Are We Targeting the Right Metric for Heart Failure?
Are We Targeting the Right Metric for Heart Failure?
The current emphasis on quality improvement and cost containment for highly prevalent chronic health conditions makes HF an ideal focus point. Understanding how to measure efficiency, resource use, and quality of care among hospitals treating patients with HF will be central to this effort. Related to this, our analysis yields several important findings. First, although variation exists in both hospital-level LOS and EOC, we demonstrate an increased variation in the latter. Second, hospital-level readmission rate was only modestly associated with overall resource use more than 30 days as assessed by total inpatient days for the EOC. Third, higher readmission rates were paradoxically associated with better adherence to quality indicators, including a composite measure of 100% adherence to CMS performance measures. Finally, hospital 30-day readmission rates were not associated with 30-day mortality rates. In contrast to readmission rates, better hospital performance on the 30-day EOC metric was significantly associated with decreased 30-day risk-adjusted mortality rates. Therefore, we conclude that CMS may wish to reconsider their emphasis on hospital 30-day readmission rates as the most important metric of hospital efficiency and quality of care.
Significantly, hospitals are under increasing pressure to reduce their 30-day readmission rate for common conditions such as HF. The CMS is now publicly reporting hospital readmission rates. In addition, under the Affordable Care Act, signed into law in 2010, CMS will likely begin piloting bundled payment programs that pay a single payment for a hospitalization, which covers all inpatient and outpatient care over a 30-day period. Under this system, hospitals will not only lose money for longer hospitalizations but also be reimbursed for readmissions for the same condition within 30 days. In this context, the concept of overall EOC (defined as all hospital days from an index admission and any readmission within 30 days for the same condition) becomes important and will likely be the key economic metric—at least from a provider- and hospital-level perspective. Surprisingly, although EOC is directly related to readmissions, we only found a modest relationship between overall EOC and 30-day readmission rate; therefore, if the health system evolves so that payers such as CMS are no longer reimbursing for readmissions within 30 days and hospitals' resource use is best measured by total inpatient days (ie, EOC), then 30-day readmission rates may become less relevant as an economic metric in the future.
To further examine the association between the concept of overall EOC and hospital quality, we analyzed the association between hospital quartile by EOC and CMS performance measures in addition to other quality measures. Similar to LOS and hospital readmission rate, we found an inconsistent relationship between EOC and quality indicators. Although in some cases, there was an association between hospitals in the longer EOC quartile and improved adherence to performance measures, this association did not hold true for all cases. Moreover, prior studies have not consistently shown an association between currently measured performance measures and patient outcomes. In part, the lack of correlation between quality measures and short-term patient outcomes may reflect the time that it takes to see a benefit of improved adherence to evidence-based therapies on outcomes. These findings likely highlight the difficulty in defining quality measures that reflect short- or long-term outcomes.
Our analysis also failed to demonstrate an association between hospitals with increased readmission rates and short-term mortality. This finding confirms a prior analysis by Heidenreich et al demonstrating a lack of correlation between readmission rates and mortality in the Veterans Affairs Health Care System. The lack of association between 30-day readmission rate and mortality was also demonstrated in a larger analysis of more than 3,800 hospitals in the Hospital Compare database. As a result, our analysis and prior studies demonstrate that 30-day readmission rate does not correlate well with total resource use or short-term mortality. Our data and the changing health care system suggest the need for a metric that better captures hospital and outpatient care quality and efficiency. As the US health care system evolves, the hospital and outpatient systems of care may be combined into a single unit termed the accountable care organization. The concept of EOC as an economic and quality metric may be ideal, reflecting total resource use, system efficiency, and quality. Indeed, our analysis suggests that hospitals in quartiles with the shortest EOC are associated with reduced 30-day mortality. In addition, variation in average hospital EOC in our analysis suggests the need for improvement in overall systems of care among health systems with the longest number of total inpatient days.
However, our inferences must be tempered with the realization that it is difficult to identify a single ideal metric of quality and inefficiency. As we have pointed out previously, established quality measures do not necessarily correlate with readmission. Furthermore, they poorly correlate with short-term mortality. Likewise, short-term readmission does not correlate well with short-term mortality. Our data confirm the more detailed analysis by Hernandez et al, describing the complexities of using 30-day readmission, 30-day mortality, or adherence to recommended therapies as measures of quality of care for patients with HF. Although we believe that EOC is a better quality measure (incorporating quality and efficiency of care), further prospective study is needed to fully understand its use. In addition, considering metrics like days alive and out of the hospital might be a useful way to incorporate both readmission and mortality.
