Enactment of federal legislation imposing hospital reimbursement penalties for excess rates of rehospitalizations among Medicare fee for service beneficiaries markedly increased interest in hospital quality improvement (QI) efforts to reduce the observed 30‐day rehospitalization of 19.6% in this elderly population.[1, 2] The Congressional Budget Office estimated that reimbursement penalties to hospitals for high readmission rates are expected to save the Medicare program approximately $7 billion between 2010 and 2019.[3] These penalties are complemented by resources from the Center for Medicare and Medicaid Innovation aiming to reduce hospital readmissions by 20% by the end of 2013 through the Partnership for Patients campaign.[4] Although potential financial penalties and provision of resources for QI intensified efforts to enhance the quality of the hospital discharge transition, patient safety risks associated with hospital discharge are well documented.[5, 6] Approximately 20% of patients discharged from the hospital may suffer adverse events,[7, 8] of which up to three‐quarters (72%) are medication related,[9] and over one‐third of required follow‐up testing after discharge is not completed.[10] Such findings indicate opportunities for improvement in the discharge process.[11]

Numerous publications describe studies aiming to improve the hospital discharge process and mitigate these hazards, though a systematic review of interventions to reduce 30‐day rehospitalization indicated that the existing evidence base for the effectiveness of transition interventions demonstrates irregular effectiveness and limitations to generalizability.[12] Most studies showing effectiveness are confined to single academic medical centers. Existing evidence supports multifaceted interventions implemented in both the pre‐ and postdischarge periods and focused on risk assessment and tailored, patient‐centered application of interventions to mitigate risk. For example Project RED (Re‐Engineered Discharge) applied a bundled intervention consisting of intensified patient education and discharge planning, improved medication reconciliation and discharge instructions, and longitudinal patient contact with follow‐up phone calls and a dedicated discharge advocate.[13] However, the mean age of patients participating in the study was 50 years, and it excluded patients admitted from or discharged to skilled nursing facilities, making generalizability to the geriatric population uncertain.

An integral aspect of QI projects is the contribution of local context to translation of best practices to disparate settings.[14, 15, 16] Most available reports of successful interventions to reduce rehospitalization have not fully described the specifics of either the intervention context or design. Moreover, the available evidence base for common interventions to reduce rehospitalization was developed in the academic setting. Validation of single academic center studies in a broader healthcare context is necessary.

Project BOOST (Better Outcomes for Older adults through Safe Transitions) recruited a diverse national cohort of both academic and nonacademic hospitals to participate in a QI effort to implement best practices for hospital discharge care transitions using a national collaborative approach facilitated by external expert mentorship. This study aimed to determine the effectiveness of BOOST in lowering hospital readmission rates and impact on length of stay.

METHODS

The study of Project BOOST was undertaken in accordance with the SQUIRE (Standards for Quality Improvement Reporting Excellence) Guidelines.[17]

RESULTS

Eleven hospitals provided rehospitalization and length‐of‐stay outcome data for both a BOOST and control unit for the pre‐ and postimplementation periods. Compared to the 19 sites that did not participate in the analysis, these 11 sites were significantly larger (559188 beds vs 350205 beds, P=0.003), more likely to be located in an urban area (100.0% [n=11] vs 78.9% [n=15], P=0.035), and more likely to be academic medical centers (45.5% [n=5] vs 26.3% [n=5], P=0.036) (Table 1).

The mean number of tools implemented by sites participating in the analysis was 3.50.9. All sites implemented at least 2 tools. The duration between attendance at the BOOST kickoff event and first tool implementation ranged from 3 months (first tool implemented prior to attending the kickoff) and 9 months (mean duration, 3.34.3 months) (Table 2).

The average rate of 30‐day rehospitalization among BOOST units was 14.7% in the preimplementation period and 12.7% during the postimplementation period (P=0.010) (Figure 1). Rehospitalization rates for matched control units were 14.0% in the preintervention period and 14.1% in the postintervention period (P=0.831). The mean absolute reduction in readmission rates over the 1‐year study period in BOOST units compared to control units was 2.0%, or a relative reduction of 13.6% (P=0.054 for signed rank test comparing differences in readmission rate reduction in BOOST units compared to site‐matched control units). Length of stay in BOOST and control units decreased an average of 0.5 days and 0.3 days, respectively. There was no difference in length of stay change between BOOST units and control units (P=0.966).

Figure 1

DISCUSSION

As hospitals strive to reduce their readmission rates to avoid Centers for Medicare and Medicaid Services penalties, Project BOOST may be a viable QI approach to achieve their goals. This initial evaluation of participation in Project BOOST by 11 hospitals of varying sizes across the United States showed an associated reduction in rehospitalization rates (absolute=2.0% and relative=13.6%, P=0.054). We did not find any significant change in length of stay among these hospitals implementing BOOST tools.

The tools provided to participating hospitals were developed from evidence found in peer‐reviewed literature established through experimental methods in well‐controlled academic settings. Further tool development was informed by recommendations of an advisory board consisting of expert representatives and advocates involved in the hospital discharge process: patients, caregivers, physicians, nurses, case managers, social workers, insurers, and regulatory and research agencies.[19] The toolkit components address multiple aspects of hospital discharge and follow‐up with the goal of improving health by optimizing the safety of care transitions. Our observation that readmission rates appeared to improve in a diverse hospital sample including nonacademic and community hospitals engaged in Project BOOST is reassuring that the benefits seen in existing research literature, developed in distinctly academic settings, can be replicated in diverse acute‐care settings.

The effect size observed in our study was modest but consistent with several studies identified in a recent review of trials measuring interventions to reduce rehospitalization, where 7 of 16 studies showing a significant improvement registered change in the 0% to 5% absolute range.[12] Impact of this project may have been tempered by the need to translate external QI content to the local setting. Additionally, in contrast to experimental studies that are limited in scope and timing and often scaled to a research budget, BOOST sites were encouraged to implement Project BOOST in the clinical setting even if no new funds were available to support the effort.[12]

The recruitment of a national sample of both academic and nonacademic hospital participants imposed several limitations on our study and analysis. We recognize that intervention units selected by hospitals may have had unmeasured unit and patient characteristics that facilitated successful change and contributed to the observed improvements. However, because external pressure to reduce readmission is present across all hospitals independent of the BOOST intervention, we felt site‐matched controls were essential to understanding effects attributable to the BOOST tools. Differences between units would be expected to be stable over the course of the study period, and comparison of outcome differences between 2 different time periods would be reasonable. Additionally, we could not collect data on readmissions to other hospitals. Theoretically, patients discharged from BOOST units might be more likely to have been rehospitalized elsewhere, but the fraction of rehospitalizations occurring at alternate facilities would also be expected to be similar on the matched control unit.

We report findings from a voluntary cohort willing and capable of designating a comparison clinical unit and contributing the requested data outcomes. Pilot sites that did not report outcomes were not analyzed, but comparison of hospital characteristics shows that participating hospitals were more likely to be large, urban, academic medical centers. Although barriers to data submission were not formally analyzed, reports from nonparticipating sites describe data submission limited by local implementation design (no geographic rollout or simultaneous rollout on all appropriate clinical units), site specific inability to generate unit level outcome statistics, and competing organizational priorities for data analyst time (electronic medical record deployment, alternative QI initiatives). The external validity of our results may be limited to organizations capable of analytics at the level of the individual clinical unit as well as those with sufficient QI resources to support reporting to a national database in the absence of a payer mandate. It is possible that additional financial support for on‐site data collection would have bolstered participation, making the example of participation rates we present potentially informative to organizations hoping to widely disseminate a QI agenda.

Nonetheless, the effectiveness demonstrated in the 11 sites that did participate is encouraging, and ongoing collaboration with subsequent BOOST cohorts has been designed to further facilitate data collection. Among the insights gained from this pilot experience, and incorporated into ongoing BOOST cohorts, is the importance of intensive mentor engagement to foster accountability among participant sites, assist with implementation troubleshooting, and offer expertise that is often particularly effective in gaining local support. We now encourage sites to have 2 mentor site visits to further these roles and more frequent conference calls. Further research to understand the marginal benefit of the mentored implementation approach is ongoing.

The limitations in data submission we experienced with the pilot cohort likely reflect resource constraints not uncommon at many hospitals. Increasing pressure placed on hospitals as a result of the Readmission Reduction Program within the Affordable Care Act as well as increasing interest from private and Medicaid payors to incorporate similar readmission‐based penalties provide encouragement for hospitals to enhance their data and analytic skills. National incentives for implementation of electronic health records (EHR) should also foster such capabilities, though we often saw EHRs as a barrier to QI, especially rapid cycle trials. Fortunately, hospitals are increasingly being afforded access to comprehensive claims databases to assist in tracking readmission rates to other facilities, and these data are becoming available in a more timely fashion. This more robust data collection, facilitated by private payors, state QI organizations, and state hospital associations, will support additional analytic methods such as multivariate regression models and interrupted time series designs to appreciate the experience of current BOOST participants.

Additional research is needed to understand the role of organizational context in the effectiveness of Project BOOST. Differences in rates of tool implementation and changes in clinical outcomes are likely dependent on local implementation context at the level of the healthcare organization and individual clinical unit.[20] Progress reports from site mentors and previously described experiences of QI implementation indicate that successful implementation of a multidimensional bundle of interventions may have reflected a higher level of institutional support, more robust team engagement in the work of reducing readmissions, increased clinical staff support for change, the presence of an effective project champion, or a key facilitating role of external mentorship.[21, 22] Ongoing data collection will continue to measure the sustainability of tool use and observed outcome changes to inform strategies to maintain gains associated with implementation. The role of mentored implementation in facilitating gains also requires further study.

Increasing attention to the problem of avoidable rehospitalization is driving hospitals, insurers, and policy makers to pursue QI efforts that favorably impact readmission rates. Our analysis of the BOOST intervention suggests that modest gains can be achieved following evidence‐based hospital process change facilitated by a mentored implementation model. However, realization of the goal of a 20% reduction in rehospitalization proposed by the Center for Medicare and Medicaid Services' Partnership for Patients initiative may be difficult to achieve on a national scale,[23] especially if efforts focus on just the hospital.

Disclosures

Project BOOST was funded by a grant from The John A. Hartford Foundation. Project BOOST is administered by the Society of Hospital Medicine (SHM). The development of the Project BOOST toolkit, recruitment of sites for this study, mentorship of the pilot cohort, project evaluation planning, and collection of pilot data were funded by a grant from The John A. Harford Foundation. Additional funding for continued data collection and analysis was funded by the SHM through funds from hospitals to participate in Project BOOST, specifically with funding support for Dr. Hansen. Dr. Williams has received funding to serve as Principal Investigator for Project BOOST. Since the time of initial cohort participation, approximately 125 additional hospitals have participated in the mentored implementation of Project BOOST. This participation was funded through a combination of site‐based tuition, third‐party payor support from private insurers, foundations, and federal funding through the Center for Medicare and Medicaid Innovation Partnership for Patients program. Drs. Greenwald, Hansen, and Williams are Project BOOST mentors for current Project BOOST sites and receive financial support through the SHM for this work. Dr. Howell has previously received funding as a Project BOOST mentor. Ms. Budnitz is the BOOST Project Director and is Chief Strategy and Development Officer for the HM. Dr. Maynard is the Senior Vice President of the SHM's Center for Hospital Innovation and Improvement.

Enactment of federal legislation imposing hospital reimbursement penalties for excess rates of rehospitalizations among Medicare fee for service beneficiaries markedly increased interest in hospital quality improvement (QI) efforts to reduce the observed 30‐day rehospitalization of 19.6% in this elderly population.[1, 2] The Congressional Budget Office estimated that reimbursement penalties to hospitals for high readmission rates are expected to save the Medicare program approximately $7 billion between 2010 and 2019.[3] These penalties are complemented by resources from the Center for Medicare and Medicaid Innovation aiming to reduce hospital readmissions by 20% by the end of 2013 through the Partnership for Patients campaign.[4] Although potential financial penalties and provision of resources for QI intensified efforts to enhance the quality of the hospital discharge transition, patient safety risks associated with hospital discharge are well documented.[5, 6] Approximately 20% of patients discharged from the hospital may suffer adverse events,[7, 8] of which up to three‐quarters (72%) are medication related,[9] and over one‐third of required follow‐up testing after discharge is not completed.[10] Such findings indicate opportunities for improvement in the discharge process.[11]

Numerous publications describe studies aiming to improve the hospital discharge process and mitigate these hazards, though a systematic review of interventions to reduce 30‐day rehospitalization indicated that the existing evidence base for the effectiveness of transition interventions demonstrates irregular effectiveness and limitations to generalizability.[12] Most studies showing effectiveness are confined to single academic medical centers. Existing evidence supports multifaceted interventions implemented in both the pre‐ and postdischarge periods and focused on risk assessment and tailored, patient‐centered application of interventions to mitigate risk. For example Project RED (Re‐Engineered Discharge) applied a bundled intervention consisting of intensified patient education and discharge planning, improved medication reconciliation and discharge instructions, and longitudinal patient contact with follow‐up phone calls and a dedicated discharge advocate.[13] However, the mean age of patients participating in the study was 50 years, and it excluded patients admitted from or discharged to skilled nursing facilities, making generalizability to the geriatric population uncertain.

An integral aspect of QI projects is the contribution of local context to translation of best practices to disparate settings.[14, 15, 16] Most available reports of successful interventions to reduce rehospitalization have not fully described the specifics of either the intervention context or design. Moreover, the available evidence base for common interventions to reduce rehospitalization was developed in the academic setting. Validation of single academic center studies in a broader healthcare context is necessary.

