpatient physician continuity was significantly associated with modest increases in hospital costs.[8] We found no association between continuity and patient satisfaction, but were likely underpowered to detect one. Interestingly, some of the models in our study suggested a slightly reduced risk of readmission with lower continuity. We were surprised by this finding and hypothesized that countervailing forces may be at play during handoffs of care from 1 hospitalist to another. Transitions of care introduce the opportunity for critical information to be lost, but they also introduce the potential for patient reassessment. A hospitalist newly taking over care from another may not be anchored to the initial diagnostic impressions and management plan established by the first. Of course, the potential benefit of a reassessment could only occur if the new hospitalist has time to perform one. At extremely high patient volumes, this theoretical benefit is unlikely to exist.

We did not include length of stay (LOS) as an outcome because hospitalist continuity and LOS are interdependent. Although discontinuity may lead to longer LOS, longer LOS definitely increases the probability of discontinuity. Thus, we controlled for LOS in our statistical models to isolate the effect of continuity. The study by Epstein and colleagues did not take into account the interdependence between LOS and hospitalist continuity.[9] Observational studies are not ideal for determining the effect of continuity on LOS. The Combing Incentives and Continuity Leading to Efficiency (CICLE) study by Chandra and colleagues was a pre‐post evaluation of a hospitalist staffing model specifically designed to improve continuity.[10] In the CICLE model, physicians work in a 4‐day rotation. On day 1, physicians exclusively admit patients. On day 2, physicians care for patients admitted on day 1 and accept patients admitted overnight. On days 3 and 4, physicians continue to care for patients received on days 1 and 2, but receive no additional patients. The remaining patients are transitioned to the next physician entering the cycle at the end of day 4. Chandra and colleagues found a 7.5% reduction in LOS and an 8.5% reduction in charges. Interestingly, they also found a 13.5% increase in readmissions that did not achieve statistical significance (P=0.08). The CICLE study suggests continuity does affect LOS, but is limited in that it did not account for a potential preexisting trend toward lower LOS.

Dr. Kerman presents data showing that it takes longer for a physician to care for a patient who is new to him or her than for a patient who is previously known. This finding has face validity. However, as we have suggested, the extra time spent by the oncoming physician may have both advantages and disadvantages. The disadvantages include time‐consuming cognitive work for the physician and the potential for information loss affecting patient care. The potential advantage is a second physician reassessing the diagnosis and management decisions established by the first, potentially correcting errors and optimizing care.

Ultimately, more research is needed to illuminate the effect of hospitalist continuity on patient outcomes. For now, we feel that hospital medicine group leaders need not institute lengthy rotations or staffing models that prioritize continuity above all other factors, as continuity appears to have little impact on patient outcomes.

We greatly appreciate the thoughtful points made by Dr. Kerman regarding our recently published study evaluating the association of hospitalist continuity on adverse events (AEs).[1] We agree that a 7‐on/7‐off staffing model may limit discontinuity relative to models using shorter rotations lengths. Many hospital medicine programs use a 7‐on/7‐off model to optimize continuity. Longer rotation lengths are uncommon, as they may lead to fatigue and negatively affect physician work‐life balance. Shorter rotation lengths do exist, and we acknowledge that a study in a setting with greater fragmentation may have detected an effect.

We respectfully disagree with Dr. Kerman's concern that our methods for AE detection and confirmation may have been insensitive. We did not rely on incident reports, as these systems suffer from under‐reporting and often represent only a fraction of true AEs. We used a modified version of the classic 2‐stage method to identify and confirm AEs.[2] In the first stage, we used computerized screens, based on criteria from the Harvard Medical Practice Study and Institute for Healthcare Improvement global trigger tool, to identify potential AEs.[3, 4, 5] A research nurse created narrative summaries of potential AEs. A physician researcher then reviewed the narrative summaries to confirm whether an AE was truly present. This time‐consuming method is much more sensitive and specific than other options for patient safety measurement, including administrative data analyses and incident reporting systems.[6, 7]

