Cross-Classified Random Effects Analysis of Hospital and Surgeon Volume
Katherine Panageas, Sloan Kettering Cancer Center
*Sujata Patil, Memorial Sloan-Kettering Cancer Center
Deborah Schrag, Memorial Sloan Kettering Cancer Center
Qin Zhou, Memorial Sloan-Kettering Cancer Center
Keywords: Cross-classified data, random effects, hospital volume, surgeon volume
Evaluating whether an association exists between hospital or surgeon procedure volume and patient outcomes exists involves complex statistical issues. These complexities arise from the fact that the unit of observation is the patient, but these studies include multiple patients per hospital or surgeon as well as multiple hospitals or surgeons. Patient outcomes tend to be correlated within hospitals and within surgeons. Distinguishing these two effects is challenging for several reasons: strengths of the trends may be relatively modest; surgeon and hospital volumes are often highly correlated; and clustering is cross-classified rather than hierarchical. In this talk, we fit a cross classified random effects model to SEER-Medicare data and compare results from a Bayesian and penalized quasi-likelihood approach. We also report results from a simulation study.