Project BOOST (Better Outcomes for Older adults through Safe Transitions) recruited a diverse national cohort of both academic and nonacademic hospitals to participate in a QI effort to implement best practices for hospital discharge care transitions using a national collaborative approach facilitated by external expert mentorship. This study aimed to determine the effectiveness of BOOST in lowering hospital readmission rates and impact on length of stay.

METHODS

The study of Project BOOST was undertaken in accordance with the SQUIRE (Standards for Quality Improvement Reporting Excellence) Guidelines.[17]

RESULTS

Eleven hospitals provided rehospitalization and length‐of‐stay outcome data for both a BOOST and control unit for the pre‐ and postimplementation periods. Compared to the 19 sites that did not participate in the analysis, these 11 sites were significantly larger (559188 beds vs 350205 beds, P=0.003), more likely to be located in an urban area (100.0% [n=11] vs 78.9% [n=15], P=0.035), and more likely to be academic medical centers (45.5% [n=5] vs 26.3% [n=5], P=0.036) (Table 1).

The mean number of tools implemented by sites participating in the analysis was 3.50.9. All sites implemented at least 2 tools. The duration between attendance at the BOOST kickoff event and first tool implementation ranged from 3 months (first tool implemented prior to attending the kickoff) and 9 months (mean duration, 3.34.3 months) (Table 2).

The average rate of 30‐day rehospitalization among BOOST units was 14.7% in the preimplementation period and 12.7% during the postimplementation period (P=0.010) (Figure 1). Rehospitalization rates for matched control units were 14.0% in the preintervention period and 14.1% in the postintervention period (P=0.831). The mean absolute reduction in readmission rates over the 1‐year study period in BOOST units compared to control units was 2.0%, or a relative reduction of 13.6% (P=0.054 for signed rank test comparing differences in readmission rate reduction in BOOST units compared to site‐matched control units). Length of stay in BOOST and control units decreased an average of 0.5 days and 0.3 days, respectively. There was no difference in length of stay change between BOOST units and control units (P=0.966).

Figure 1

DISCUSSION

As hospitals strive to reduce their readmission rates to avoid Centers for Medicare and Medicaid Services penalties, Project BOOST may be a viable QI approach to achieve their goals. This initial evaluation of participation in Project BOOST by 11 hospitals of varying sizes across the United States showed an associated reduction in rehospitalization rates (absolute=2.0% and relative=13.6%, P=0.054). We did not find any significant change in length of stay among these hospitals implementing BOOST tools.

The tools provided to participating hospitals were developed from evidence found in peer‐reviewed literature established through experimental methods in well‐controlled academic settings. Further tool development was informed by recommendations of an advisory board consisting of expert representatives and advocates involved in the hospital discharge process: patients, caregivers, physicians, nurses, case managers, social workers, insurers, and regulatory and research agencies.[19] The toolkit components address multiple aspects of hospital discharge and follow‐up with the goal of improving health by optimizing the safety of care transitions. Our observation that readmission rates appeared to improve in a diverse hospital sample including nonacademic and community hospitals engaged in Project BOOST is reassuring that the benefits seen in existing research literature, developed in distinctly academic settings, can be replicated in diverse acute‐care settings.

The effect size observed in our study was modest but consistent with several studies identified in a recent review of trials measuring interventions to reduce rehospitalization, where 7 of 16 studies showing a significant improvement registered change in the 0% to 5% absolute range.[12] Impact of this project may have been tempered by the need to translate external QI content to the local setting. Additionally, in contrast to experimental studies that are limited in scope and timing and often scaled to a research budget, BOOST sites were encouraged to implement Project BOOST in the clinical setting even if no new funds were available to support the effort.[12]

The recruitment of a national sample of both academic and nonacademic hospital participants imposed several limitations on our study and analysis. We recognize that intervention units selected by hospitals may have had unmeasured unit and patient characteristics that facilitated successful change and contributed to the observed improvements. However, because external pressure to reduce readmission is present across all hospitals independent of the BOOST intervention, we felt site‐matched controls were essential to understanding effects attributable to the BOOST tools. Differences between units would be expected to be stable over the course of the study period, and comparison of outcome differences between 2 different time periods would be reasonable. Additionally, we could not collect data on readmissions to other hospitals. Theoretically, patients discharged from BOOST units might be more likely to have been rehospitalized elsewhere, but the fraction of rehospitalizations occurring at alternate facilities would also be expected to be similar on the matched control unit.

We report findings from a voluntary cohort willing and capable of designating a comparison clinical unit and contributing the requested data outcomes. Pilot sites that did not report outcomes were not analyzed, but comparison of hospital characteristics shows that participating hospitals were more likely to be large, urban, academic medical centers. Although barriers to data submission were not formally analyzed, reports from nonparticipating sites describe data submission limited by local implementation design (no geographic rollout or simultaneous rollout on all appropriate clinical units), site specific inability to generate unit level outcome statistics, and competing organizational priorities for data analyst time (electronic medical record deployment, alternative QI initiatives). The external validity of our results may be limited to organizations capable of analytics at the level of the individual clinical unit as well as those with sufficient QI resources to support reporting to a national database in the absence of a payer mandate. It is possible that additional financial support for on‐site data collection would have bolstered participation, making the example of participation rates we present potentially informative to organizations hoping to widely disseminate a QI agenda.

Nonetheless, the effectiveness demonstrated in the 11 sites that did participate is encouraging, and ongoing collaboration with subsequent BOOST cohorts has been designed to further facilitate data collection. Among the insights gained from this pilot experience, and incorporated into ongoing BOOST cohorts, is the importance of intensive mentor engagement to foster accountability among participant sites, assist with implementation troubleshooting, and offer expertise that is often particularly effective in gaining local support. We now encourage sites to have 2 mentor site visits to further these roles and more frequent conference calls. Further research to understand the marginal benefit of the mentored implementation approach is ongoing.

The limitations in data submission we experienced with the pilot cohort likely reflect resource constraints not uncommon at many hospitals. Increasing pressure placed on hospitals as a result of the Readmission Reduction Program within the Affordable Care Act as well as increasing interest from private and Medicaid payors to incorporate similar readmission‐based penalties provide encouragement for hospitals to enhance their data and analytic skills. National incentives for implementation of electronic health records (EHR) should also foster such capabilities, though we often saw EHRs as a barrier to QI, especially rapid cycle trials. Fortunately, hospitals are increasingly being afforded access to comprehensive claims databases to assist in tracking readmission rates to other facilities, and these data are becoming available in a more timely fashion. This more robust data collection, facilitated by private payors, state QI organizations, and state hospital associations, will support additional analytic methods such as multivariate regression models and interrupted time series designs to appreciate the experience of current BOOST participants.

Additional research is needed to understand the role of organizational context in the effectiveness of Project BOOST. Differences in rates of tool implementation and changes in clinical outcomes are likely dependent on local implementation context at the level of the healthcare organization and individual clinical unit.[20] Progress reports from site mentors and previously described experiences of QI implementation indicate that successful implementation of a multidimensional bundle of interventions may have reflected a higher level of institutional support, more robust team engagement in the work of reducing readmissions, increased clinical staff support for change, the presence of an effective project champion, or a key facilitating role of external mentorship.[21, 22] Ongoing data collection will continue to measure the sustainability of tool use and observed outcome changes to inform strategies to maintain gains associated with implementation. The role of mentored implementation in facilitating gains also requires further study.

Increasing attention to the problem of avoidable rehospitalization is driving hospitals, insurers, and policy makers to pursue QI efforts that favorably impact readmission rates. Our analysis of the BOOST intervention suggests that modest gains can be achieved following evidence‐based hospital process change facilitated by a mentored implementation model. However, realization of the goal of a 20% reduction in rehospitalization proposed by the Center for Medicare and Medicaid Services' Partnership for Patients initiative may be difficult to achieve on a national scale,[23] especially if efforts focus on just the hospital.

Disclosures

Project BOOST was funded by a grant from The John A. Hartford Foundation. Project BOOST is administered by the Society of Hospital Medicine (SHM). The development of the Project BOOST toolkit, recruitment of sites for this study, mentorship of the pilot cohort, project evaluation planning, and collection of pilot data were funded by a grant from The John A. Harford Foundation. Additional funding for continued data collection and analysis was funded by the SHM through funds from hospitals to participate in Project BOOST, specifically with funding support for Dr. Hansen. Dr. Williams has received funding to serve as Principal Investigator for Project BOOST. Since the time of initial cohort participation, approximately 125 additional hospitals have participated in the mentored implementation of Project BOOST. This participation was funded through a combination of site‐based tuition, third‐party payor support from private insurers, foundations, and federal funding through the Center for Medicare and Medicaid Innovation Partnership for Patients program. Drs. Greenwald, Hansen, and Williams are Project BOOST mentors for current Project BOOST sites and receive financial support through the SHM for this work. Dr. Howell has previously received funding as a Project BOOST mentor. Ms. Budnitz is the BOOST Project Director and is Chief Strategy and Development Officer for the HM. Dr. Maynard is the Senior Vice President of the SHM's Center for Hospital Innovation and Improvement.

Enactment of federal legislation imposing hospital reimbursement penalties for excess rates of rehospitalizations among Medicare fee for service beneficiaries markedly increased interest in hospital quality improvement (QI) efforts to reduce the observed 30‐day rehospitalization of 19.6% in this elderly population.[1, 2] The Congressional Budget Office estimated that reimbursement penalties to hospitals for high readmission rates are expected to save the Medicare program approximately $7 billion between 2010 and 2019.[3] These penalties are complemented by resources from the Center for Medicare and Medicaid Innovation aiming to reduce hospital readmissions by 20% by the end of 2013 through the Partnership for Patients campaign.[4] Although potential financial penalties and provision of resources for QI intensified efforts to enhance the quality of the hospital discharge transition, patient safety risks associated with hospital discharge are well documented.[5, 6] Approximately 20% of patients discharged from the hospital may suffer adverse events,[7, 8] of which up to three‐quarters (72%) are medication related,[9] and over one‐third of required follow‐up testing after discharge is not completed.[10] Such findings indicate opportunities for improvement in the discharge process.[11]

Numerous publications describe studies aiming to improve the hospital discharge process and mitigate these hazards, though a systematic review of interventions to reduce 30‐day rehospitalization indicated that the existing evidence base for the effectiveness of transition interventions demonstrates irregular effectiveness and limitations to generalizability.[12] Most studies showing effectiveness are confined to single academic medical centers. Existing evidence supports multifaceted interventions implemented in both the pre‐ and postdischarge periods and focused on risk assessment and tailored, patient‐centered application of interventions to mitigate risk. For example Project RED (Re‐Engineered Discharge) applied a bundled intervention consisting of intensified patient education and discharge planning, improved medication reconciliation and discharge instructions, and longitudinal patient contact with follow‐up phone calls and a dedicated discharge advocate.[13] However, the mean age of patients participating in the study was 50 years, and it excluded patients admitted from or discharged to skilled nursing facilities, making generalizability to the geriatric population uncertain.

An integral aspect of QI projects is the contribution of local context to translation of best practices to disparate settings.[14, 15, 16] Most available reports of successful interventions to reduce rehospitalization have not fully described the specifics of either the intervention context or design. Moreover, the available evidence base for common interventions to reduce rehospitalization was developed in the academic setting. Validation of single academic center studies in a broader healthcare context is necessary.

Project BOOST (Better Outcomes for Older adults through Safe Transitions) recruited a diverse national cohort of both academic and nonacademic hospitals to participate in a QI effort to implement best practices for hospital discharge care transitions using a national collaborative approach facilitated by external expert mentorship. This study aimed to determine the effectiveness of BOOST in lowering hospital readmission rates and impact on length of stay.

METHODS

The study of Project BOOST was undertaken in accordance with the SQUIRE (Standards for Quality Improvement Reporting Excellence) Guidelines.[17]

RESULTS

Eleven hospitals provided rehospitalization and length‐of‐stay outcome data for both a BOOST and control unit for the pre‐ and postimplementation periods. Compared to the 19 sites that did not participate in the analysis, these 11 sites were significantly larger (559188 beds vs 350205 beds, P=0.003), more likely to be located in an urban area (100.0% [n=11] vs 78.9% [n=15], P=0.035), and more likely to be academic medical centers (45.5% [n=5] vs 26.3% [n=5], P=0.036) (Table 1).

The mean number of tools implemented by sites participating in the analysis was 3.50.9. All sites implemented at least 2 tools. The duration between attendance at the BOOST kickoff event and first tool implementation ranged from 3 months (first tool implemented prior to attending the kickoff) and 9 months (mean duration, 3.34.3 months) (Table 2).

The average rate of 30‐day rehospitalization among BOOST units was 14.7% in the preimplementation period and 12.7% during the postimplementation period (P=0.010) (Figure 1). Rehospitalization rates for matched control units were 14.0% in the preintervention period and 14.1% in the postintervention period (P=0.831). The mean absolute reduction in readmission rates over the 1‐year study period in BOOST units compared to control units was 2.0%, or a relative reduction of 13.6% (P=0.054 for signed rank test comparing differences in readmission rate reduction in BOOST units compared to site‐matched control units). Length of stay in BOOST and control units decreased an average of 0.5 days and 0.3 days, respectively. There was no difference in length of stay change between BOOST units and control units (P=0.966).