With respect to other outcomes that may be affected by hospitalist continuity, we recently published a separate study showing that lower inpatient physician continuity was significantly associated with modest increases in hospital costs.[8] We found no association between continuity and patient satisfaction, but were likely underpowered to detect one. Interestingly, some of the models in our study suggested a slightly reduced risk of readmission with lower continuity. We were surprised by this finding and hypothesized that countervailing forces may be at play during handoffs of care from 1 hospitalist to another. Transitions of care introduce the opportunity for critical information to be lost, but they also introduce the potential for patient reassessment. A hospitalist newly taking over care from another may not be anchored to the initial diagnostic impressions and management plan established by the first. Of course, the potential benefit of a reassessment could only occur if the new hospitalist has time to perform one. At extremely high patient volumes, this theoretical benefit is unlikely to exist.

We did not include length of stay (LOS) as an outcome because hospitalist continuity and LOS are interdependent. Although discontinuity may lead to longer LOS, longer LOS definitely increases the probability of discontinuity. Thus, we controlled for LOS in our statistical models to isolate the effect of continuity. The study by Epstein and colleagues did not take into account the interdependence between LOS and hospitalist continuity.[9] Observational studies are not ideal for determining the effect of continuity on LOS. The Combing Incentives and Continuity Leading to Efficiency (CICLE) study by Chandra and colleagues was a pre‐post evaluation of a hospitalist staffing model specifically designed to improve continuity.[10] In the CICLE model, physicians work in a 4‐day rotation. On day 1, physicians exclusively admit patients. On day 2, physicians care for patients admitted on day 1 and accept patients admitted overnight. On days 3 and 4, physicians continue to care for patients received on days 1 and 2, but receive no additional patients. The remaining patients are transitioned to the next physician entering the cycle at the end of day 4. Chandra and colleagues found a 7.5% reduction in LOS and an 8.5% reduction in charges. Interestingly, they also found a 13.5% increase in readmissions that did not achieve statistical significance (P=0.08). The CICLE study suggests continuity does affect LOS, but is limited in that it did not account for a potential preexisting trend toward lower LOS.

Dr. Kerman presents data showing that it takes longer for a physician to care for a patient who is new to him or her than for a patient who is previously known. This finding has face validity. However, as we have suggested, the extra time spent by the oncoming physician may have both advantages and disadvantages. The disadvantages include time‐consuming cognitive work for the physician and the potential for information loss affecting patient care. The potential advantage is a second physician reassessing the diagnosis and management decisions established by the first, potentially correcting errors and optimizing care.

Ultimately, more research is needed to illuminate the effect of hospitalist continuity on patient outcomes. For now, we feel that hospital medicine group leaders need not institute lengthy rotations or staffing models that prioritize continuity above all other factors, as continuity appears to have little impact on patient outcomes.

Although definitions vary, continuity of care can be thought of as the patient's experience of a continuous caring relationship with an identified healthcare professional.[1] Research in ambulatory settings has found that patients who see their primary care physician for a higher proportion of office visits have higher patient satisfaction, better hypertensive control, lower risk of hospitalization, and fewer emergency department visits.[2, 3, 4, 5] Continuity with a single hospital‐based physician is difficult to achieve because of the need to provide care 24 hours a day, 7 days a week. Key clinical information may be lost during physician‐to‐physician handoffs (eg, at admission, at the end of rotations on service) during hospitalization. Our research group recently found that lower hospital physician continuity was associated with modestly increased hospital costs, but also a trend toward lower readmissions.[6] We speculated that physicians newly taking over patient care from colleagues reassess diagnoses and treatment plans. This reassessment may identify errors missed by the previous hospital physician. Thus, discontinuity may theoretically help or hinder the provision of safe hospital care.

