In 2010 Congress and the Centers for Medicare & Medicaid Services concocted a program designed to mitigate the evil of hospital readmissions. Now it would be a good thing to avoid people going back to the hospital if it can be avoided. But that “if it can be avoided” can, in my judgment, only truly be determined by individual examination of each readmission by qualified clinicians, not by statistical legerdemain. The construct and outcomes of this program are important because substantial penalties are imposed on hospitals deemed to have excessive readmissions. The program appears to be a success because readmissions have been reduced. But, as a critique in Health Affairs points out, like a lot of research, when proper logical analysis and statistics are applied, the program may not have had any effect at all. (HA Article) The author’s primary analysis is extremely simple. The assumption behind the readmissions reduction program is that the readmission is somehow related to quality of care in the primary admission. And CMS assumed a benchmark readmission rate based on the pre-existing admission rates. But during the period of the program, admission rates were falling. You might expect a drop in readmission rates as well.
There is no reason to think the drop in overall admissions had anything to do with the readmissions reduction program. That drop has occurred across all payers and is due to attempts to avoid hospitalizations as a method of care, since it is the most expensive way to treat a condition. The authors took Medicare admissions from 2009 to 2014 and randomly removed admissions to mimic the decline in admission rates. They conducted this experiment with multiple random samples. Admission rates in Medicare declined from 19.65 per year per 100 beneficiaries in 2009 to 17.27 in 2014. The simulated rate of readmissions, under the assumption that readmissions are not related to care deficiencies in the prior admission, indicated that readmission rates declined at the same rate as admissions. This outcome is exactly what was observed in reality. Under the converse assumption that a readmission was caused by poor care in the admission, simulated readmission rates stayed the same throughout the study period, which is obviously not what actually happened. The results of this experiment are consistent with other research suggesting that at most a quarter of readmissions might be avoidable. And they are somewhat surprising because as admission rates fall, sicker patients on average are probably being admitted and so you might expect a rising readmission rate, even on a health status adjusted analysis. Just another reason for CMS to pull the program and give the hospitals back the money it took away from them based on a faulty analysis.