National Patterns of Risk-Standardized Mortality and Readmission After Hospitalization for Acute Myocardial Infarction, Heart Failure, and Pneumonia: Update on Publicly Reported Outcomes Measures Based on the 2013 Release

Between July 2009 and June 2012, the current 3-year measurement period for CMS’s public reporting of hospital performance, our results demonstrate improvement in all annual risk-standardized outcome rates for patients admitted with AMI, HF, and pneumonia, except an increase of 0.2% absolute percentage points in HF mortality. Despite this improvement, in half of U.S. hospitals, one in seven patients admitted for AMI will die within 30 days of admission and one in four patients admitted with HF will be readmitted with an unplanned readmission within 30 days of discharge. Given the complexity of improving outcomes, improvement in national rates is likely to be slow, but these results allow an assessment of the trajectory of outcomes within the 3 years included in current publicly-reported results, showing that mortality following admission for AMI continues to decline and unplanned readmission rates for all three conditions indicate early signs of improvement. However, HF mortality rose slightly over the reporting period and pneumonia mortality did not demonstrate the marked declines shown prior to 2005.19 These results reflect outcomes after the start of public reporting and inclusion of these measures in CMS’s pay-for-performance programs.20 In addition, the readmission rates now focus solely on unplanned readmissions.

The measures continue to demonstrate wide performance variation, with outcome rates for high performing hospitals (10^th percentile) at least three absolute percentage points lower than those for poor performing hospitals (90^th percentile) for five of six measures; the AMI readmission measure demonstrates 2.4 absolute percentage points between top and bottom deciles. Documentation of this variation provides evidence of continued room for improvement in care quality. The publicly reported measures are designed to benchmark an individual hospital against how an average hospital performs when caring for a similar case-mix. The goal of public reporting is not to reduce mortality or readmission to zero, but rather to bring as many hospitals as possible to the performance standards set by the nation’s highest performing hospitals.

Although cross-sectional and including only the 3-year current measurement period, these data are the first indication that measure-specific readmission rates are declining and support evidence published by Gerhardt, et. al in 2013 that all-condition readmission rates are declining.21 Unlike that study as well as previous hospital-level results reported on CMS’s Hospital Compare, our findings indicate declining rates of unplanned readmissions after hospitalization for AMI, HF, and pneumonia. These are the first data to exclude planned readmissions from the measures’ outcome and therefore it is inappropriate to compare them directly to prior publicly reported results. However, by excluding planned readmissions, such as elective surgeries or planned chemotherapy, these results better assess hospital care quality. Using a uniform definition of unplanned readmission, median annual RSRRs for AMI and HF decreased by nearly a full (absolute) percentage point within the last 3 years of available data, compared to year-to-year changes in annual median RSRRs ≤ 0.3 absolute percentage points between 2007 and 2010.22,23 While some year-to-year variation in rates cannot be excluded, given the magnitude of recent changes compared to recent historic data, these improvements are unlikely to be random variation. If all hospitals with RSRRs in excess of the median value had improved to the median hospital RSRR, 5,072 AMI, 11,839 HF, and 14,494 pneumonia unplanned readmissions would have been averted during the 2013 public reporting period.

One concern expressed by critics of examining hospital readmissions as a measure of hospital quality is that focus on readmissions will divert attention from other important quality improvement efforts and/or result in unintended consequences.24 The most serious concern is that efforts to reduce readmissions may lead hospitals to avoid re-hospitalizing patients who need more intensive care, thus leading to increased mortality rates in the post-discharge period. With the exception of HF, which demonstrated a 0.2% absolute percentage point increase in median hospital RSMRs between the first and last year of the current 3-year reporting period, AMI and pneumonia mortality rates fell or were stable simultaneous to declining readmission rates. While this does not exclude a relationship between hospital performance on mortality and readmission measures, this finding is reassuring and supports earlier data that good performance on readmission measures does not necessarily occur at the expense of performance on the mortality measures or vice versa.25 Gorodeski et al. demonstrated that, at higher readmission rates, lower HF mortality was associated with higher readmission; however, this relationship did not exist for hospitals with the lowest readmission rates, where mortality performance remained flat,26 and the study did not examine outcomes for other conditions that have been studied previously and found to have little correlation.25 There is evidence that patients admitted now are sicker, as each measure shows small increases in the frequency of several risk covariates, such as cardiorespiratory failure and shock, over the measurement period (online Appendix), supporting that the patient case mix may be increasing in severity overtime. Despite this, only HF mortality showed an overall increase in mortality, and in fact, there was a decrease in HF mortality between the second and third years of the reporting period. These data merit further consideration, but reinforce the benefit of measuring and reducing both mortality and readmission rates simultaneously to prevent unintended consequences of measurement and reporting. Further, these measures supplement data captured by existing publicly reported process measures that examine and incentivize guideline-concordant care, but often correlate poorly with actual patient outcomes.27

There are several limitations to consider in interpreting these results. First, administrative claims carry less face validity for adequate case-mix adjustment compared to clinical data. The risk models used in these measures have been validated against and demonstrate similar performance assessments to models based on medical record data.9‐13 Second, reported rates only reflect the experience of Medicare FFS and VA patients ages ≥ 65 and are not necessarily generalizable to other populations. Third, Medicare and VA data are combined prior to fitting the models, allowing between hospital means and variances to be different for the two systems. We did not explore whether VA hospitals represented a disproportionate number of improving hospitals; however, VA hospitals contribute less than 3% of measured patients and therefore the rates are largely driven by outcomes for index admissions to non-federal hospitals. Fourth, although we have used a robust risk-adjustment approach endorsed by statistical experts,28 there may be other sources of variation contributing to the differences among hospital RSMRs and RSRRs, including differences in coding practices that could impact either risk adjustment or cohort definitions. This has been identified as a concern for the pneumonia measures, with recent evidence suggesting that more seriously ill pneumonia patients may be coded with respiratory failure or sepsis and thus may not be included in measurement.29 Further, the measures do not adjust for code status or do not resuscitate orders, as these are inconsistently applied by clinicians and not captured in claims. However, the mortality measures exclude patients enrolled in hospice at any time during the 12-months prior to admission or who enrolled in hospice on the first day of admission, as the goals of care for such patients likely differ from patients not participating in hospice. These measures have been previously validated using medical record-based models, minimizing the likelihood that coding differences are the main source of variation for all measures. Variation in the propensity to admit patients or in care intensity may also contribute to performance differences.

Examination of 2009–2012 data reveals variation in the quality of care provided to patients with AMI, HF, and pneumonia. Our study summarizes unplanned readmission rates nationally and signals potential early declines in readmission rates. The 2013 publicly reported measures update suggests payment incentive programs may be influencing hospital readmission performance without a concomitant increase in mortality, at least for AMI and pneumonia patients, and supports the need for simultaneously measuring mortality and readmission to encourage better quality across all aspects of care and monitor for unintended consequences.