The Effects of Medicare's Hospital Readmission Reduction Program

Nearly 20% of Medicare beneficiaries are readmitted within 30 days of hospital discharge. Because many readmissions are thought to be avoidable, high readmission rates are widely considered a sign of poor quality of care. They are also very costly- it is estimated that readmissions of Medicare patients cost the Centers for Medicare and Medicaid Services (CMS) more than $17 billion dollars per year. As a result, reducing re-hospitalizations of Medicare beneficiaries has become a top priority. As part of the Affordable Care Act (ACA), the Hospital Readmission Reduction Program (HRRP) was established in an effort to prevent re-hospitalizations and thereby improve health care quality and lower costs.

Despite the rapid scaling up of HRRP, the success of the program is unknown. Early evaluations have found that the HRRP was associated with reductions in Medicare readmission rates. However, the inferences from these studies are compromised due to weak evaluation designs and statistical methods. The primary challenge to evaluating the impact of HRRP concerns understanding what would have happened to readmission rates if HRRP had not been implemented. Evidence from published evaluations shows strong signs of program spillovers, in which the HRRP appeared to decrease readmission rates for non-Medicare patients and for non-targeted diagnoses, such as gastrointestinal diseases. As a result, estimates of the effect of HRRP from these studies are likely to be biased towards the null. This study attempts to overcome these evaluation challenges when estimating the impact of the HRRP.

We evaluated the effect of the HRRP on hospital readmissions for heart attack, heart failure, and pneumonia during its first 51 months of implementation (April 2010 through June 2014). Our study population includes all US hospitals in the Inpatient Prospective Payment System. We perform a difference-in-differences analysis testing whether there were greater reductions in risk adjusted 30-day readmission rates, relative to risk adjusted 30-day mortality rates, after the HRRP was initiated. Because the baseline levels of readmission rates are somewhat different than baseline mortality rates, we log transformed these rates so that we could interpret the impact of the HRRP as a percentage change, rather than a level change. Because we are interested in testing whether the impact of HRRP changed over time, we specified exposure to the HRRP as the cumulative monthly exposure to the program.

We found that each month of exposure to the program was associated with a 0.14% reduction in pneumonia patients’ risk of readmission (p < 0.001), a 0.29% reduction in heart failure patients’ risk of readmission (p < 0.001), and a 0.06% reduction in heart attack patient’s risk of readmission (p < 0.001).