Figure 1

DISCUSSION

As hospitals strive to reduce their readmission rates to avoid Centers for Medicare and Medicaid Services penalties, Project BOOST may be a viable QI approach to achieve their goals. This initial evaluation of participation in Project BOOST by 11 hospitals of varying sizes across the United States showed an associated reduction in rehospitalization rates (absolute=2.0% and relative=13.6%, P=0.054). We did not find any significant change in length of stay among these hospitals implementing BOOST tools.

The tools provided to participating hospitals were developed from evidence found in peer‐reviewed literature established through experimental methods in well‐controlled academic settings. Further tool development was informed by recommendations of an advisory board consisting of expert representatives and advocates involved in the hospital discharge process: patients, caregivers, physicians, nurses, case managers, social workers, insurers, and regulatory and research agencies.[19] The toolkit components address multiple aspects of hospital discharge and follow‐up with the goal of improving health by optimizing the safety of care transitions. Our observation that readmission rates appeared to improve in a diverse hospital sample including nonacademic and community hospitals engaged in Project BOOST is reassuring that the benefits seen in existing research literature, developed in distinctly academic settings, can be replicated in diverse acute‐care settings.

The effect size observed in our study was modest but consistent with several studies identified in a recent review of trials measuring interventions to reduce rehospitalization, where 7 of 16 studies showing a significant improvement registered change in the 0% to 5% absolute range.[12] Impact of this project may have been tempered by the need to translate external QI content to the local setting. Additionally, in contrast to experimental studies that are limited in scope and timing and often scaled to a research budget, BOOST sites were encouraged to implement Project BOOST in the clinical setting even if no new funds were available to support the effort.[12]

The recruitment of a national sample of both academic and nonacademic hospital participants imposed several limitations on our study and analysis. We recognize that intervention units selected by hospitals may have had unmeasured unit and patient characteristics that facilitated successful change and contributed to the observed improvements. However, because external pressure to reduce readmission is present across all hospitals independent of the BOOST intervention, we felt site‐matched controls were essential to understanding effects attributable to the BOOST tools. Differences between units would be expected to be stable over the course of the study period, and comparison of outcome differences between 2 different time periods would be reasonable. Additionally, we could not collect data on readmissions to other hospitals. Theoretically, patients discharged from BOOST units might be more likely to have been rehospitalized elsewhere, but the fraction of rehospitalizations occurring at alternate facilities would also be expected to be similar on the matched control unit.

We report findings from a voluntary cohort willing and capable of designating a comparison clinical unit and contributing the requested data outcomes. Pilot sites that did not report outcomes were not analyzed, but comparison of hospital characteristics shows that participating hospitals were more likely to be large, urban, academic medical centers. Although barriers to data submission were not formally analyzed, reports from nonparticipating sites describe data submission limited by local implementation design (no geographic rollout or simultaneous rollout on all appropriate clinical units), site specific inability to generate unit level outcome statistics, and competing organizational priorities for data analyst time (electronic medical record deployment, alternative QI initiatives). The external validity of our results may be limited to organizations capable of analytics at the level of the individual clinical unit as well as those with sufficient QI resources to support reporting to a national database in the absence of a payer mandate. It is possible that additional financial support for on‐site data collection would have bolstered participation, making the example of participation rates we present potentially informative to organizations hoping to widely disseminate a QI agenda.

Nonetheless, the effectiveness demonstrated in the 11 sites that did participate is encouraging, and ongoing collaboration with subsequent BOOST cohorts has been designed to further facilitate data collection. Among the insights gained from this pilot experience, and incorporated into ongoing BOOST cohorts, is the importance of intensive mentor engagement to foster accountability among participant sites, assist with implementation troubleshooting, and offer expertise that is often particularly effective in gaining local support. We now encourage sites to have 2 mentor site visits to further these roles and more frequent conference calls. Further research to understand the marginal benefit of the mentored implementation approach is ongoing.

The limitations in data submission we experienced with the pilot cohort likely reflect resource constraints not uncommon at many hospitals. Increasing pressure placed on hospitals as a result of the Readmission Reduction Program within the Affordable Care Act as well as increasing interest from private and Medicaid payors to incorporate similar readmission‐based penalties provide encouragement for hospitals to enhance their data and analytic skills. National incentives for implementation of electronic health records (EHR) should also foster such capabilities, though we often saw EHRs as a barrier to QI, especially rapid cycle trials. Fortunately, hospitals are increasingly being afforded access to comprehensive claims databases to assist in tracking readmission rates to other facilities, and these data are becoming available in a more timely fashion. This more robust data collection, facilitated by private payors, state QI organizations, and state hospital associations, will support additional analytic methods such as multivariate regression models and interrupted time series designs to appreciate the experience of current BOOST participants.

Additional research is needed to understand the role of organizational context in the effectiveness of Project BOOST. Differences in rates of tool implementation and changes in clinical outcomes are likely dependent on local implementation context at the level of the healthcare organization and individual clinical unit.[20] Progress reports from site mentors and previously described experiences of QI implementation indicate that successful implementation of a multidimensional bundle of interventions may have reflected a higher level of institutional support, more robust team engagement in the work of reducing readmissions, increased clinical staff support for change, the presence of an effective project champion, or a key facilitating role of external mentorship.[21, 22] Ongoing data collection will continue to measure the sustainability of tool use and observed outcome changes to inform strategies to maintain gains associated with implementation. The role of mentored implementation in facilitating gains also requires further study.

Increasing attention to the problem of avoidable rehospitalization is driving hospitals, insurers, and policy makers to pursue QI efforts that favorably impact readmission rates. Our analysis of the BOOST intervention suggests that modest gains can be achieved following evidence‐based hospital process change facilitated by a mentored implementation model. However, realization of the goal of a 20% reduction in rehospitalization proposed by the Center for Medicare and Medicaid Services' Partnership for Patients initiative may be difficult to achieve on a national scale,[23] especially if efforts focus on just the hospital.

Disclosures

Project BOOST was funded by a grant from The John A. Hartford Foundation. Project BOOST is administered by the Society of Hospital Medicine (SHM). The development of the Project BOOST toolkit, recruitment of sites for this study, mentorship of the pilot cohort, project evaluation planning, and collection of pilot data were funded by a grant from The John A. Harford Foundation. Additional funding for continued data collection and analysis was funded by the SHM through funds from hospitals to participate in Project BOOST, specifically with funding support for Dr. Hansen. Dr. Williams has received funding to serve as Principal Investigator for Project BOOST. Since the time of initial cohort participation, approximately 125 additional hospitals have participated in the mentored implementation of Project BOOST. This participation was funded through a combination of site‐based tuition, third‐party payor support from private insurers, foundations, and federal funding through the Center for Medicare and Medicaid Innovation Partnership for Patients program. Drs. Greenwald, Hansen, and Williams are Project BOOST mentors for current Project BOOST sites and receive financial support through the SHM for this work. Dr. Howell has previously received funding as a Project BOOST mentor. Ms. Budnitz is the BOOST Project Director and is Chief Strategy and Development Officer for the HM. Dr. Maynard is the Senior Vice President of the SHM's Center for Hospital Innovation and Improvement.

In this issue of the Journal of Hospital Medicine, Hansen and colleagues provide a first, early look at the effectiveness of the BOOST intervention to reduce 30‐day readmissions among hospitalized patients.[1] BOOST[2] is 1 of a number of care transition improvement methodologies that have been applied to the problem of readmissions, each of which has evidence to support its effectiveness in its initial settings[3, 4] but has proven to be difficult to translate to other sites.[5, 6, 7]

BOOST stands in contrast with other, largely research protocol‐derived, programs in that it allows sites to tailor adoption of recommendations to local contexts and is therefore potentially more feasible to implement. Feasibility and practicality has led BOOST to be adopted in large national settings, even if it has had little evidence to support its effectiveness to date.

Given the nonstandardized and ad hoc nature of most multicenter collaboratives generally, and the flexibility of the BOOST model specifically, the BOOST authors are to be commended for undertaking any evaluation at all. Perhaps, not surprisingly, they encountered many of the problems associated with a multicenter studydropout of sites, problematic data, and limited evidence for adoption of the intervention at participating hospitals. Although these represent real‐world experiences of a quality‐improvement program, as a group they pose a number of problems that limit the study's robustness, and generate important caveats that readers should use to temper their interpretation of the authors' findings.

The first caveat relates to the substantial number of sites that either dropped out of BOOST or failed to submit data after enlisting in the collaborative. Although this may be common in quality improvement collaboratives, similar problems would not be permissible in a trial of a new drug or device. Dropout and selected ability to contribute data suggest that the ability to fully adopt BOOST may not be universal, and raises the possibility of bias, because the least successful sites may have had less interest in remaining engaged and submitting data.

The second caveat relates to how readmission rates were assessed. Because sites provided rates of readmissions at the unit level rather than the actual counts of admissions or readmissions, the authors were unable to conduct statistical analyses typically performed for these interventions, such as time series or difference‐in‐difference analyses. More importantly, one cannot discern whether their results are driven by a small absolute but large relative change in the number of readmissions at small sites. That is, large percentage changes of low statistical significance could have misleadingly affected the overall results. Conversely, we cannot identify large sites where a similar relative reduction could be statistically significant and more broadly interpreted as representing the real effectiveness of BOOST efforts.

The third caveat is in regard to the data describing the sites' performance. The effectiveness of BOOST in this analysis varied greatly among sites, with only 1 site showing a strong reduction in readmission rate, and nearly all others showing no statistical improvements. In fact, it appears that their overall results were almost entirely driven by the improvements at that 1 site.

Variable effectiveness of an intervention can be related to variable adoption or contextual factors (such as availability of personnel to implement the program). Although these authors have data on BOOST programmatic adoption, they do not have qualitative data on local barriers and facilitators to BOOST implementation, which at this stage of evaluation would be particularly valuable in understanding the results. Analyzing site‐level effectiveness is of growing relevance to multicenter quality improvement collaboratives,[8, 9] but this evaluation provides little insight into reasons for variable success across institutions.

Finally, their study design does not allow us to understand a number of key questions. How many patients were involved in the intervention? How many patients received all BOOST‐recommended interventions? Which of these interventions seemed most effective in which patients? To what degree did patient severity of illness, cognitive status, social supports, or access to primary care influence readmission risk? Such information would help frame cost‐effective deployment of BOOST or related tools.

In the end, it seems unlikely that this iteration of the BOOST program produced broad reductions in readmission rates. Having said this, the authors provide the necessary start down the road toward a fuller understanding of real‐world efforts to reduce readmissions. Stated alternately, the nuances and flaws of this study provide ample fodder for others working in the field. BOOST is in good stead with other care transition models that have not translated well from their initial research environment to real‐world practices. The question now is: Do any of these interventions actually work in clinical practice settings, and will we ever know? Even more fundamentally, how important and meaningful are these hospital‐based care transition interventions? Where is the engagement with primary care? Where are the primary care outcomes? Does BOOST truly impact outcomes other than readmission?[10]

Doing high‐quality research in the context of a rapidly evolving quality improvement program is hard. Doing it at more than 1 site is harder. BOOST's flexibility is both a great source of strength and a clear challenge to rigorous evaluation. However, when the costs of care transition programs are so high, and the potential consequences of high readmission rates are so great for patients and for hospitals, the need to address these issues with real data and better evidence is paramount. We look forward to the next phase of BOOST and to the growth and refinement of the evidence base for how to improve care coordination and transitions effectively.

In this issue of the Journal of Hospital Medicine, Hansen and colleagues provide a first, early look at the effectiveness of the BOOST intervention to reduce 30‐day readmissions among hospitalized patients.[1] BOOST[2] is 1 of a number of care transition improvement methodologies that have been applied to the problem of readmissions, each of which has evidence to support its effectiveness in its initial settings[3, 4] but has proven to be difficult to translate to other sites.[5, 6, 7]

BOOST stands in contrast with other, largely research protocol‐derived, programs in that it allows sites to tailor adoption of recommendations to local contexts and is therefore potentially more feasible to implement. Feasibility and practicality has led BOOST to be adopted in large national settings, even if it has had little evidence to support its effectiveness to date.

Given the nonstandardized and ad hoc nature of most multicenter collaboratives generally, and the flexibility of the BOOST model specifically, the BOOST authors are to be commended for undertaking any evaluation at all. Perhaps, not surprisingly, they encountered many of the problems associated with a multicenter studydropout of sites, problematic data, and limited evidence for adoption of the intervention at participating hospitals. Although these represent real‐world experiences of a quality‐improvement program, as a group they pose a number of problems that limit the study's robustness, and generate important caveats that readers should use to temper their interpretation of the authors' findings.

The first caveat relates to the substantial number of sites that either dropped out of BOOST or failed to submit data after enlisting in the collaborative. Although this may be common in quality improvement collaboratives, similar problems would not be permissible in a trial of a new drug or device. Dropout and selected ability to contribute data suggest that the ability to fully adopt BOOST may not be universal, and raises the possibility of bias, because the least successful sites may have had less interest in remaining engaged and submitting data.

The second caveat relates to how readmission rates were assessed. Because sites provided rates of readmissions at the unit level rather than the actual counts of admissions or readmissions, the authors were unable to conduct statistical analyses typically performed for these interventions, such as time series or difference‐in‐difference analyses. More importantly, one cannot discern whether their results are driven by a small absolute but large relative change in the number of readmissions at small sites. That is, large percentage changes of low statistical significance could have misleadingly affected the overall results. Conversely, we cannot identify large sites where a similar relative reduction could be statistically significant and more broadly interpreted as representing the real effectiveness of BOOST efforts.

