The study examined demographic variables, psychosocial outcome measures, and two treatment variables as possible predictors of attrition. The sample consisted of 222 women, of whom 129 were assigned to the experimental day treatment group and 93 were assigned to the comparison outpatient group. The participants were assessed by a psychiatric social worker and, if appropriate for outpatient treatment, randomly assigned to a treatment program where they were interviewed by trained research personnel. They were interviewed again when they left treatment and at 3, 6, and 12 months after discharge. The interviews were structured and lasted approximately 2.5 hours. Data were analyzed for the 98 women who completed treatment and the 124 women who dropped out of treatment after September 1992, when the revised research design was implemented. Attrition rates increased from admission, peaked at the 3-month follow-up, and then declined substantially by the 12-month follow-up period. Logistic regression analysis revealed that only treatment discharge status significantly predicted attrition. Tables, figures, author photograph and biography, and 40 references (Author abstract modified)