One further consideration to be taken into account when determining the best hospital-level quality indicators is the influence of geographic region, socioeconomic make-up, and resource availability of individual institutions. These factors may independently predict patient outcomes regardless of adherence to core quality or performance measures. Indeed, under a value-based purchasing reimbursement program, so-called safety net hospitals (ie, rural, resource-poor, serving socioeconomically disadvantaged patients) may suffer financially.
Our study has several important limitations. First, this is a retrospective observational study, so there may not be adjustment for unmeasured or residual confounding. For example, our finding that hospitals with shorter LOS do not have higher readmission rates may be confounded by illness severity in the patient population that we cannot fully adjust for, despite using a well-validated approach. Likewise, illness severity may confound the relationship between quality measures and outcomes, despite adjustment. Second, the hospitals in this study were all participating in a national quality improvement program, and this may limit the generalizability of the results. Third, the relationships observed in this cross-sectional group do not imply causality. For example, one cannot conclude that increasing LOS might not improve 30-day readmission rate based on this analysis. Our analysis of 30-day mortality excluded inpatient mortality. Because the primary scope of the study was to examine the relationship among LOS, EOC, and 30-day readmissions, patients had to survive to hospital discharge to be included in the primary analysis. Finally, patients included in this study were all fee-for-service Medicare beneficiaries, which also could limit the generalizability of our findings with respect to the larger population of patients with HF. The median age of this Medicare population was 80 years, and patients were mostly white. This limits generalizability of our data to a younger, more ethnically diverse population.
In conclusion, the 30-day rehospitalization metric was not associated with hospital LOS, better performance on quality measures, or 30-day mortality risk. Furthermore, there was only a modest association between hospital ranking by overall 30-day EOC and readmission rates. Hospitals with the shortest rates of 30-day readmission paradoxically performed worse on current HF performance measures. The 30-day rehospitalization measure was not associated with lower 30-day mortality. In contrast, the shorter EOC quartile was associated with decreased odds of 30-day mortality. Total hospital days over a 30-day EOC may be a novel and more accurate metric of health system quality, resource use, and outcomes.
Discussion
The current emphasis on quality improvement and cost containment for highly prevalent chronic health conditions makes HF an ideal focus point. Understanding how to measure efficiency, resource use, and quality of care among hospitals treating patients with HF will be central to this effort. Related to this, our analysis yields several important findings. First, although variation exists in both hospital-level LOS and EOC, we demonstrate an increased variation in the latter. Second, hospital-level readmission rate was only modestly associated with overall resource use more than 30 days as assessed by total inpatient days for the EOC. Third, higher readmission rates were paradoxically associated with better adherence to quality indicators, including a composite measure of 100% adherence to CMS performance measures. Finally, hospital 30-day readmission rates were not associated with 30-day mortality rates. In contrast to readmission rates, better hospital performance on the 30-day EOC metric was significantly associated with decreased 30-day risk-adjusted mortality rates. Therefore, we conclude that CMS may wish to reconsider their emphasis on hospital 30-day readmission rates as the most important metric of hospital efficiency and quality of care.
Significantly, hospitals are under increasing pressure to reduce their 30-day readmission rate for common conditions such as HF. The CMS is now publicly reporting hospital readmission rates. In addition, under the Affordable Care Act, signed into law in 2010, CMS will likely begin piloting bundled payment programs that pay a single payment for a hospitalization, which covers all inpatient and outpatient care over a 30-day period. Under this system, hospitals will not only lose money for longer hospitalizations but also be reimbursed for readmissions for the same condition within 30 days. In this context, the concept of overall EOC (defined as all hospital days from an index admission and any readmission within 30 days for the same condition) becomes important and will likely be the key economic metric—at least from a provider- and hospital-level perspective. Surprisingly, although EOC is directly related to readmissions, we only found a modest relationship between overall EOC and 30-day readmission rate; therefore, if the health system evolves so that payers such as CMS are no longer reimbursing for readmissions within 30 days and hospitals' resource use is best measured by total inpatient days (ie, EOC), then 30-day readmission rates may become less relevant as an economic metric in the future.
