Using Latent Variable Models for Assessing Quality of Health Care: Issues and Applications
*Sharon-Lise T Normand, Harvard Medical School
Keywords: profiling, latent variables, mixed modeling, bayesian
Institutional profiling involves a comparison of a health care provider’s processes of care or outcomes to a standard in the form of a report card. The purpose is to quantify quality of care. Given the ubiquity of report cards and similar consumer ratings in contemporary American culture, it is notable that these are recent phenomenon in health care. There are several statistical considerations in quantifying quality of care. This talk describes the use of hierarchical models for characterizing quality of care and of Item Response Theory models to model multiple measures. Methods are illustrated to profile hospital 30-day mortality rates following isolated coronary artery bypass graft surgery in Massachusetts hospitals and US hospitals on the basis of multiple measures for diseased-based cohorts: 5 AMI measures from 2761 acute care hospitals; 4 CHF measures from 3271 hospitals.