Earlier this month, MedPAC staff presented their latest thinking on the Hospital Readmissions Reduction Program (HRRP). That’s the Medicare program that gives hospitals a small haircut for high rates of readmissions. I was not at the presentation, but the slides are online (PDF).
They raise several issues. I’ll focus on two. First, they note that low income patients have higher readmission rates due, perhaps, to reduced access to resources for post-hospitalization care. Their suggestion is not to adjust for income directly but to compare readmission rates among hospitals that treat a similar proportion of beneficiaries receiving cash assistance (SSI). They do not suggest changes based on any other socioeconomic factor. They note that race may also be relevant, but it’s effect is attenuated once income is accounted for.
The second issue is the relationship between mortality and readmissions. Rates of each could be inversely correlated for different reasons:
Hypothesis 1: Hospitals that keep very ill patients alive may have lower mortality, but higher readmission rates.
Hypothesis 2: Hospitals that admit more patients that could be treated on an outpatient basis may have more admissions, readmissions, and lower mortality per admission.
The point, I think, is that under hypothesis 1, readmissions are good because they keep patients alive that might have otherwise died. Under hypothesis 2 they are not because they use more expensive hospital resources for situations in which outpatient care would suffice.
MedPAC staff report that heart failure mortality rates are negative correlated with heart failure, acute myocardial infarction, pneumonia, and all-condition readmission rates (see, for example, this). However, they report an “insignificant correlation between [a] more inclusive mortality measure and all-condition readmissions.”
It’s not clear to me what the take away message from MedPAC is on the mortality/readmissions interaction. As Ashish Jha pointed out, lack of a positive correlation calls into question the validity of readmissions as a quality measure. Would we be comforted by a positive correlation? Perhaps. But consider again the negative correlation implied by hypothesis 1 above, under which readmissions are welcome.
