Intensive residential treatment (IRT) is an effective management approach for those with severe obsessive-compulsive disorder (OCD). This study aimed to identify IRT response predictors for clinical and research use. Consecutive subjects admitted to the Massachusetts General Hospital/McLean OCD Institute (OCDI) between February 1997 and June 2003 were included (N=476). IRT responder and non-responder group characteristics were compared using t-tests and chi(2) analyses. Multiple regression analysis modeled relationships between final OCD severity (Yale-Brown Obsessive-Compulsive scale scores) and predictor variables, while accounting for multicollinearity and potential outliers. Treatment responders comprised 59.3% of the treatment sample. Responders had significantly fewer males (p=0.02), lower depression severity (p=0.03), poorer psychosocial functioning (p=0.03) and fewer tic disorders (0.04), but were not different with respect to admission length, age, marital or employment status, OCD onset, family OCD history, treatment or admission history. In the final regression model, decreased initial OCD severity (p<0.001), female gender (p=0.003) and better initial psychosocial functioning (Work and Social Adjustment scale scores) (p=0.003) were predictors of less severe OCD at discharge (adjusted R-square=0.28). Depression severity (Beck Depression Inventory scores) and insight were not predictive of treatment outcome. Future research is necessary to elucidate putative relationships between gender and OCD psychopathology, and to understand the interplay of psychosocial factors, OCD severity and treatment outcome.