Predicting depression in a chronic disease population using automated data

UnrestrictedBackground and Objective: Depression prevalence is elevated in chronic disease populations, and it can adversely affect health. This study develops a model to predict depression within a chronic disease population using administrative data.; Methods: The study population consisted of Medi-Cal beneficiaries with chronic heart or respiratory conditions or diabetes at baseline (1999). A split-sample approach was employed, with Cox proportional hazard regression used to predict depression from 2000-2002 from beneficiaries ' baseline characteristics. Variables were selected for inclusion based on published literature and pre-established thresholds. Model performance was assessed using discrimination and calibration.; Results: Depression occurred in 10.8% of beneficiaries. Age, gender, race, healthcare utilization, and specific medical conditions were significant predictors. The model had moderate discriminatory capability, and comparisons of observed versus expected depression cases across deciles of risk were not statistically significant.; Conclusions: This model may be useful in identifying individuals at risk of depression and focusing limited resources