The third caveat is in regard to the data describing the sites' performance. The effectiveness of BOOST in this analysis varied greatly among sites, with only 1 site showing a strong reduction in readmission rate, and nearly all others showing no statistical improvements. In fact, it appears that their overall results were almost entirely driven by the improvements at that 1 site.

Variable effectiveness of an intervention can be related to variable adoption or contextual factors (such as availability of personnel to implement the program). Although these authors have data on BOOST programmatic adoption, they do not have qualitative data on local barriers and facilitators to BOOST implementation, which at this stage of evaluation would be particularly valuable in understanding the results. Analyzing site‐level effectiveness is of growing relevance to multicenter quality improvement collaboratives,[8, 9] but this evaluation provides little insight into reasons for variable success across institutions.

Finally, their study design does not allow us to understand a number of key questions. How many patients were involved in the intervention? How many patients received all BOOST‐recommended interventions? Which of these interventions seemed most effective in which patients? To what degree did patient severity of illness, cognitive status, social supports, or access to primary care influence readmission risk? Such information would help frame cost‐effective deployment of BOOST or related tools.

In the end, it seems unlikely that this iteration of the BOOST program produced broad reductions in readmission rates. Having said this, the authors provide the necessary start down the road toward a fuller understanding of real‐world efforts to reduce readmissions. Stated alternately, the nuances and flaws of this study provide ample fodder for others working in the field. BOOST is in good stead with other care transition models that have not translated well from their initial research environment to real‐world practices. The question now is: Do any of these interventions actually work in clinical practice settings, and will we ever know? Even more fundamentally, how important and meaningful are these hospital‐based care transition interventions? Where is the engagement with primary care? Where are the primary care outcomes? Does BOOST truly impact outcomes other than readmission?[10]

Doing high‐quality research in the context of a rapidly evolving quality improvement program is hard. Doing it at more than 1 site is harder. BOOST's flexibility is both a great source of strength and a clear challenge to rigorous evaluation. However, when the costs of care transition programs are so high, and the potential consequences of high readmission rates are so great for patients and for hospitals, the need to address these issues with real data and better evidence is paramount. We look forward to the next phase of BOOST and to the growth and refinement of the evidence base for how to improve care coordination and transitions effectively.

In this issue of the Journal of Hospital Medicine, Hansen and colleagues provide a first, early look at the effectiveness of the BOOST intervention to reduce 30‐day readmissions among hospitalized patients.[1] BOOST[2] is 1 of a number of care transition improvement methodologies that have been applied to the problem of readmissions, each of which has evidence to support its effectiveness in its initial settings[3, 4] but has proven to be difficult to translate to other sites.[5, 6, 7]

BOOST stands in contrast with other, largely research protocol‐derived, programs in that it allows sites to tailor adoption of recommendations to local contexts and is therefore potentially more feasible to implement. Feasibility and practicality has led BOOST to be adopted in large national settings, even if it has had little evidence to support its effectiveness to date.

Given the nonstandardized and ad hoc nature of most multicenter collaboratives generally, and the flexibility of the BOOST model specifically, the BOOST authors are to be commended for undertaking any evaluation at all. Perhaps, not surprisingly, they encountered many of the problems associated with a multicenter studydropout of sites, problematic data, and limited evidence for adoption of the intervention at participating hospitals. Although these represent real‐world experiences of a quality‐improvement program, as a group they pose a number of problems that limit the study's robustness, and generate important caveats that readers should use to temper their interpretation of the authors' findings.

The first caveat relates to the substantial number of sites that either dropped out of BOOST or failed to submit data after enlisting in the collaborative. Although this may be common in quality improvement collaboratives, similar problems would not be permissible in a trial of a new drug or device. Dropout and selected ability to contribute data suggest that the ability to fully adopt BOOST may not be universal, and raises the possibility of bias, because the least successful sites may have had less interest in remaining engaged and submitting data.

The second caveat relates to how readmission rates were assessed. Because sites provided rates of readmissions at the unit level rather than the actual counts of admissions or readmissions, the authors were unable to conduct statistical analyses typically performed for these interventions, such as time series or difference‐in‐difference analyses. More importantly, one cannot discern whether their results are driven by a small absolute but large relative change in the number of readmissions at small sites. That is, large percentage changes of low statistical significance could have misleadingly affected the overall results. Conversely, we cannot identify large sites where a similar relative reduction could be statistically significant and more broadly interpreted as representing the real effectiveness of BOOST efforts.

The third caveat is in regard to the data describing the sites' performance. The effectiveness of BOOST in this analysis varied greatly among sites, with only 1 site showing a strong reduction in readmission rate, and nearly all others showing no statistical improvements. In fact, it appears that their overall results were almost entirely driven by the improvements at that 1 site.

Variable effectiveness of an intervention can be related to variable adoption or contextual factors (such as availability of personnel to implement the program). Although these authors have data on BOOST programmatic adoption, they do not have qualitative data on local barriers and facilitators to BOOST implementation, which at this stage of evaluation would be particularly valuable in understanding the results. Analyzing site‐level effectiveness is of growing relevance to multicenter quality improvement collaboratives,[8, 9] but this evaluation provides little insight into reasons for variable success across institutions.

Finally, their study design does not allow us to understand a number of key questions. How many patients were involved in the intervention? How many patients received all BOOST‐recommended interventions? Which of these interventions seemed most effective in which patients? To what degree did patient severity of illness, cognitive status, social supports, or access to primary care influence readmission risk? Such information would help frame cost‐effective deployment of BOOST or related tools.

In the end, it seems unlikely that this iteration of the BOOST program produced broad reductions in readmission rates. Having said this, the authors provide the necessary start down the road toward a fuller understanding of real‐world efforts to reduce readmissions. Stated alternately, the nuances and flaws of this study provide ample fodder for others working in the field. BOOST is in good stead with other care transition models that have not translated well from their initial research environment to real‐world practices. The question now is: Do any of these interventions actually work in clinical practice settings, and will we ever know? Even more fundamentally, how important and meaningful are these hospital‐based care transition interventions? Where is the engagement with primary care? Where are the primary care outcomes? Does BOOST truly impact outcomes other than readmission?[10]

Doing high‐quality research in the context of a rapidly evolving quality improvement program is hard. Doing it at more than 1 site is harder. BOOST's flexibility is both a great source of strength and a clear challenge to rigorous evaluation. However, when the costs of care transition programs are so high, and the potential consequences of high readmission rates are so great for patients and for hospitals, the need to address these issues with real data and better evidence is paramount. We look forward to the next phase of BOOST and to the growth and refinement of the evidence base for how to improve care coordination and transitions effectively.

A 39‐year‐old woman presented to the emergency department (ED) with fever and headache. One to two weeks prior to presentation, she developed nightly fevers that gradually increased to as high as 39.4C. She subsequently developed generalized throbbing headaches, malaise, and diffuse body pain. The headache gradually worsened. The day prior to presentation, she developed photophobia, nausea, and vomiting. She also reported right scalp pain while combing her hair, difficulty emptying her bladder, and left buttock pain radiating down the leg. She denied rash, joint pain, visual changes, dysarthria, cough, chest pain, abdominal pain, or diarrhea.

Fever and headache can be explained by meningitis, encephalitis, or brain abscess. The combination is seen far more frequently, however, in patients with common systemic infections such as influenza. For either bacterial meningitis or influenza, a 2‐week course is prolonged and atypical. The progressive nature of the symptoms and photophobia suggest a chronic meningitis, and the development of nausea and vomiting, although nonspecific, is also consistent with elevated intracranial pressure. In a young woman, subacute fever and aches should prompt consideration of an autoimmune disorder such as systemic lupus erythematosus (SLE), although early central nervous system (CNS) involvement is atypical. Migraine headaches are characterized by light sensitivity, nausea, and vomiting and can be precipitated by a viral syndrome, but in this case, the headaches were present at the outset, and 2 weeks is too long for a migraine attack.

Pain while combing hair is not characteristic of the aforementioned syndromes. The scalp should be examined to confirm that there are no skin lesions associated with herpes zoster and no arterial prominence associated with temporal arteritis. She is young for the latter, which would otherwise be a suitable explanation for fever, headache, scalp tenderness, and visual complaints (usually impairment not photophobia).

Incomplete bladder emptying and left buttock pain suggest that there might be a concomitant lumbosacral myelopathy or radiculopathy. Some nonbacterial causes of meningitis such as cytomegalovirus (CMV), syphilis, and cancer simultaneously involve the CNS and peripheral nerve roots. It is also possible that the scalp tenderness associated with combing reflects a cervical sensory radiculopathy.

She had presented to the ED 2 and 4 days before the current (third) ED visit. Both times her main complaint was left buttock pain and left leg paresthesias. Although she had no skin lesions, she was diagnosed with prodromal herpes zoster in the S2 dermatomal distribution and was prescribed valacyclovir (to be started should eruptions develop, which never occurred).

She reported intermittent self‐limited fevers at 3‐ to 4‐week intervals during the prior 6 months; two fever episodes were accompanied by an influenza‐like illness, and one was associated with gastrointestinal symptoms. Her last fever prior to this evaluation was 6 weeks earlier when she was treated with azithromycin for suspected pneumonia at an outside facility.

Her past medical history included hypothyroidism, gastroesophageal reflux disease, diverticulitis, and gluten intolerance. Her medications included porcine (natural) thyroid, fish oil, ibuprofen, and acetaminophen. She lived in Michigan and traveled to the northeast United States (Maine, Cape Cod, New Hampshire, Connecticut, and Vermont) 7 months prior to this evaluation. She was married and had no pets at home. She denied any tobacco, alcohol, or illicit drug use.

Her illness now appears to be chronic, associated with fever, and multisystem (potentially involving the pulmonary and gastrointestinal tract). None of her medical problems would predispose her to subacute meningitis, myelopathy, or radiculopathy. Hypothyroidism raises the possibility of a concomitant autoimmune disorder which causes meningitis, such as SLE or Behet's disease. Sarcoidosis can cause chronic meningitis and neuropathy with concomitant lung and gastrointestinal involvement and rarely fever.

Residency in the upper Midwest increases exposure to chronic infections that rarely cause subacute meningitis such as histoplasmosis, blastomycosis, or human granulocytic anaplasmosis. Travel to the northeast United States 1 month before the onset of her symptoms raises the possibility of other endemic infections like Lyme disease, babesiosis, and tularemia, which may account for her recurrent fevers. Of these, Lyme is most likely to present as chronic meningitis with cranial neuropathy and radiculoneuropathy.

Although the diagnosis of pneumonia was made late in her 6‐month illness, its etiology and treatment may be relevant. If the recent pneumonia was viral, a subsequent viral meningitis may be manifesting now or may have triggered an autoimmune process, such as acute disseminated encephalomyelitis. Bacterial pneumonia is a common precursor to bacterial meningitis, and treatment with azithromycin for the pneumonia may have delayed the meningitis onset or muted its course; this should be taken into account when interpreting cerebrospinal fluid (CSF) culture results.

On physical examination, her temperature was 39.1C, blood pressure was 135/91 mm Hg, with pulse of 87 beats per minute, respiratory rate of 16 breaths per minute, and oxygenation saturation of 97% on room air. She appeared in distress and was covering her eyes. She was alert and oriented. She had photophobia and mild nuchal rigidity. Pupils were equal and reactive to light, but she could not tolerate the eye exam for papilledema. Lung, heart, and abdominal exam were normal. No cranial nerve abnormalities were noted, and muscle strength was 5/5 in all 4 extremities. She had decreased sensation to light touch with allodynia throughout her lower extremities in addition to the lateral portion of the right scalp, which was also tender to palpation. Deep tendon reflexes were 2+ and symmetric in her bilateral upper and lower extremities. She did not have joint swelling, edema, lymphadenopathy, or a rash.

Her fever, headache, nuchal rigidity and photophobia collectively suggest meningitis, which requires evaluation by a lumbar puncture. There is no rash that supports herpes zoster or SLE. She does not have signs of myelopathy that would explain the urinary complaints, but lower motor neuron involvement has not been excluded. The sensory abnormalities in the scalp and leg are consistent with a polyneuroradiculopathy. Anterior lateral scalp tenderness may signal trigeminal nerve involvement, whereas posterior scalp tenderness would localize to the upper cervical cord nerve roots. The contralateral distribution of the scalp and leg sensory deficits suggests a multifocal peripheral nervous system process rather than a single CNS lesion.

Initial laboratory data showed serum white blood cell count (WBC) of 12,000/mm³ (79% polymorphonuclear leukocytes). Hemoglobin was 14.2 g/dL, and platelets were 251,000/mm³. Electrolytes, renal function, and liver function were normal. Thyroid‐stimulating hormone, erythrocyte sedimentation rate, and C‐reactive protein were normal. Urinalysis was negative. Chest x‐ray was normal. Noncontrast head computed tomography (CT) was normal. The patient was unable to void; 500 mL of urine returned when catheterization was performed.

CSF WBC count was 1,280/mm³ (39% neutrophils and 49% lymphocytes). CSF total protein was 175 mg/dL, and glucose was 48 mg/dL; serum glucose was 104 mg/dL. Opening pressure was not recorded. Gram stain was negative. Ceftriaxone, vancomycin, ampicillin, and acyclovir were administered for presumed bacterial or viral meningitis. Magnetic resonance imaging (MRI) of the brain and spine showed diffuse leptomeningeal enhancement (Figure 1).

