Models and Methods for Profiling Medical Providers

• Activity Number: 398
• Type: Invited
• Date/Time: Tuesday, August 11, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Health Policy Statistics Section
• Author(s): John D. Kalbfleisch* and Kevin He and Yanming Li
• Companies: University of Michigan and University of Michigan and University of Michigan
• Keywords: fixed effects, random effects, hierarchical models, risk adjustment, empirical null

Abstract

Methods that appropriately account for patient heterogeneity and natural variation among medical providers are important in the monitoring and profiling the providers for the purpose of quality improvement, information for patients and medical personnel, and as information to funders. In particular, we discuss methods used in defining measures of performance that compare outcomes of individual providers with those from a national cohort. We compare fixed and random effects as representations of provider effects in both linear and nonlinear models. Methods based primarily on fixed effects are proposed since they account appropriately for confounding between risk adjusters and provider effects, provide relatively simpler measures of facility effects, and provide more accurate estimates of provider effects for those facilities of primary interest, whose outcomes deviate substantially from the national norm. Natural variation between facilities is accounted for using robust methods based on the empirical null distribution and connections with random effects will be noted. The methods are illustrated in an example monitoring outcomes of dialysis facilities.