Pump the brakes. In work with a vastly stronger empirical design, Atul Gupta found the program reduced mortality and readmissions. Observational studies have a role, but we can’t simply interpret them as causal. dropbox.com/s/rfwok9en2c58… @amitabhchandra2 @asacarny
The Gupta study is really good. Below is a summary and some thoughts. He notes that hospitals could respond to readmissions penalties in two ways: 1) by improving quality or 2) by changing composition of patients they admit (e.g. not admitting patients
Atul uses Medicare claims data. He exploits the fact that hospitals’ present readmissions rate impacts its penalty in the future. Hospitals with low readmissions in the past face little incentive to change. Hospitals with high recent readmits are exposed to the penalty
He constructs a measure of expectation of a penalty based on 2006-7 data. He also uses readmissions predicted by patient characteristics (e.g. not actual readmissions rates).
He finds that the penalty is associated with a 5% decrease in readmissions rates on average (9% for hospitals with high expected readmissions). He finds, on average, a 2% decrease in mortality.
Ditto, “The figure suggests that one-year mortality at the most at-risk-of-penalty hospitals has declined by about 1pp relative to the remaining hospitals”.
There are some important nuances worth pointing out. He does find that there’s a big decrease in the probability a hospital readmitts their own patients when they return to the ED, but no reduction for patients at other hospitals. This is evidence of some gaming.
This gaming makes these patients more likely to return later to the hospital, but there are no observable mortality effects.
His results show why instrumenting (unlike JAMA study) is important. His OLS estimates of the policy on readmissions are a third larger than IV estimates.
Overall, this is a really thoughtful study that uses rigorous, careful methods to examine a policy with no random variation. This paper viewed in the contend of the @Noahpinion and @amitabhchandra2 discussion of causality highlights the value of economics research.
While the JAMA study today made headlines, there’s not been much attention to Atul’s paper. This is unfortunate. There’s a long lag for Econ publications and economists are reticent about discussing their results with policy-makers in the press before peer review.
We also have to be upfront about two issues. First, no study is definitive. We need to shift opinion when a body of rigorous work points in a direction. Second, there’s a real issue with rigor and the empirical literacy present in the peer review process at some medical journals