Figure 1

The urinary retention in the absence of myelopathic findings on exam or MRI suggests a sacral polyradiculoneuropathy. Diffuse leptomeningeal enhancement is consistent with many, if not all, causes of meningitis. The high WBC count, elevated protein, and low glucosecollectively signaling active inflammation in the CNSare highly compatible with bacterial meningitis, although the lymphocytic predominance and other clinical data point to nonbacterial etiologies. The negative Gram stain further lowers the probability of bacterial meningitis, but it has limited sensitivity, may be affected by recent antibiotics, and is typically negative with Listeria. Enterovirus, acute human immunodeficiency virus (HIV), and herpes viruses (eg, CMV or herpes simplex virus [HSV]) are important considerations, with the latter 2 causing associated polyneuroradiculopathy. Patients with genital HSV (not detected here) can have a concomitant sacral radiculitis leading to urinary retention.

Fungal and mycobacterial meningitis is a possibility (especially with the high protein), but the patient does not have the typical multisystem disease or immunosuppression that frequently accompanies those conditions when CNS disease is present. Autoimmune conditions like SLE, Behet's disease, and sarcoidosis remain important conditions, especially with the polyneuroradiculopathy or mononeuritis multiplex, which may reflect multifocal nerve infarction or invasion. Similarly, lymphomatous or carcinomatous meningitis should be considered, although an isolated manifestation in the CNS is unusual. Based on the multifocal neurologic deficits, I favor a viral, spirochete, or malignant etiology of her meningoencephalitis.

Despite ongoing broad spectrum antibiotics and supportive care, she became confused on hospital day 3 and developed anomia, agitation, and worsening headache. A repeat CT of the brain did not show any new abnormalities, but repeat lumbar puncture demonstrated elevated intracranial pressure (opening pressure of 47 cm water) with 427 WBC/mm³. Blood and CSF cultures remained negative.

Detailed questioning of the family revealed that she had been horseback riding 3 weeks prior to admission; there were no other livestock where she rode horses. In addition, the family reported that she and other family members routinely drank raw milk from a cow share program.

HIV antibody test was negative. Herpes simplex, varicella zoster, enteroviruses, and adenovirus CSF polymerase chain reaction (PCR) were negative. Cytomegalovirus and Epstein‐Barr virus PCR were negative in serum and CSF. Arbovirus, lymphocytic choriomeningitis, Coccidioides, Blastomyces, Histoplasma, Brucella, and Lyme serologies were negative. Cryptococcus neoformans antigen was negative in CSF. Serum QuantiFERON‐TB test was negative. Blood and CSF acid‐fast bacilli smears (and eventually mycobacterial cultures) were also negative. Her CSF flow cytometry and cytology were negative for lymphoma.

Unpasteurized milk conveys multiple infectious risks. Listeriosis is a food‐borne illness that can cause meningoencephalitis, but peripheral neuropathies are not characteristic. Brucellosis is usually characterized by severe bone pain, pancytopenia, and hepatosplenomegaly, which are absent. Infection with Mycobacterium bovis mimics Mycobacterium tuberculosis and can cause multisystem disease, typically involving the lung. Campylobacter infection is characterized by gastroenteritis, which has not been prominent.

Rhodococcus equi is a horse‐related pathogen which leads to pulmonary infections in immunocompromised hosts but not meningitis. Rather than focusing on horse exposure alone, however, it may be useful to consider her at risk for vector‐borne pathogens based on her time outdoors, such as Lyme disease (which can cause radiculopathy and encephalopathy), West Nile virus (although motor weakness rather than sensory symptoms is typical), or eastern equine encephalitis.

The absence of weight loss, cytopenias, lymphadenopathy, and organomegaly with the negative CSF cytology and flow cytometry makes lymphomatous meningitis unlikely. The case for an autoimmune disorder is not strong in the absence of joint pains, rash, or autoimmune serologies. In a young woman with unexplained encephalitis, antibodies to the N‐methyl‐D‐aspartate receptor should be assayed.

Although the CSF leukocytosis is declining, the elevated pressure and clinical deterioration signal that the disease process is not controlled. At this point I am uncertain as to the cause of her progressive meningoencephalitis with polyneuroradiculopathy. The latter feature makes me favor a viral or spirochete etiology.

On hospital day 4, Coxiella burnetii serologies were reported as positive (phase II immunoglobulin [Ig] G 1:256; phase II IgM <1:16; phase I IgG <1:16; phase I IgM <1:16) suggesting acute Q fever. Antibiotics were changed to intravenous doxycycline and ciprofloxacin. Her increased intracranial pressure was managed with serial lumbar punctures. The patient was discharged after 12 days of hospitalization taking oral doxycycline and ciprofloxacin. Her symptoms resolved over 10 weeks. No vegetations were seen on transesophageal echocardiogram. She had no evidence of chronic Q fever on repeat serologies.

I was not aware that Q fever causes meningitis or meningoencephalitis. However, I should have considered it in light of her indirect exposure to cows. It is possible that her pneumonia 6 weeks earlier represented acute Q fever, as pneumonia and hepatitis are among the most typical acute manifestations of this infection.

COMMENTARY

Hospitalists are commonly confronted by the combination of fever, headache, and confusion and are familiar with the diagnostic and therapeutic dilemmas related to prompt discrimination between CNS and non‐CNS processes, particularly infections. At the time of this patient's final ED presentation, her illness unambiguously localized to the CNS. As common and emergent conditions such as acute bacterial meningitis were excluded, the greatest challenge was finding the clue that could direct investigations into less common causes of meningoencephalitis.

The Infectious Disease Society of America has developed clinical practice guidelines for the diagnosis and management of encephalitis which highlight the importance of epidemiology and risk factor assessment.[1] This approach requires the clinician to examine potential clues and to go beyond initial associationsfor instance, not simply linking horseback riding to horse‐associated pathogens, but interpreting horseback riding as a proxy for outdoor exposure, which places her at risk for contact with mosquitos, which transmit West Nile virus or eastern equine encephalitis. Similarly, ingestion of raw milk, which is typically linked to Listeria monocytogenes, Brucella, and other pathogens prompted the infectious disease consultant to think more broadly and include livestock (cow)‐associated pathogens including C. burnetii.

Although involvement of the CNS is common in chronic Q fever endocarditis due to septic embolism, neurologic involvement in acute Q fever varies in prevalence (range of 1.7%22%).[2, 3, 4] The 3 major neurological syndromes of acute Q fever are (1) meningoencephalitis or encephalitis, (2) lymphocytic meningitis, and (3) peripheral neuropathy (myelitis, polyradiculoneuritis, or peripheral neuritis). CSF analysis usually shows mild pleocytosis with a predominance of lymphocytic cells; CSF protein elevation is variable, and glucose is usually normal. Neuroradiologic examination is usually normal, and there are no pathognomonic imaging abnormalities for Q fever meningoencephalitis.[2, 3] The mechanism by which C. burnetii causes neurologic injury and dysfunction is unknown.

The diagnosis of Q fever is usually established by serologic testing. In acute Q fever, antibodies to phase II antigen are higher than the phase I antibody titer. Phase II IgM antibodies are the first to appear, but then decline on average after week 8, often reaching undetectable levels 10 to 12 weeks after disease onset.[5] If this patient's pneumonia 6 weeks prior to this presentation was acute Q fever pneumonia, her IgM titers may have been declining by the time her neurologic illness developed. A false negative test result is also possible; immunofluorescence assays are more specific than sensitive in acute Q fever.[5]

Evaluating this case in isolation may raise some doubt as to the accuracy of the diagnosis as she did not have a 4‐fold rise in the phase II IgG titer and did not have a detectable phase II IgM. However, she was part of a cluster of individuals who regularly consumed raw milk from the same dairy and had evidence of C. burnetii infection. This group included her spouse, who had a robust serologic evidence of C. burnetii, characterized by a >4‐fold rise in phase II IgM and IgG titers.[6]

C. burnetii is found primarily in cattle, sheep, and goats and is shed in large quantities by infected periparturient animals in their urine, feces, and milk.[7] Inhalation of contaminated aerosols is the principal route of transmission.[7, 8] Acute Q fever is underdiagnosed because the majority of acute infections are asymptomatic (60%) or present as a nonspecific flu‐like illness.[7] This case represents a rare manifestation of a rare infection acquired through a rare route of transmission, but highlights the importance of epidemiology and risk factor assessment when clinicians are faced with a diagnostic challenge.

TEACHING POINTS

• Exploration of epidemiology and exposure history is central to diagnosing meningoencephalitis with negative bacterial cultures and undetectable HSV PCR, although the etiology of meningoencephalitis can elude identification even after exhaustive investigation.
• Inhalation of contaminated aerosols is the principal route of transmission for C. burnetii, but it can also be transmitted via infected unpasteurized milk.[7, 9]
• Acute presentations of Q fever, which may warrant admission, include pneumonia, hepatitis, or meningoencephalitis.
• Q fever is diagnosed by serologic testing, and doxycycline is the antibiotic of choice.

Disclosures

This case was presented at the 2012 Annual Meeting of the Society of Hospital Medicine. It was subsequently reported in the epidemiologic report of the outbreak.[6] The authors report no conflicts of interest.

A 39‐year‐old woman presented to the emergency department (ED) with fever and headache. One to two weeks prior to presentation, she developed nightly fevers that gradually increased to as high as 39.4C. She subsequently developed generalized throbbing headaches, malaise, and diffuse body pain. The headache gradually worsened. The day prior to presentation, she developed photophobia, nausea, and vomiting. She also reported right scalp pain while combing her hair, difficulty emptying her bladder, and left buttock pain radiating down the leg. She denied rash, joint pain, visual changes, dysarthria, cough, chest pain, abdominal pain, or diarrhea.

Fever and headache can be explained by meningitis, encephalitis, or brain abscess. The combination is seen far more frequently, however, in patients with common systemic infections such as influenza. For either bacterial meningitis or influenza, a 2‐week course is prolonged and atypical. The progressive nature of the symptoms and photophobia suggest a chronic meningitis, and the development of nausea and vomiting, although nonspecific, is also consistent with elevated intracranial pressure. In a young woman, subacute fever and aches should prompt consideration of an autoimmune disorder such as systemic lupus erythematosus (SLE), although early central nervous system (CNS) involvement is atypical. Migraine headaches are characterized by light sensitivity, nausea, and vomiting and can be precipitated by a viral syndrome, but in this case, the headaches were present at the outset, and 2 weeks is too long for a migraine attack.

Pain while combing hair is not characteristic of the aforementioned syndromes. The scalp should be examined to confirm that there are no skin lesions associated with herpes zoster and no arterial prominence associated with temporal arteritis. She is young for the latter, which would otherwise be a suitable explanation for fever, headache, scalp tenderness, and visual complaints (usually impairment not photophobia).

Incomplete bladder emptying and left buttock pain suggest that there might be a concomitant lumbosacral myelopathy or radiculopathy. Some nonbacterial causes of meningitis such as cytomegalovirus (CMV), syphilis, and cancer simultaneously involve the CNS and peripheral nerve roots. It is also possible that the scalp tenderness associated with combing reflects a cervical sensory radiculopathy.

She had presented to the ED 2 and 4 days before the current (third) ED visit. Both times her main complaint was left buttock pain and left leg paresthesias. Although she had no skin lesions, she was diagnosed with prodromal herpes zoster in the S2 dermatomal distribution and was prescribed valacyclovir (to be started should eruptions develop, which never occurred).

She reported intermittent self‐limited fevers at 3‐ to 4‐week intervals during the prior 6 months; two fever episodes were accompanied by an influenza‐like illness, and one was associated with gastrointestinal symptoms. Her last fever prior to this evaluation was 6 weeks earlier when she was treated with azithromycin for suspected pneumonia at an outside facility.

Her past medical history included hypothyroidism, gastroesophageal reflux disease, diverticulitis, and gluten intolerance. Her medications included porcine (natural) thyroid, fish oil, ibuprofen, and acetaminophen. She lived in Michigan and traveled to the northeast United States (Maine, Cape Cod, New Hampshire, Connecticut, and Vermont) 7 months prior to this evaluation. She was married and had no pets at home. She denied any tobacco, alcohol, or illicit drug use.

Her illness now appears to be chronic, associated with fever, and multisystem (potentially involving the pulmonary and gastrointestinal tract). None of her medical problems would predispose her to subacute meningitis, myelopathy, or radiculopathy. Hypothyroidism raises the possibility of a concomitant autoimmune disorder which causes meningitis, such as SLE or Behet's disease. Sarcoidosis can cause chronic meningitis and neuropathy with concomitant lung and gastrointestinal involvement and rarely fever.

Residency in the upper Midwest increases exposure to chronic infections that rarely cause subacute meningitis such as histoplasmosis, blastomycosis, or human granulocytic anaplasmosis. Travel to the northeast United States 1 month before the onset of her symptoms raises the possibility of other endemic infections like Lyme disease, babesiosis, and tularemia, which may account for her recurrent fevers. Of these, Lyme is most likely to present as chronic meningitis with cranial neuropathy and radiculoneuropathy.

Although the diagnosis of pneumonia was made late in her 6‐month illness, its etiology and treatment may be relevant. If the recent pneumonia was viral, a subsequent viral meningitis may be manifesting now or may have triggered an autoimmune process, such as acute disseminated encephalomyelitis. Bacterial pneumonia is a common precursor to bacterial meningitis, and treatment with azithromycin for the pneumonia may have delayed the meningitis onset or muted its course; this should be taken into account when interpreting cerebrospinal fluid (CSF) culture results.

