Accounting for the Growth of Observation Stays in the Assessment of Medicare’s Hospital Readmissions Reduction Program

Importance Decreases in 30-day readmissions following the implementation of the Medicare Hospital Readmissions Reduction Program (HRRP) have occurred against the backdrop of increasing hospital observation stay use, yet observation stays are not captured in readmission measures. Objective To examine whether the HRRP was associated with decreases in 30-day readmissions after accounting for observation stays. Design, Setting, and Participants This retrospective cohort study included a 20% sample of inpatient admissions and observation stays among Medicare fee-for-service beneficiaries from January 1, 2009, to December 31, 2015. Data analysis was performed from November 2021 to June 2022. A differences-in-differences analysis assessed changes in 30-day readmissions after the announcement of the HRRP and implementation of penalties for target conditions (heart failure, acute myocardial infarction, and pneumonia) vs nontarget conditions under scenarios that excluded and included observation stays. Main Outcomes and Measures Thirty-day inpatient admissions and observation stays. Results The study included 8 944 295 hospitalizations (mean SD age, 78.7 8.2 years; 58.6% were female; 1.3% Asian; 10.0% Black; 2.0% Hispanic; 0.5% North American Native; 85.0% White; and 1.2% other or unknown). Observation stays increased from 2.3% to 4.4% (91.3% relative increase) of index hospitalizations among target conditions and 14.1% to 21.3% (51.1% relative increase) of index hospitalizations for nontarget conditions. Readmission rates decreased significantly after the announcement of the HRRP and returned to baseline by the time penalties were implemented for both target and nontarget conditions regardless of whether observation stays were included. When only inpatient hospitalizations were counted, decreasing readmissions accrued into a −1.48 percentage point (95% CI, −1.65 to −1.31 percentage points) absolute reduction in readmission rates by the postpenalty period for target conditions and −1.13 percentage point (95% CI, −1.30 to −0.96 percentage points) absolute reduction in readmission rates by the postpenalty period for nontarget conditions. This reduction corresponded to a statistically significant differential change of −0.35 percentage points (95% CI, −0.59 to −0.11 percentage points). Accounting for observation stays more than halved the absolute decrease in readmission rates for target conditions (−0.66 percentage points; 95% CI, −0.83 to –0.49 percentage points). Nontarget conditions showed an overall greater decrease during the same period (−0.76 percentage points; 95% CI, −0.92 to −0.59 percentage points), corresponding to a differential change in readmission rates of 0.10 percentage points (95% CI, −0.14 to 0.33 percentage points) that was not statistically significant. Conclusions and Relevance The findings of this study suggest that the reduction of readmissions associated with the implementation of the HRRP was smaller than originally reported. More than half of the decrease in readmissions for target conditions appears to be attributable to the reclassification of inpatient admission to observation stays.