OBJECTIVE: Two assertive community treatment teams were compared with a usual-care control condition based on their ability to engage and retain clients with serious and persistent mental illness in community- based mental health services. METHODS: Clients were randomly assigned to one of two assertive community treatment teams (N = 116) or to usual care (N = 58). Survival analysis was used to compare clients' length of engagement and retention in service in the two treatment conditions and in usual care. Cox regression analyses were conducted to determine whether demographic, program, or client variables were significantly associated with length of retention in treatment. Data on these baseline variables were collected after clients made initial contact with their community mental health provider. Clients were observed for up to 870 days. RESULTS: By the end of the observation period, the assertive community treatment teams retained 68 percent of their clients, compared with 43 percent in usual care. In both types of treatment, clients were at greatest risk of dropping out of services during the first nine months. The risk of dropout was associated with the type of treatment. Usual-care clients were more than twice as likely as assertive community treatment clients to drop out for reasons related to dissatisfaction with treatment. Each additional night homeless during the six months before enrollment in the study resulted in a 14 percent increase in the probability of dropout. CONCLUSIONS: Assertive community treatment clearly demonstrated a greater ability than usual care services to engage and retain clients in community mental health care.