Unemployment and emergency room visits predict poor treatment outcome in primary care panic disorder.

Abstract

To complement existing data on predictors of treatment response in groups of "pure" panic disorder patients studied in clinical trials or in poorly controlled naturalistic follow-up, we sought to elucidate predictors of treatment response over 1 year in a diagnostically heterogeneous and comorbidly ill group of primary care patients with panic disorder participating in a randomized effectiveness study. Patients with DSM-IV panic disorder (N = 115), mostly without agoraphobia, were recruited from 3 primary care clinics in Seattle, Wash., and randomly assigned to an on-site collaborative care intervention (N = 57), in which psychiatrists provided education, 2 visits, follow-up phone calls, and paroxetine, or to usual care by their primary care physician (N = 58). Predictors of response at 3-month intervals over 1 year were determined using logistic regression analysis. Patients with consistent response over the year (response at the majority of available timepoints) were significantly (p <.05) more likely to be white, employed, in higher income strata, and in the intervention group and had less medical comorbidity and phobia severity, fewer recent hospitalizations and emergency room visits, and higher reported Medical Outcomes Study 36-Item Short Form physical and role functioning. The final regression model indicated that responders were more likely to be in the intervention group, be employed, and lack a recent emergency room visit. While some of the univariate findings partially replicate previous results linking greater illness severity with poorer response, univariate findings linking medical comorbidity and low socioeconomic status with poor response, as well as multivariate findings that unemployment and recent emergency room use are the most potent predictors of poor response, have not been previously reported.