To further examine the association between the concept of overall EOC and hospital quality, we analyzed the association between hospital quartile by EOC and CMS performance measures in addition to other quality measures. Similar to LOS and hospital readmission rate, we found an inconsistent relationship between EOC and quality indicators. Although in some cases, there was an association between hospitals in the longer EOC quartile and improved adherence to performance measures, this association did not hold true for all cases. Moreover, prior studies have not consistently shown an association between currently measured performance measures and patient outcomes. In part, the lack of correlation between quality measures and short-term patient outcomes may reflect the time that it takes to see a benefit of improved adherence to evidence-based therapies on outcomes. These findings likely highlight the difficulty in defining quality measures that reflect short- or long-term outcomes.
Our analysis also failed to demonstrate an association between hospitals with increased readmission rates and short-term mortality. This finding confirms a prior analysis by Heidenreich et al demonstrating a lack of correlation between readmission rates and mortality in the Veterans Affairs Health Care System. The lack of association between 30-day readmission rate and mortality was also demonstrated in a larger analysis of more than 3,800 hospitals in the Hospital Compare database. As a result, our analysis and prior studies demonstrate that 30-day readmission rate does not correlate well with total resource use or short-term mortality. Our data and the changing health care system suggest the need for a metric that better captures hospital and outpatient care quality and efficiency. As the US health care system evolves, the hospital and outpatient systems of care may be combined into a single unit termed the accountable care organization. The concept of EOC as an economic and quality metric may be ideal, reflecting total resource use, system efficiency, and quality. Indeed, our analysis suggests that hospitals in quartiles with the shortest EOC are associated with reduced 30-day mortality. In addition, variation in average hospital EOC in our analysis suggests the need for improvement in overall systems of care among health systems with the longest number of total inpatient days.
However, our inferences must be tempered with the realization that it is difficult to identify a single ideal metric of quality and inefficiency. As we have pointed out previously, established quality measures do not necessarily correlate with readmission. Furthermore, they poorly correlate with short-term mortality. Likewise, short-term readmission does not correlate well with short-term mortality. Our data confirm the more detailed analysis by Hernandez et al, describing the complexities of using 30-day readmission, 30-day mortality, or adherence to recommended therapies as measures of quality of care for patients with HF. Although we believe that EOC is a better quality measure (incorporating quality and efficiency of care), further prospective study is needed to fully understand its use. In addition, considering metrics like days alive and out of the hospital might be a useful way to incorporate both readmission and mortality.
One further consideration to be taken into account when determining the best hospital-level quality indicators is the influence of geographic region, socioeconomic make-up, and resource availability of individual institutions. These factors may independently predict patient outcomes regardless of adherence to core quality or performance measures. Indeed, under a value-based purchasing reimbursement program, so-called safety net hospitals (ie, rural, resource-poor, serving socioeconomically disadvantaged patients) may suffer financially.
Our study has several important limitations. First, this is a retrospective observational study, so there may not be adjustment for unmeasured or residual confounding. For example, our finding that hospitals with shorter LOS do not have higher readmission rates may be confounded by illness severity in the patient population that we cannot fully adjust for, despite using a well-validated approach. Likewise, illness severity may confound the relationship between quality measures and outcomes, despite adjustment. Second, the hospitals in this study were all participating in a national quality improvement program, and this may limit the generalizability of the results. Third, the relationships observed in this cross-sectional group do not imply causality. For example, one cannot conclude that increasing LOS might not improve 30-day readmission rate based on this analysis. Our analysis of 30-day mortality excluded inpatient mortality. Because the primary scope of the study was to examine the relationship among LOS, EOC, and 30-day readmissions, patients had to survive to hospital discharge to be included in the primary analysis. Finally, patients included in this study were all fee-for-service Medicare beneficiaries, which also could limit the generalizability of our findings with respect to the larger population of patients with HF. The median age of this Medicare population was 80 years, and patients were mostly white. This limits generalizability of our data to a younger, more ethnically diverse population.
In conclusion, the 30-day rehospitalization metric was not associated with hospital LOS, better performance on quality measures, or 30-day mortality risk. Furthermore, there was only a modest association between hospital ranking by overall 30-day EOC and readmission rates. Hospitals with the shortest rates of 30-day readmission paradoxically performed worse on current HF performance measures. The 30-day rehospitalization measure was not associated with lower 30-day mortality. In contrast, the shorter EOC quartile was associated with decreased odds of 30-day mortality. Total hospital days over a 30-day EOC may be a novel and more accurate metric of health system quality, resource use, and outcomes.