Using data analytics to improve nursing home quality performance

Objective We propose an alternative to the Risk Adjusted Mortality Rate (RAMR), about which we identify four serious concerns. We apply our method to cardiac surgery. Design We present a methodology that uses the upper and lower tail probabilities (UTP/LTP) of the binomial distribution to screen for poor/high performing providers. Study Setting The New York State Department of Health (NYS DOH) publicly releases data on all cardiac surgery patients in the state. We download cardiac surgery data from the NYS DOH website for the years 2011 through 2013. The state's objective is to identify poorly performing hospitals and surgeons and thereby reduce deaths. NYS employs the RAMR. Results The UTP/LTP approach agrees with the RAMR in its classification of all 132 surgeons and all 40 hospitals. However, performance is a continuous construct and strict categorization can lead to failure to identify marginal providers. Conclusions Our methodology addresses all four concerns regarding the RAMR. The UTP/LTP approach avoids inappropriate hypothesis testing and is consistent with standard statistical theory and practice in its approach to case volume. It does not require confidence intervals and it applies to all providers regardless of case volume.