On physical examination, her temperature was 39.1C, blood pressure was 135/91 mm Hg, with pulse of 87 beats per minute, respiratory rate of 16 breaths per minute, and oxygenation saturation of 97% on room air. She appeared in distress and was covering her eyes. She was alert and oriented. She had photophobia and mild nuchal rigidity. Pupils were equal and reactive to light, but she could not tolerate the eye exam for papilledema. Lung, heart, and abdominal exam were normal. No cranial nerve abnormalities were noted, and muscle strength was 5/5 in all 4 extremities. She had decreased sensation to light touch with allodynia throughout her lower extremities in addition to the lateral portion of the right scalp, which was also tender to palpation. Deep tendon reflexes were 2+ and symmetric in her bilateral upper and lower extremities. She did not have joint swelling, edema, lymphadenopathy, or a rash.

Her fever, headache, nuchal rigidity and photophobia collectively suggest meningitis, which requires evaluation by a lumbar puncture. There is no rash that supports herpes zoster or SLE. She does not have signs of myelopathy that would explain the urinary complaints, but lower motor neuron involvement has not been excluded. The sensory abnormalities in the scalp and leg are consistent with a polyneuroradiculopathy. Anterior lateral scalp tenderness may signal trigeminal nerve involvement, whereas posterior scalp tenderness would localize to the upper cervical cord nerve roots. The contralateral distribution of the scalp and leg sensory deficits suggests a multifocal peripheral nervous system process rather than a single CNS lesion.

Initial laboratory data showed serum white blood cell count (WBC) of 12,000/mm³ (79% polymorphonuclear leukocytes). Hemoglobin was 14.2 g/dL, and platelets were 251,000/mm³. Electrolytes, renal function, and liver function were normal. Thyroid‐stimulating hormone, erythrocyte sedimentation rate, and C‐reactive protein were normal. Urinalysis was negative. Chest x‐ray was normal. Noncontrast head computed tomography (CT) was normal. The patient was unable to void; 500 mL of urine returned when catheterization was performed.

CSF WBC count was 1,280/mm³ (39% neutrophils and 49% lymphocytes). CSF total protein was 175 mg/dL, and glucose was 48 mg/dL; serum glucose was 104 mg/dL. Opening pressure was not recorded. Gram stain was negative. Ceftriaxone, vancomycin, ampicillin, and acyclovir were administered for presumed bacterial or viral meningitis. Magnetic resonance imaging (MRI) of the brain and spine showed diffuse leptomeningeal enhancement (Figure 1).

Figure 1

The urinary retention in the absence of myelopathic findings on exam or MRI suggests a sacral polyradiculoneuropathy. Diffuse leptomeningeal enhancement is consistent with many, if not all, causes of meningitis. The high WBC count, elevated protein, and low glucosecollectively signaling active inflammation in the CNSare highly compatible with bacterial meningitis, although the lymphocytic predominance and other clinical data point to nonbacterial etiologies. The negative Gram stain further lowers the probability of bacterial meningitis, but it has limited sensitivity, may be affected by recent antibiotics, and is typically negative with Listeria. Enterovirus, acute human immunodeficiency virus (HIV), and herpes viruses (eg, CMV or herpes simplex virus [HSV]) are important considerations, with the latter 2 causing associated polyneuroradiculopathy. Patients with genital HSV (not detected here) can have a concomitant sacral radiculitis leading to urinary retention.

Fungal and mycobacterial meningitis is a possibility (especially with the high protein), but the patient does not have the typical multisystem disease or immunosuppression that frequently accompanies those conditions when CNS disease is present. Autoimmune conditions like SLE, Behet's disease, and sarcoidosis remain important conditions, especially with the polyneuroradiculopathy or mononeuritis multiplex, which may reflect multifocal nerve infarction or invasion. Similarly, lymphomatous or carcinomatous meningitis should be considered, although an isolated manifestation in the CNS is unusual. Based on the multifocal neurologic deficits, I favor a viral, spirochete, or malignant etiology of her meningoencephalitis.

Despite ongoing broad spectrum antibiotics and supportive care, she became confused on hospital day 3 and developed anomia, agitation, and worsening headache. A repeat CT of the brain did not show any new abnormalities, but repeat lumbar puncture demonstrated elevated intracranial pressure (opening pressure of 47 cm water) with 427 WBC/mm³. Blood and CSF cultures remained negative.

Detailed questioning of the family revealed that she had been horseback riding 3 weeks prior to admission; there were no other livestock where she rode horses. In addition, the family reported that she and other family members routinely drank raw milk from a cow share program.

HIV antibody test was negative. Herpes simplex, varicella zoster, enteroviruses, and adenovirus CSF polymerase chain reaction (PCR) were negative. Cytomegalovirus and Epstein‐Barr virus PCR were negative in serum and CSF. Arbovirus, lymphocytic choriomeningitis, Coccidioides, Blastomyces, Histoplasma, Brucella, and Lyme serologies were negative. Cryptococcus neoformans antigen was negative in CSF. Serum QuantiFERON‐TB test was negative. Blood and CSF acid‐fast bacilli smears (and eventually mycobacterial cultures) were also negative. Her CSF flow cytometry and cytology were negative for lymphoma.

Unpasteurized milk conveys multiple infectious risks. Listeriosis is a food‐borne illness that can cause meningoencephalitis, but peripheral neuropathies are not characteristic. Brucellosis is usually characterized by severe bone pain, pancytopenia, and hepatosplenomegaly, which are absent. Infection with Mycobacterium bovis mimics Mycobacterium tuberculosis and can cause multisystem disease, typically involving the lung. Campylobacter infection is characterized by gastroenteritis, which has not been prominent.

Rhodococcus equi is a horse‐related pathogen which leads to pulmonary infections in immunocompromised hosts but not meningitis. Rather than focusing on horse exposure alone, however, it may be useful to consider her at risk for vector‐borne pathogens based on her time outdoors, such as Lyme disease (which can cause radiculopathy and encephalopathy), West Nile virus (although motor weakness rather than sensory symptoms is typical), or eastern equine encephalitis.

The absence of weight loss, cytopenias, lymphadenopathy, and organomegaly with the negative CSF cytology and flow cytometry makes lymphomatous meningitis unlikely. The case for an autoimmune disorder is not strong in the absence of joint pains, rash, or autoimmune serologies. In a young woman with unexplained encephalitis, antibodies to the N‐methyl‐D‐aspartate receptor should be assayed.

Although the CSF leukocytosis is declining, the elevated pressure and clinical deterioration signal that the disease process is not controlled. At this point I am uncertain as to the cause of her progressive meningoencephalitis with polyneuroradiculopathy. The latter feature makes me favor a viral or spirochete etiology.

On hospital day 4, Coxiella burnetii serologies were reported as positive (phase II immunoglobulin [Ig] G 1:256; phase II IgM <1:16; phase I IgG <1:16; phase I IgM <1:16) suggesting acute Q fever. Antibiotics were changed to intravenous doxycycline and ciprofloxacin. Her increased intracranial pressure was managed with serial lumbar punctures. The patient was discharged after 12 days of hospitalization taking oral doxycycline and ciprofloxacin. Her symptoms resolved over 10 weeks. No vegetations were seen on transesophageal echocardiogram. She had no evidence of chronic Q fever on repeat serologies.

I was not aware that Q fever causes meningitis or meningoencephalitis. However, I should have considered it in light of her indirect exposure to cows. It is possible that her pneumonia 6 weeks earlier represented acute Q fever, as pneumonia and hepatitis are among the most typical acute manifestations of this infection.

COMMENTARY

Hospitalists are commonly confronted by the combination of fever, headache, and confusion and are familiar with the diagnostic and therapeutic dilemmas related to prompt discrimination between CNS and non‐CNS processes, particularly infections. At the time of this patient's final ED presentation, her illness unambiguously localized to the CNS. As common and emergent conditions such as acute bacterial meningitis were excluded, the greatest challenge was finding the clue that could direct investigations into less common causes of meningoencephalitis.

The Infectious Disease Society of America has developed clinical practice guidelines for the diagnosis and management of encephalitis which highlight the importance of epidemiology and risk factor assessment.[1] This approach requires the clinician to examine potential clues and to go beyond initial associationsfor instance, not simply linking horseback riding to horse‐associated pathogens, but interpreting horseback riding as a proxy for outdoor exposure, which places her at risk for contact with mosquitos, which transmit West Nile virus or eastern equine encephalitis. Similarly, ingestion of raw milk, which is typically linked to Listeria monocytogenes, Brucella, and other pathogens prompted the infectious disease consultant to think more broadly and include livestock (cow)‐associated pathogens including C. burnetii.

Although involvement of the CNS is common in chronic Q fever endocarditis due to septic embolism, neurologic involvement in acute Q fever varies in prevalence (range of 1.7%22%).[2, 3, 4] The 3 major neurological syndromes of acute Q fever are (1) meningoencephalitis or encephalitis, (2) lymphocytic meningitis, and (3) peripheral neuropathy (myelitis, polyradiculoneuritis, or peripheral neuritis). CSF analysis usually shows mild pleocytosis with a predominance of lymphocytic cells; CSF protein elevation is variable, and glucose is usually normal. Neuroradiologic examination is usually normal, and there are no pathognomonic imaging abnormalities for Q fever meningoencephalitis.[2, 3] The mechanism by which C. burnetii causes neurologic injury and dysfunction is unknown.

The diagnosis of Q fever is usually established by serologic testing. In acute Q fever, antibodies to phase II antigen are higher than the phase I antibody titer. Phase II IgM antibodies are the first to appear, but then decline on average after week 8, often reaching undetectable levels 10 to 12 weeks after disease onset.[5] If this patient's pneumonia 6 weeks prior to this presentation was acute Q fever pneumonia, her IgM titers may have been declining by the time her neurologic illness developed. A false negative test result is also possible; immunofluorescence assays are more specific than sensitive in acute Q fever.[5]

Evaluating this case in isolation may raise some doubt as to the accuracy of the diagnosis as she did not have a 4‐fold rise in the phase II IgG titer and did not have a detectable phase II IgM. However, she was part of a cluster of individuals who regularly consumed raw milk from the same dairy and had evidence of C. burnetii infection. This group included her spouse, who had a robust serologic evidence of C. burnetii, characterized by a >4‐fold rise in phase II IgM and IgG titers.[6]

C. burnetii is found primarily in cattle, sheep, and goats and is shed in large quantities by infected periparturient animals in their urine, feces, and milk.[7] Inhalation of contaminated aerosols is the principal route of transmission.[7, 8] Acute Q fever is underdiagnosed because the majority of acute infections are asymptomatic (60%) or present as a nonspecific flu‐like illness.[7] This case represents a rare manifestation of a rare infection acquired through a rare route of transmission, but highlights the importance of epidemiology and risk factor assessment when clinicians are faced with a diagnostic challenge.

TEACHING POINTS

• Exploration of epidemiology and exposure history is central to diagnosing meningoencephalitis with negative bacterial cultures and undetectable HSV PCR, although the etiology of meningoencephalitis can elude identification even after exhaustive investigation.
• Inhalation of contaminated aerosols is the principal route of transmission for C. burnetii, but it can also be transmitted via infected unpasteurized milk.[7, 9]
• Acute presentations of Q fever, which may warrant admission, include pneumonia, hepatitis, or meningoencephalitis.
• Q fever is diagnosed by serologic testing, and doxycycline is the antibiotic of choice.

Disclosures

This case was presented at the 2012 Annual Meeting of the Society of Hospital Medicine. It was subsequently reported in the epidemiologic report of the outbreak.[6] The authors report no conflicts of interest.

A 39‐year‐old woman presented to the emergency department (ED) with fever and headache. One to two weeks prior to presentation, she developed nightly fevers that gradually increased to as high as 39.4C. She subsequently developed generalized throbbing headaches, malaise, and diffuse body pain. The headache gradually worsened. The day prior to presentation, she developed photophobia, nausea, and vomiting. She also reported right scalp pain while combing her hair, difficulty emptying her bladder, and left buttock pain radiating down the leg. She denied rash, joint pain, visual changes, dysarthria, cough, chest pain, abdominal pain, or diarrhea.

Fever and headache can be explained by meningitis, encephalitis, or brain abscess. The combination is seen far more frequently, however, in patients with common systemic infections such as influenza. For either bacterial meningitis or influenza, a 2‐week course is prolonged and atypical. The progressive nature of the symptoms and photophobia suggest a chronic meningitis, and the development of nausea and vomiting, although nonspecific, is also consistent with elevated intracranial pressure. In a young woman, subacute fever and aches should prompt consideration of an autoimmune disorder such as systemic lupus erythematosus (SLE), although early central nervous system (CNS) involvement is atypical. Migraine headaches are characterized by light sensitivity, nausea, and vomiting and can be precipitated by a viral syndrome, but in this case, the headaches were present at the outset, and 2 weeks is too long for a migraine attack.

Pain while combing hair is not characteristic of the aforementioned syndromes. The scalp should be examined to confirm that there are no skin lesions associated with herpes zoster and no arterial prominence associated with temporal arteritis. She is young for the latter, which would otherwise be a suitable explanation for fever, headache, scalp tenderness, and visual complaints (usually impairment not photophobia).

Incomplete bladder emptying and left buttock pain suggest that there might be a concomitant lumbosacral myelopathy or radiculopathy. Some nonbacterial causes of meningitis such as cytomegalovirus (CMV), syphilis, and cancer simultaneously involve the CNS and peripheral nerve roots. It is also possible that the scalp tenderness associated with combing reflects a cervical sensory radiculopathy.

