Title: Outcome-dependent sampling in cluster-correlated settings with application to hospital profiling
Authors: Glen McGee - Harvard University (United States) [presenting]
Jonathan Schildcrout - Vanderbilt University (United States)
Sharon-Lise Normand - Harvard University (United States)
Sebastien Haneuse - Harvard TH Chan School of Public Health (United States)
Abstract: Hospital readmission is a key marker of quality of health care used by the Centers for Medicare and Medicaid Services to determine hospital reimbursement rates. Analyses of readmission are based on a logistic-normal generalized linear mixed model (GLMM) that permits estimation of hospital-specific measures while adjusting for case-mix differences. Moves to address healthcare disparities call for expanding adjustment to include socioeconomic measures while minimizing burden to hospitals associated with data collection. We propose the detailed socioeconomic data to be collected on a sub-sample of patients via an outcome-dependent sampling scheme, specifically the cluster-stratified case-control (CSCC) design. Estimation and inference, for both fixed and random effects components, is performed via pseudo-maximum likelihood wherein inverse-probability weights are incorporated into the integrated likelihood to account for the design. In simulations, CSCC sampling proves to be an efficient design whenever interest lies in fixed or random effects of a GLMM and covariates are unobserved or expensive to collect. Methods are illustrated via a motivating analysis of Medicare beneficiaries hospitalized for congestive heart failure at one of 3,116 hospitals. Results show that the proposed framework provides a means of mitigating disparities in terms of which hospitals are indicated as being poor performers, relative to a naive analysis that fails to adjust for missing case-mix variables.