We sought to examine the relationship between hospital physician continuity and the incidence of adverse events (AEs). We combined data from 2 previously published studies by our research group; one investigated the relationship between hospital physician continuity and costs and 30‐day readmissions, the other assessed the impact of unit‐based interventions on AEs.[6, 7]

METHODS

RESULTS

DISCUSSION

We found that hospitalist physician continuity was not associated with the incidence of AEs. Our findings are somewhat surprising because of the high value placed on continuity of care and patient safety concerns related to handoffs. Key clinical information may be lost when patient care is transitioned to a new hospitalist shortly after admission (eg, from a night hospitalist) or at the end of a rotation. Thus, it is logical to assume that discontinuity inherently increases the risk for harm. On the other hand, a physician newly taking over patient care from another may not be anchored to the initial diagnosis and treatment plan established by the first. This second look could potentially prevent missed/delayed diagnoses and optimize the plan of care.[19] These countervailing forces may explain our findings.

Several other potential explanations for our findings should be considered. First, the quality of handoffs may have been sufficient to overcome the potential for information loss. We feel this is unlikely given that little attention had been dedicated to improving the quality of patient handoffs among hospitalists in our institution. Notably, though a number of studies have evaluated resident physician handoffs, most of the work has focused on night coverage, and little is known about the quality of attending handoffs.[20] Second, access to a fully integrated electronic health record may have assisted hospitalists in complementing information received during handoffs. For example, a hospitalist about to start his or her rotation may have remotely accessed and reviewed patient medical records prior to receiving the phone handoff from the outgoing hospitalist. Third, other efforts to improve patient safety may have reduced the overall risk and provided some resilience in the system. Unit‐based interventions, including structured interdisciplinary rounds and nurse‐physician coleadership, improved teamwork climate and reduced AEs in the study hospital over time.[7]

Another factor to consider relates to the fact that hospital care is provided by teams of clinicians (eg, nurses, specialist physicians, therapists, social workers). Hospital teams are often large and have dynamic team membership. Similar to hospitalists, nurses, physician specialists, and other team members handoff care throughout the course of a patient's hospital stay. Yet, discontinuity for each professional type may occur at different times and frequencies. For example, a patient may be handed off from one hospitalist to another, yet the care continues with the same cardiologist or nurse. Future research should better characterize hospital team complexity (eg, size, relationships among members) and dynamics (eg, continuity for various professional types) and the impact of these factors on patient outcomes.

Our findings are important because hospitalist physician discontinuity is common during hospital stays. Hospital medicine groups vary in their staffing and scheduling models. Policies related to admission distribution and rotation length (consecutive days worked) systematically impact physician continuity. Few studies have evaluated the effect on continuity on hospitalized patient outcomes, and no prior research, to our knowledge, has explored the association of continuity on measures of patient safety.[6, 21, 22] Though our study might suggest that staffing models have little impact on patient safety, as previously mentioned, other team factors may influence patient outcomes.

Our study has several limitations. First, we assessed the impact of continuity on AEs in a single site. Although the 7 days on/7 days off model is the most common scheduling pattern used by adult hospital medicine groups,[23] staffing models and patient safety practices vary across hospitals, potentially limiting the generalizability of our study. Second, continuity can be defined and measured in a variety of ways. We used 2 different measures of physician continuity. As previously mentioned, assessing continuity of other clinicians may allow for a more complete understanding of the potential problems related to fragmentation of care. Third, this study excluded patients who experienced care transitions from other hospitals or other units within the hospital. Patients transferred from other hospitals are often complex, severely ill, and may be at higher risk for loss of key clinical information. Fourth, we used automated screens of an EDW to identify potential AEs. Although our prior research found that this method identified a similar number of AEs as manual medical record review screening, there was poor agreement between the 2 methods. Unfortunately, there is no gold standard to identify AEs. The EDW‐facilitated method allowed us to feasibly screen a larger number of charts, increasing statistical power, and minimized any potential bias that might occur during a manual review to identify potential AEs. Finally, we used data available from 2 prior studies and may have been underpowered to detect a significant association between continuity and AEs due to the relatively low percentage of patients experiencing an AE. In a post hoc power calculation, we estimated that we had 70% power to detect a 33% change in the proportion of patients with 1 or more AE for each 1‐unit increase in NPI, and 80% power to detect a 20% change for each 0.1‐unit decrease in UPC.