She had presented to the ED 2 and 4 days before the current (third) ED visit. Both times her main complaint was left buttock pain and left leg paresthesias. Although she had no skin lesions, she was diagnosed with prodromal herpes zoster in the S2 dermatomal distribution and was prescribed valacyclovir (to be started should eruptions develop, which never occurred).

She reported intermittent self‐limited fevers at 3‐ to 4‐week intervals during the prior 6 months; two fever episodes were accompanied by an influenza‐like illness, and one was associated with gastrointestinal symptoms. Her last fever prior to this evaluation was 6 weeks earlier when she was treated with azithromycin for suspected pneumonia at an outside facility.

Her past medical history included hypothyroidism, gastroesophageal reflux disease, diverticulitis, and gluten intolerance. Her medications included porcine (natural) thyroid, fish oil, ibuprofen, and acetaminophen. She lived in Michigan and traveled to the northeast United States (Maine, Cape Cod, New Hampshire, Connecticut, and Vermont) 7 months prior to this evaluation. She was married and had no pets at home. She denied any tobacco, alcohol, or illicit drug use.

Her illness now appears to be chronic, associated with fever, and multisystem (potentially involving the pulmonary and gastrointestinal tract). None of her medical problems would predispose her to subacute meningitis, myelopathy, or radiculopathy. Hypothyroidism raises the possibility of a concomitant autoimmune disorder which causes meningitis, such as SLE or Behet's disease. Sarcoidosis can cause chronic meningitis and neuropathy with concomitant lung and gastrointestinal involvement and rarely fever.

Residency in the upper Midwest increases exposure to chronic infections that rarely cause subacute meningitis such as histoplasmosis, blastomycosis, or human granulocytic anaplasmosis. Travel to the northeast United States 1 month before the onset of her symptoms raises the possibility of other endemic infections like Lyme disease, babesiosis, and tularemia, which may account for her recurrent fevers. Of these, Lyme is most likely to present as chronic meningitis with cranial neuropathy and radiculoneuropathy.

Although the diagnosis of pneumonia was made late in her 6‐month illness, its etiology and treatment may be relevant. If the recent pneumonia was viral, a subsequent viral meningitis may be manifesting now or may have triggered an autoimmune process, such as acute disseminated encephalomyelitis. Bacterial pneumonia is a common precursor to bacterial meningitis, and treatment with azithromycin for the pneumonia may have delayed the meningitis onset or muted its course; this should be taken into account when interpreting cerebrospinal fluid (CSF) culture results.

On physical examination, her temperature was 39.1C, blood pressure was 135/91 mm Hg, with pulse of 87 beats per minute, respiratory rate of 16 breaths per minute, and oxygenation saturation of 97% on room air. She appeared in distress and was covering her eyes. She was alert and oriented. She had photophobia and mild nuchal rigidity. Pupils were equal and reactive to light, but she could not tolerate the eye exam for papilledema. Lung, heart, and abdominal exam were normal. No cranial nerve abnormalities were noted, and muscle strength was 5/5 in all 4 extremities. She had decreased sensation to light touch with allodynia throughout her lower extremities in addition to the lateral portion of the right scalp, which was also tender to palpation. Deep tendon reflexes were 2+ and symmetric in her bilateral upper and lower extremities. She did not have joint swelling, edema, lymphadenopathy, or a rash.

Her fever, headache, nuchal rigidity and photophobia collectively suggest meningitis, which requires evaluation by a lumbar puncture. There is no rash that supports herpes zoster or SLE. She does not have signs of myelopathy that would explain the urinary complaints, but lower motor neuron involvement has not been excluded. The sensory abnormalities in the scalp and leg are consistent with a polyneuroradiculopathy. Anterior lateral scalp tenderness may signal trigeminal nerve involvement, whereas posterior scalp tenderness would localize to the upper cervical cord nerve roots. The contralateral distribution of the scalp and leg sensory deficits suggests a multifocal peripheral nervous system process rather than a single CNS lesion.

Initial laboratory data showed serum white blood cell count (WBC) of 12,000/mm³ (79% polymorphonuclear leukocytes). Hemoglobin was 14.2 g/dL, and platelets were 251,000/mm³. Electrolytes, renal function, and liver function were normal. Thyroid‐stimulating hormone, erythrocyte sedimentation rate, and C‐reactive protein were normal. Urinalysis was negative. Chest x‐ray was normal. Noncontrast head computed tomography (CT) was normal. The patient was unable to void; 500 mL of urine returned when catheterization was performed.

CSF WBC count was 1,280/mm³ (39% neutrophils and 49% lymphocytes). CSF total protein was 175 mg/dL, and glucose was 48 mg/dL; serum glucose was 104 mg/dL. Opening pressure was not recorded. Gram stain was negative. Ceftriaxone, vancomycin, ampicillin, and acyclovir were administered for presumed bacterial or viral meningitis. Magnetic resonance imaging (MRI) of the brain and spine showed diffuse leptomeningeal enhancement (Figure 1).

Figure 1

The urinary retention in the absence of myelopathic findings on exam or MRI suggests a sacral polyradiculoneuropathy. Diffuse leptomeningeal enhancement is consistent with many, if not all, causes of meningitis. The high WBC count, elevated protein, and low glucosecollectively signaling active inflammation in the CNSare highly compatible with bacterial meningitis, although the lymphocytic predominance and other clinical data point to nonbacterial etiologies. The negative Gram stain further lowers the probability of bacterial meningitis, but it has limited sensitivity, may be affected by recent antibiotics, and is typically negative with Listeria. Enterovirus, acute human immunodeficiency virus (HIV), and herpes viruses (eg, CMV or herpes simplex virus [HSV]) are important considerations, with the latter 2 causing associated polyneuroradiculopathy. Patients with genital HSV (not detected here) can have a concomitant sacral radiculitis leading to urinary retention.

Fungal and mycobacterial meningitis is a possibility (especially with the high protein), but the patient does not have the typical multisystem disease or immunosuppression that frequently accompanies those conditions when CNS disease is present. Autoimmune conditions like SLE, Behet's disease, and sarcoidosis remain important conditions, especially with the polyneuroradiculopathy or mononeuritis multiplex, which may reflect multifocal nerve infarction or invasion. Similarly, lymphomatous or carcinomatous meningitis should be considered, although an isolated manifestation in the CNS is unusual. Based on the multifocal neurologic deficits, I favor a viral, spirochete, or malignant etiology of her meningoencephalitis.

Despite ongoing broad spectrum antibiotics and supportive care, she became confused on hospital day 3 and developed anomia, agitation, and worsening headache. A repeat CT of the brain did not show any new abnormalities, but repeat lumbar puncture demonstrated elevated intracranial pressure (opening pressure of 47 cm water) with 427 WBC/mm³. Blood and CSF cultures remained negative.

Detailed questioning of the family revealed that she had been horseback riding 3 weeks prior to admission; there were no other livestock where she rode horses. In addition, the family reported that she and other family members routinely drank raw milk from a cow share program.

HIV antibody test was negative. Herpes simplex, varicella zoster, enteroviruses, and adenovirus CSF polymerase chain reaction (PCR) were negative. Cytomegalovirus and Epstein‐Barr virus PCR were negative in serum and CSF. Arbovirus, lymphocytic choriomeningitis, Coccidioides, Blastomyces, Histoplasma, Brucella, and Lyme serologies were negative. Cryptococcus neoformans antigen was negative in CSF. Serum QuantiFERON‐TB test was negative. Blood and CSF acid‐fast bacilli smears (and eventually mycobacterial cultures) were also negative. Her CSF flow cytometry and cytology were negative for lymphoma.

Unpasteurized milk conveys multiple infectious risks. Listeriosis is a food‐borne illness that can cause meningoencephalitis, but peripheral neuropathies are not characteristic. Brucellosis is usually characterized by severe bone pain, pancytopenia, and hepatosplenomegaly, which are absent. Infection with Mycobacterium bovis mimics Mycobacterium tuberculosis and can cause multisystem disease, typically involving the lung. Campylobacter infection is characterized by gastroenteritis, which has not been prominent.

Rhodococcus equi is a horse‐related pathogen which leads to pulmonary infections in immunocompromised hosts but not meningitis. Rather than focusing on horse exposure alone, however, it may be useful to consider her at risk for vector‐borne pathogens based on her time outdoors, such as Lyme disease (which can cause radiculopathy and encephalopathy), West Nile virus (although motor weakness rather than sensory symptoms is typical), or eastern equine encephalitis.

The absence of weight loss, cytopenias, lymphadenopathy, and organomegaly with the negative CSF cytology and flow cytometry makes lymphomatous meningitis unlikely. The case for an autoimmune disorder is not strong in the absence of joint pains, rash, or autoimmune serologies. In a young woman with unexplained encephalitis, antibodies to the N‐methyl‐D‐aspartate receptor should be assayed.

Although the CSF leukocytosis is declining, the elevated pressure and clinical deterioration signal that the disease process is not controlled. At this point I am uncertain as to the cause of her progressive meningoencephalitis with polyneuroradiculopathy. The latter feature makes me favor a viral or spirochete etiology.

On hospital day 4, Coxiella burnetii serologies were reported as positive (phase II immunoglobulin [Ig] G 1:256; phase II IgM <1:16; phase I IgG <1:16; phase I IgM <1:16) suggesting acute Q fever. Antibiotics were changed to intravenous doxycycline and ciprofloxacin. Her increased intracranial pressure was managed with serial lumbar punctures. The patient was discharged after 12 days of hospitalization taking oral doxycycline and ciprofloxacin. Her symptoms resolved over 10 weeks. No vegetations were seen on transesophageal echocardiogram. She had no evidence of chronic Q fever on repeat serologies.

I was not aware that Q fever causes meningitis or meningoencephalitis. However, I should have considered it in light of her indirect exposure to cows. It is possible that her pneumonia 6 weeks earlier represented acute Q fever, as pneumonia and hepatitis are among the most typical acute manifestations of this infection.

COMMENTARY

Hospitalists are commonly confronted by the combination of fever, headache, and confusion and are familiar with the diagnostic and therapeutic dilemmas related to prompt discrimination between CNS and non‐CNS processes, particularly infections. At the time of this patient's final ED presentation, her illness unambiguously localized to the CNS. As common and emergent conditions such as acute bacterial meningitis were excluded, the greatest challenge was finding the clue that could direct investigations into less common causes of meningoencephalitis.

The Infectious Disease Society of America has developed clinical practice guidelines for the diagnosis and management of encephalitis which highlight the importance of epidemiology and risk factor assessment.[1] This approach requires the clinician to examine potential clues and to go beyond initial associationsfor instance, not simply linking horseback riding to horse‐associated pathogens, but interpreting horseback riding as a proxy for outdoor exposure, which places her at risk for contact with mosquitos, which transmit West Nile virus or eastern equine encephalitis. Similarly, ingestion of raw milk, which is typically linked to Listeria monocytogenes, Brucella, and other pathogens prompted the infectious disease consultant to think more broadly and include livestock (cow)‐associated pathogens including C. burnetii.

Although involvement of the CNS is common in chronic Q fever endocarditis due to septic embolism, neurologic involvement in acute Q fever varies in prevalence (range of 1.7%22%).[2, 3, 4] The 3 major neurological syndromes of acute Q fever are (1) meningoencephalitis or encephalitis, (2) lymphocytic meningitis, and (3) peripheral neuropathy (myelitis, polyradiculoneuritis, or peripheral neuritis). CSF analysis usually shows mild pleocytosis with a predominance of lymphocytic cells; CSF protein elevation is variable, and glucose is usually normal. Neuroradiologic examination is usually normal, and there are no pathognomonic imaging abnormalities for Q fever meningoencephalitis.[2, 3] The mechanism by which C. burnetii causes neurologic injury and dysfunction is unknown.

The diagnosis of Q fever is usually established by serologic testing. In acute Q fever, antibodies to phase II antigen are higher than the phase I antibody titer. Phase II IgM antibodies are the first to appear, but then decline on average after week 8, often reaching undetectable levels 10 to 12 weeks after disease onset.[5] If this patient's pneumonia 6 weeks prior to this presentation was acute Q fever pneumonia, her IgM titers may have been declining by the time her neurologic illness developed. A false negative test result is also possible; immunofluorescence assays are more specific than sensitive in acute Q fever.[5]

Evaluating this case in isolation may raise some doubt as to the accuracy of the diagnosis as she did not have a 4‐fold rise in the phase II IgG titer and did not have a detectable phase II IgM. However, she was part of a cluster of individuals who regularly consumed raw milk from the same dairy and had evidence of C. burnetii infection. This group included her spouse, who had a robust serologic evidence of C. burnetii, characterized by a >4‐fold rise in phase II IgM and IgG titers.[6]

C. burnetii is found primarily in cattle, sheep, and goats and is shed in large quantities by infected periparturient animals in their urine, feces, and milk.[7] Inhalation of contaminated aerosols is the principal route of transmission.[7, 8] Acute Q fever is underdiagnosed because the majority of acute infections are asymptomatic (60%) or present as a nonspecific flu‐like illness.[7] This case represents a rare manifestation of a rare infection acquired through a rare route of transmission, but highlights the importance of epidemiology and risk factor assessment when clinicians are faced with a diagnostic challenge.

