Regression to the mean is a known statistical phenomenon. It occurs when an outcome is measured multiple times. Outcomes that are extreme relative to the statistical average, or mean, during the first measurement are more likely to be closer to the mean in subsequent measurement periods simply by chance, because more extreme values have a lower probability of occurring. What is important when considering whether regression to the mean might play a role is whether the variable being measured has a component that can be considered random, or chance.
Sood and his colleagues used 2006 through 2014 Medicare data to identify hospitalizations and 30-day readmission episodes to analyze whether or not documented declines in readmissions at hospitals penalized for excess readmissions under the hospital readmissions reduction program was due to regression to the mean or an effect of the policy.
In 2010 Medicare introduced the Hospital Readmission Reduction Program (HRRP), one of a series of reforms under the Affordable Care Act aimed at improving quality and lowering costs across the healthcare sector.
Under HRRP, Medicare started penalizing hospitals with excess readmissions for heart failure, acute myocardial infarction, and pneumonia in fiscal year 2013. To identify hospitals considered to have excess readmissions, Medicare used fiscal years 2008 through 2011 as the baseline.
After calculating the change in readmission rates (based on the baseline 3 years of 2008-11) for hospitals that were deemed to have excess readmissions in 2013, Sood and his colleagues ran a series of thought experiments to tease out whether the policy incentives really had an effect on the outlier hospitals that reported excess readmissions and were consequently penalized.
For example, they set up a scenario to measure the effect of an artificial HRRP treatment 3 years before the actual program was implemented and observed changes in excess readmissions as if the policy were in effect.
They found that hospitals with below-mean performance during the 2008-2011 baseline window experienced a decline of 4.8 to 7.1 percentage points in excess readmissions during the subsequent 3 years. However, 74.3 percent to 86.5 percent of the improvement observed was explained by regression to the mean. Similarly, hospitals that had better than average baseline performance during the baseline period experienced an increase in the subsequent reporting period, of which 83.6 percent to 91.8 percent could be explained by this statistical anomaly.
“Most of the decline in readmission rates in hospitals with high rates during the measurement window for the first year of the HRRP appeared to be due to RTM. These findings seem to call into question the notion of an HRRP policy effect on readmissions,” conclude the authors.
The full study can be found at JAMA Internal Medicine. A press release about the study is here.