A number of studies (e.g., Wasfy et al. 2017, Desai et al. 2016) have found that the Centers for Medicare and Medicaid Services (CMS) Hospital Readmissions Reduction Program (HRRP) reduce hospital readmissions. However, are these findings valid? A recent paper by Ody et al. (2019) argues maybe not. Because readmission rates are risk-adjusted, changing coding practices may explain differences in risk-adjusted readmission rates.
Between the March 2010 establishment of the HRRP and the October 2012 introduction of penalties, the Centers for Medicare and Medicaid Services (CMS) changed the electronic transaction standards that hospitals use to submit Medicare claims, allowing for an increased number of diagnosis codes. This change coincided with the time window in which risk-adjusted readmission rates declined the fastest.
Specifically, the number of diagnosis codes that could be coded on an inpatient claim increased from 10 prior to 2011 to 25 in 2011 and after.
To test this hypothesis, the authors use the 100% sample of the 2007-2014 Medicare’s Research Identifiable Files (RIF). The authors compare changes in readmissions for admissions targeted by HRRP (i.e., acute myocardial infarction, heart failure, and pneumonia) against changes in hospital admissions not targeted by HRRP. The authors then applied two risk adjustment approaches based on patient age, gender, and health conditions categories. The first risk adjustment approach predicted readmission risk based on all available diagnosis codes from prior-year claims; the second approach predicted readmission only based on the first 9 diagnosis codes from prior-year claims.
This study confirms the findings of Ibrahim and coauthors that increased patient risk scores
explain a meaningful share of the decrease in readmission rates following the passage of the HRRP…Readmission rates for targeted conditions and targeted hospitals decreased by 1.35 percentage points more during the period directly after implementation than would have been anticipated based upon the rate of decreases before implementation. However, a similar decrease occurred for nontargeted conditions at targeted hospitals and targeted conditions at nontargeted hospitals. Thus, we cannot conclude that the HRRP led to a differential decline relative to the observed decline in the comparison groups.
In short, if we believe that hospitals enacted hospital-wide policies to reduce readmissions as a direct result of HRRP, then HRRP may have still had a causal effect (1.35 percentage point reduction) on readmissions. If we believe that hospitals responded to HRRP by focusing on reducing readmissions only for the targeted conditions, then HRRP’s ability to reduce readmissions may be illusory.
The authors wisely note that if HRRP does not actually reduce readmissions, it may be harmful. First because one way to reduce readmissions is to refuse to readmit patients who need care; this option negatively affects patient health. Second, value-based payment exposes providers to volatility in reimbursements; if the pay-for-performance program does not actually improve quality but adds risks to hospital financials, then it would serve little purpose. This questions are yet to be answered definitively but the Ody et al. (2019) study provides a high-quality study to add to the debate.