The purpose of this study was to formulate and test two case-mix models for
depression treatment that permit comparisons of patient outcomes across di
verse clinical settings. It assessed demographics; eight, diagnostic-specif
ic,case-mix variables; and clinical status at baseline and follow-up for 18
7 patients. Regressions were performed to test two models for four dependen
t variables including depression severity and diagnosis. Individual treatme
nt settings were then ranked based on a comparison of actual versus predict
ed outcomes using regression coefficients and predictor variables. A model
inclusive of baseline physical health status and depression severity, predi
cted depression severity,, mental health, and physical health functioning a
t follow-tip. A simpler model performed well in predicting depression remis
sion. This study identifies variables to be included in case-mix adjustment
models and demonstrates statistical methods to control for differences acr
oss settings when comparing depression outcomes.