CONCLUSION

In conclusion, we found that hospitalist physician continuity was not associated with the incidence of AEs. We speculate that hospitalist continuity is only 1 of many team factors that may influence patient safety, and that prior efforts within our institution may have reduced our ability to detect an association. Future research should better characterize hospital team complexity and dynamics and the impact of these factors on patient outcomes.

Although definitions vary, continuity of care can be thought of as the patient's experience of a continuous caring relationship with an identified healthcare professional.[1] Research in ambulatory settings has found that patients who see their primary care physician for a higher proportion of office visits have higher patient satisfaction, better hypertensive control, lower risk of hospitalization, and fewer emergency department visits.[2, 3, 4, 5] Continuity with a single hospital‐based physician is difficult to achieve because of the need to provide care 24 hours a day, 7 days a week. Key clinical information may be lost during physician‐to‐physician handoffs (eg, at admission, at the end of rotations on service) during hospitalization. Our research group recently found that lower hospital physician continuity was associated with modestly increased hospital costs, but also a trend toward lower readmissions.[6] We speculated that physicians newly taking over patient care from colleagues reassess diagnoses and treatment plans. This reassessment may identify errors missed by the previous hospital physician. Thus, discontinuity may theoretically help or hinder the provision of safe hospital care.

We sought to examine the relationship between hospital physician continuity and the incidence of adverse events (AEs). We combined data from 2 previously published studies by our research group; one investigated the relationship between hospital physician continuity and costs and 30‐day readmissions, the other assessed the impact of unit‐based interventions on AEs.[6, 7]

METHODS

RESULTS

DISCUSSION

We found that hospitalist physician continuity was not associated with the incidence of AEs. Our findings are somewhat surprising because of the high value placed on continuity of care and patient safety concerns related to handoffs. Key clinical information may be lost when patient care is transitioned to a new hospitalist shortly after admission (eg, from a night hospitalist) or at the end of a rotation. Thus, it is logical to assume that discontinuity inherently increases the risk for harm. On the other hand, a physician newly taking over patient care from another may not be anchored to the initial diagnosis and treatment plan established by the first. This second look could potentially prevent missed/delayed diagnoses and optimize the plan of care.[19] These countervailing forces may explain our findings.

Several other potential explanations for our findings should be considered. First, the quality of handoffs may have been sufficient to overcome the potential for information loss. We feel this is unlikely given that little attention had been dedicated to improving the quality of patient handoffs among hospitalists in our institution. Notably, though a number of studies have evaluated resident physician handoffs, most of the work has focused on night coverage, and little is known about the quality of attending handoffs.[20] Second, access to a fully integrated electronic health record may have assisted hospitalists in complementing information received during handoffs. For example, a hospitalist about to start his or her rotation may have remotely accessed and reviewed patient medical records prior to receiving the phone handoff from the outgoing hospitalist. Third, other efforts to improve patient safety may have reduced the overall risk and provided some resilience in the system. Unit‐based interventions, including structured interdisciplinary rounds and nurse‐physician coleadership, improved teamwork climate and reduced AEs in the study hospital over time.[7]

Another factor to consider relates to the fact that hospital care is provided by teams of clinicians (eg, nurses, specialist physicians, therapists, social workers). Hospital teams are often large and have dynamic team membership. Similar to hospitalists, nurses, physician specialists, and other team members handoff care throughout the course of a patient's hospital stay. Yet, discontinuity for each professional type may occur at different times and frequencies. For example, a patient may be handed off from one hospitalist to another, yet the care continues with the same cardiologist or nurse. Future research should better characterize hospital team complexity (eg, size, relationships among members) and dynamics (eg, continuity for various professional types) and the impact of these factors on patient outcomes.