TEACHING POINTS

• Exploration of epidemiology and exposure history is central to diagnosing meningoencephalitis with negative bacterial cultures and undetectable HSV PCR, although the etiology of meningoencephalitis can elude identification even after exhaustive investigation.
• Inhalation of contaminated aerosols is the principal route of transmission for C. burnetii, but it can also be transmitted via infected unpasteurized milk.[7, 9]
• Acute presentations of Q fever, which may warrant admission, include pneumonia, hepatitis, or meningoencephalitis.
• Q fever is diagnosed by serologic testing, and doxycycline is the antibiotic of choice.

Disclosures

This case was presented at the 2012 Annual Meeting of the Society of Hospital Medicine. It was subsequently reported in the epidemiologic report of the outbreak.[6] The authors report no conflicts of interest.

Although hospital readmissions have been a problem for at least the past 5 decades, they are now receiving more attention than ever before. Starting with the 2007 Medicare Payment Advisory Commission report detailing the vast scope of the problem,[1] readmissions have garnered substantial policy interest, culminating with Congress' inclusion of a penalty for hospitals with excessive readmission rates in the Affordable Care Act. Clinical leaders have become increasingly active in this issue as well, and hospitals around the nation have become engaged in finding ways to reduce the number of times patients return after discharge.

The Hospital Readmissions Reduction Program (HRRP), which is the penalty program put in place by Congress to address readmissions, has been controversial from its inception. Supporters point to the large number of patients whose discharge is fraught with poor communication, ineffective medication management, and inadequate handoffs to the primary care physician. Critics have countered that only a small proportion of readmissions are likely preventable by what hospitals can control,[2] and that patient factors, especially social and economic circumstances,[3] primarily drive readmissions. Despite this debate, we can all agree there is ample opportunity to improve the care of patients at the time of discharge.

In this context, we see important evidence emerging from the Better Outcomes by Optimizing Safe Transitions (BOOST) program. Funded by the Hartford Foundation among others, BOOST is specifically aimed at improving care transitions among older hospitalized adults. BOOST focuses on identifying those at highest risk for readmissions, communicating the discharge plan effectively, and ensuring close follow‐up, both through phone calls after discharge and timely appointments with primary care providers. These are all interventions that seem intuitively like good ideas. In this issue of the Journal of Hospital Medicine, leaders of the BOOST program report on the impact on readmissions rate.[4] However, as the accompanying editorial points out, the data are disappointing.[5] The evidence, seen in the best possible light, suggests a small improvement among a very select group of hospitals. Although the authors should be commended for writing up their findings, the fact that 19 of the 30 hospitals that received substantial training and assistance through the BOOST program chose not to report their data is unconscionable. The decision by those 19 hospitals to withhold data makes the results nearly uninterpretable and jeopardizes the hard work that so many others have engaged in. BOOST should require that hospitals agree to share data as a condition of participation in the program.

The Hansen study,[4] despite its disappointing findings, may signal that it is time for a new approach. First of all, we may need to focus on different metrics. Looking ahead, the most important question may not be Does BOOST lower readmission rates? but rather Does BOOST improve the care for patients at the time of discharge from the hospital? There are several good measures of the quality of a care transition, such as those by Coleman and colleagues,[6] and these could be used to measure the quality of care hospitals deliver at discharge. We could also develop new metrics of transitions of care. For example, hospitals truly committed to improvement could field an ongoing survey of primary care physicians in their community to ensure that care transitions are happening smoothly from the primary care providers' perspective. Patient experience metrics, beyond those captured in the Hospital Consumer Assessment of Healthcare Providers and Systems survey, may be necessary to better assess patient and family perspectives on the transition from the hospital to home. These and other approaches can help hospitals better understand how effectively they manage the handoff as patients leave their doors.

However, we should also recognize that although such approaches may improve care transitions, they are unlikely to substantially reduce readmissions. Instead, hospitals serious about reducing readmissions may need to reconsider their business model.[7] In the days following a discharge, patients are medically and socially vulnerable. Patients without robust social support at home may need more than just the right medications, a phone call, or a follow‐up appointment. They may need help with groceries, having their meals prepared, or getting a ride to the doctor's office. Hospitals that want to reduce readmissions may need to make investments in creating the community and social support that so many patients lack when they leave the hospital. This has never been part of the hospital business model before, but it may be time for a change.

The HRRP, an effort by federal policymakers to drive down readmissions through penalties, has clearly begun to make hospitals think about changing their business models in precisely these ways. Readmission rates are falling, although a concurrent increase in the number of patients being admitted to observation status makes it unclear whether patient care has actually improved. More data and time will tell. Furthermore, the program as currently designed targets hospitals that care for the sickest and poorest patients for penalties.[8] There are plenty of good options for addressing these unintended consequences, such as comparing safety‐net hospitals' performance to other similar institutions, or focusing only on preventable readmissions. However, regardless of its limitations, the HRRP in some form or another is here to stay. Therefore, hospitals will need to find ways to reduce readmissions, and programs like BOOST, even when executed perfectly, will be necessary but likely insufficient. Improving the quality of care transitions is critically important. But to truly get to better outcomes for older Americans, hospitals will need to think beyond their 4 walls.

Although hospital readmissions have been a problem for at least the past 5 decades, they are now receiving more attention than ever before. Starting with the 2007 Medicare Payment Advisory Commission report detailing the vast scope of the problem,[1] readmissions have garnered substantial policy interest, culminating with Congress' inclusion of a penalty for hospitals with excessive readmission rates in the Affordable Care Act. Clinical leaders have become increasingly active in this issue as well, and hospitals around the nation have become engaged in finding ways to reduce the number of times patients return after discharge.

The Hospital Readmissions Reduction Program (HRRP), which is the penalty program put in place by Congress to address readmissions, has been controversial from its inception. Supporters point to the large number of patients whose discharge is fraught with poor communication, ineffective medication management, and inadequate handoffs to the primary care physician. Critics have countered that only a small proportion of readmissions are likely preventable by what hospitals can control,[2] and that patient factors, especially social and economic circumstances,[3] primarily drive readmissions. Despite this debate, we can all agree there is ample opportunity to improve the care of patients at the time of discharge.

In this context, we see important evidence emerging from the Better Outcomes by Optimizing Safe Transitions (BOOST) program. Funded by the Hartford Foundation among others, BOOST is specifically aimed at improving care transitions among older hospitalized adults. BOOST focuses on identifying those at highest risk for readmissions, communicating the discharge plan effectively, and ensuring close follow‐up, both through phone calls after discharge and timely appointments with primary care providers. These are all interventions that seem intuitively like good ideas. In this issue of the Journal of Hospital Medicine, leaders of the BOOST program report on the impact on readmissions rate.[4] However, as the accompanying editorial points out, the data are disappointing.[5] The evidence, seen in the best possible light, suggests a small improvement among a very select group of hospitals. Although the authors should be commended for writing up their findings, the fact that 19 of the 30 hospitals that received substantial training and assistance through the BOOST program chose not to report their data is unconscionable. The decision by those 19 hospitals to withhold data makes the results nearly uninterpretable and jeopardizes the hard work that so many others have engaged in. BOOST should require that hospitals agree to share data as a condition of participation in the program.

The Hansen study,[4] despite its disappointing findings, may signal that it is time for a new approach. First of all, we may need to focus on different metrics. Looking ahead, the most important question may not be Does BOOST lower readmission rates? but rather Does BOOST improve the care for patients at the time of discharge from the hospital? There are several good measures of the quality of a care transition, such as those by Coleman and colleagues,[6] and these could be used to measure the quality of care hospitals deliver at discharge. We could also develop new metrics of transitions of care. For example, hospitals truly committed to improvement could field an ongoing survey of primary care physicians in their community to ensure that care transitions are happening smoothly from the primary care providers' perspective. Patient experience metrics, beyond those captured in the Hospital Consumer Assessment of Healthcare Providers and Systems survey, may be necessary to better assess patient and family perspectives on the transition from the hospital to home. These and other approaches can help hospitals better understand how effectively they manage the handoff as patients leave their doors.

However, we should also recognize that although such approaches may improve care transitions, they are unlikely to substantially reduce readmissions. Instead, hospitals serious about reducing readmissions may need to reconsider their business model.[7] In the days following a discharge, patients are medically and socially vulnerable. Patients without robust social support at home may need more than just the right medications, a phone call, or a follow‐up appointment. They may need help with groceries, having their meals prepared, or getting a ride to the doctor's office. Hospitals that want to reduce readmissions may need to make investments in creating the community and social support that so many patients lack when they leave the hospital. This has never been part of the hospital business model before, but it may be time for a change.

The HRRP, an effort by federal policymakers to drive down readmissions through penalties, has clearly begun to make hospitals think about changing their business models in precisely these ways. Readmission rates are falling, although a concurrent increase in the number of patients being admitted to observation status makes it unclear whether patient care has actually improved. More data and time will tell. Furthermore, the program as currently designed targets hospitals that care for the sickest and poorest patients for penalties.[8] There are plenty of good options for addressing these unintended consequences, such as comparing safety‐net hospitals' performance to other similar institutions, or focusing only on preventable readmissions. However, regardless of its limitations, the HRRP in some form or another is here to stay. Therefore, hospitals will need to find ways to reduce readmissions, and programs like BOOST, even when executed perfectly, will be necessary but likely insufficient. Improving the quality of care transitions is critically important. But to truly get to better outcomes for older Americans, hospitals will need to think beyond their 4 walls.

Although hospital readmissions have been a problem for at least the past 5 decades, they are now receiving more attention than ever before. Starting with the 2007 Medicare Payment Advisory Commission report detailing the vast scope of the problem,[1] readmissions have garnered substantial policy interest, culminating with Congress' inclusion of a penalty for hospitals with excessive readmission rates in the Affordable Care Act. Clinical leaders have become increasingly active in this issue as well, and hospitals around the nation have become engaged in finding ways to reduce the number of times patients return after discharge.

The Hospital Readmissions Reduction Program (HRRP), which is the penalty program put in place by Congress to address readmissions, has been controversial from its inception. Supporters point to the large number of patients whose discharge is fraught with poor communication, ineffective medication management, and inadequate handoffs to the primary care physician. Critics have countered that only a small proportion of readmissions are likely preventable by what hospitals can control,[2] and that patient factors, especially social and economic circumstances,[3] primarily drive readmissions. Despite this debate, we can all agree there is ample opportunity to improve the care of patients at the time of discharge.

In this context, we see important evidence emerging from the Better Outcomes by Optimizing Safe Transitions (BOOST) program. Funded by the Hartford Foundation among others, BOOST is specifically aimed at improving care transitions among older hospitalized adults. BOOST focuses on identifying those at highest risk for readmissions, communicating the discharge plan effectively, and ensuring close follow‐up, both through phone calls after discharge and timely appointments with primary care providers. These are all interventions that seem intuitively like good ideas. In this issue of the Journal of Hospital Medicine, leaders of the BOOST program report on the impact on readmissions rate.[4] However, as the accompanying editorial points out, the data are disappointing.[5] The evidence, seen in the best possible light, suggests a small improvement among a very select group of hospitals. Although the authors should be commended for writing up their findings, the fact that 19 of the 30 hospitals that received substantial training and assistance through the BOOST program chose not to report their data is unconscionable. The decision by those 19 hospitals to withhold data makes the results nearly uninterpretable and jeopardizes the hard work that so many others have engaged in. BOOST should require that hospitals agree to share data as a condition of participation in the program.

The Hansen study,[4] despite its disappointing findings, may signal that it is time for a new approach. First of all, we may need to focus on different metrics. Looking ahead, the most important question may not be Does BOOST lower readmission rates? but rather Does BOOST improve the care for patients at the time of discharge from the hospital? There are several good measures of the quality of a care transition, such as those by Coleman and colleagues,[6] and these could be used to measure the quality of care hospitals deliver at discharge. We could also develop new metrics of transitions of care. For example, hospitals truly committed to improvement could field an ongoing survey of primary care physicians in their community to ensure that care transitions are happening smoothly from the primary care providers' perspective. Patient experience metrics, beyond those captured in the Hospital Consumer Assessment of Healthcare Providers and Systems survey, may be necessary to better assess patient and family perspectives on the transition from the hospital to home. These and other approaches can help hospitals better understand how effectively they manage the handoff as patients leave their doors.

However, we should also recognize that although such approaches may improve care transitions, they are unlikely to substantially reduce readmissions. Instead, hospitals serious about reducing readmissions may need to reconsider their business model.[7] In the days following a discharge, patients are medically and socially vulnerable. Patients without robust social support at home may need more than just the right medications, a phone call, or a follow‐up appointment. They may need help with groceries, having their meals prepared, or getting a ride to the doctor's office. Hospitals that want to reduce readmissions may need to make investments in creating the community and social support that so many patients lack when they leave the hospital. This has never been part of the hospital business model before, but it may be time for a change.

The HRRP, an effort by federal policymakers to drive down readmissions through penalties, has clearly begun to make hospitals think about changing their business models in precisely these ways. Readmission rates are falling, although a concurrent increase in the number of patients being admitted to observation status makes it unclear whether patient care has actually improved. More data and time will tell. Furthermore, the program as currently designed targets hospitals that care for the sickest and poorest patients for penalties.[8] There are plenty of good options for addressing these unintended consequences, such as comparing safety‐net hospitals' performance to other similar institutions, or focusing only on preventable readmissions. However, regardless of its limitations, the HRRP in some form or another is here to stay. Therefore, hospitals will need to find ways to reduce readmissions, and programs like BOOST, even when executed perfectly, will be necessary but likely insufficient. Improving the quality of care transitions is critically important. But to truly get to better outcomes for older Americans, hospitals will need to think beyond their 4 walls.