Our findings are important because hospitalist physician discontinuity is common during hospital stays. Hospital medicine groups vary in their staffing and scheduling models. Policies related to admission distribution and rotation length (consecutive days worked) systematically impact physician continuity. Few studies have evaluated the effect on continuity on hospitalized patient outcomes, and no prior research, to our knowledge, has explored the association of continuity on measures of patient safety.[6, 21, 22] Though our study might suggest that staffing models have little impact on patient safety, as previously mentioned, other team factors may influence patient outcomes.

Our study has several limitations. First, we assessed the impact of continuity on AEs in a single site. Although the 7 days on/7 days off model is the most common scheduling pattern used by adult hospital medicine groups,[23] staffing models and patient safety practices vary across hospitals, potentially limiting the generalizability of our study. Second, continuity can be defined and measured in a variety of ways. We used 2 different measures of physician continuity. As previously mentioned, assessing continuity of other clinicians may allow for a more complete understanding of the potential problems related to fragmentation of care. Third, this study excluded patients who experienced care transitions from other hospitals or other units within the hospital. Patients transferred from other hospitals are often complex, severely ill, and may be at higher risk for loss of key clinical information. Fourth, we used automated screens of an EDW to identify potential AEs. Although our prior research found that this method identified a similar number of AEs as manual medical record review screening, there was poor agreement between the 2 methods. Unfortunately, there is no gold standard to identify AEs. The EDW‐facilitated method allowed us to feasibly screen a larger number of charts, increasing statistical power, and minimized any potential bias that might occur during a manual review to identify potential AEs. Finally, we used data available from 2 prior studies and may have been underpowered to detect a significant association between continuity and AEs due to the relatively low percentage of patients experiencing an AE. In a post hoc power calculation, we estimated that we had 70% power to detect a 33% change in the proportion of patients with 1 or more AE for each 1‐unit increase in NPI, and 80% power to detect a 20% change for each 0.1‐unit decrease in UPC.

CONCLUSION

In conclusion, we found that hospitalist physician continuity was not associated with the incidence of AEs. We speculate that hospitalist continuity is only 1 of many team factors that may influence patient safety, and that prior efforts within our institution may have reduced our ability to detect an association. Future research should better characterize hospital team complexity and dynamics and the impact of these factors on patient outcomes.

Although definitions vary, continuity of care can be thought of as the patient's experience of a continuous caring relationship with an identified healthcare professional.[1] Research in ambulatory settings has found that patients who see their primary care physician for a higher proportion of office visits have higher patient satisfaction, better hypertensive control, lower risk of hospitalization, and fewer emergency department visits.[2, 3, 4, 5] Continuity with a single hospital‐based physician is difficult to achieve because of the need to provide care 24 hours a day, 7 days a week. Key clinical information may be lost during physician‐to‐physician handoffs (eg, at admission, at the end of rotations on service) during hospitalization. Our research group recently found that lower hospital physician continuity was associated with modestly increased hospital costs, but also a trend toward lower readmissions.[6] We speculated that physicians newly taking over patient care from colleagues reassess diagnoses and treatment plans. This reassessment may identify errors missed by the previous hospital physician. Thus, discontinuity may theoretically help or hinder the provision of safe hospital care.

We sought to examine the relationship between hospital physician continuity and the incidence of adverse events (AEs). We combined data from 2 previously published studies by our research group; one investigated the relationship between hospital physician continuity and costs and 30‐day readmissions, the other assessed the impact of unit‐based interventions on AEs.[6, 7]

METHODS

RESULTS

DISCUSSION

We found that hospitalist physician continuity was not associated with the incidence of AEs. Our findings are somewhat surprising because of the high value placed on continuity of care and patient safety concerns related to handoffs. Key clinical information may be lost when patient care is transitioned to a new hospitalist shortly after admission (eg, from a night hospitalist) or at the end of a rotation. Thus, it is logical to assume that discontinuity inherently increases the risk for harm. On the other hand, a physician newly taking over patient care from another may not be anchored to the initial diagnosis and treatment plan established by the first. This second look could potentially prevent missed/delayed diagnoses and optimize the plan of care.[19] These countervailing forces may explain our findings.

Several other potential explanations for our findings should be considered. First, the quality of handoffs may have been sufficient to overcome the potential for information loss. We feel this is unlikely given that little attention had been dedicated to improving the quality of patient handoffs among hospitalists in our institution. Notably, though a number of studies have evaluated resident physician handoffs, most of the work has focused on night coverage, and little is known about the quality of attending handoffs.[20] Second, access to a fully integrated electronic health record may have assisted hospitalists in complementing information received during handoffs. For example, a hospitalist about to start his or her rotation may have remotely accessed and reviewed patient medical records prior to receiving the phone handoff from the outgoing hospitalist. Third, other efforts to improve patient safety may have reduced the overall risk and provided some resilience in the system. Unit‐based interventions, including structured interdisciplinary rounds and nurse‐physician coleadership, improved teamwork climate and reduced AEs in the study hospital over time.[7]

Another factor to consider relates to the fact that hospital care is provided by teams of clinicians (eg, nurses, specialist physicians, therapists, social workers). Hospital teams are often large and have dynamic team membership. Similar to hospitalists, nurses, physician specialists, and other team members handoff care throughout the course of a patient's hospital stay. Yet, discontinuity for each professional type may occur at different times and frequencies. For example, a patient may be handed off from one hospitalist to another, yet the care continues with the same cardiologist or nurse. Future research should better characterize hospital team complexity (eg, size, relationships among members) and dynamics (eg, continuity for various professional types) and the impact of these factors on patient outcomes.

Our findings are important because hospitalist physician discontinuity is common during hospital stays. Hospital medicine groups vary in their staffing and scheduling models. Policies related to admission distribution and rotation length (consecutive days worked) systematically impact physician continuity. Few studies have evaluated the effect on continuity on hospitalized patient outcomes, and no prior research, to our knowledge, has explored the association of continuity on measures of patient safety.[6, 21, 22] Though our study might suggest that staffing models have little impact on patient safety, as previously mentioned, other team factors may influence patient outcomes.

Our study has several limitations. First, we assessed the impact of continuity on AEs in a single site. Although the 7 days on/7 days off model is the most common scheduling pattern used by adult hospital medicine groups,[23] staffing models and patient safety practices vary across hospitals, potentially limiting the generalizability of our study. Second, continuity can be defined and measured in a variety of ways. We used 2 different measures of physician continuity. As previously mentioned, assessing continuity of other clinicians may allow for a more complete understanding of the potential problems related to fragmentation of care. Third, this study excluded patients who experienced care transitions from other hospitals or other units within the hospital. Patients transferred from other hospitals are often complex, severely ill, and may be at higher risk for loss of key clinical information. Fourth, we used automated screens of an EDW to identify potential AEs. Although our prior research found that this method identified a similar number of AEs as manual medical record review screening, there was poor agreement between the 2 methods. Unfortunately, there is no gold standard to identify AEs. The EDW‐facilitated method allowed us to feasibly screen a larger number of charts, increasing statistical power, and minimized any potential bias that might occur during a manual review to identify potential AEs. Finally, we used data available from 2 prior studies and may have been underpowered to detect a significant association between continuity and AEs due to the relatively low percentage of patients experiencing an AE. In a post hoc power calculation, we estimated that we had 70% power to detect a 33% change in the proportion of patients with 1 or more AE for each 1‐unit increase in NPI, and 80% power to detect a 20% change for each 0.1‐unit decrease in UPC.

CONCLUSION

In conclusion, we found that hospitalist physician continuity was not associated with the incidence of AEs. We speculate that hospitalist continuity is only 1 of many team factors that may influence patient safety, and that prior efforts within our institution may have reduced our ability to detect an association. Future research should better characterize hospital team complexity and dynamics and the impact of these factors on patient outcomes.