Abstract: Background: Disengagement from treatment is a major concern in psychiatry. This is of particular concern for those presenting for care at their first episode of psychosis (FEP). The purpose of this study is to determine the rate of disengagement from a three year FE treatment program and the predictors of disengagement.Method: We used a longitudinal cohort design. The cohort consisted of 286 FEP individuals. Measures included assessments of positive and negative symptoms, depression, substance use, premorbid and current functioning, cognition and duration of untreated psychosis. Disengagement from treatment was defined as leaving the program before the 30months.Results: At 30months after treatment, the estimated rate of disengagement from treatment was 31%. Predictors of disengagement were examined via Cox proportional hazards models which revealed that lower ratings on negative symptom scores at baseline (HR=0.946; CI=0.909–0.985), a shorter duration of untreated psychosis (HR=0.997; CI=0.994–0.999), and not having a family member involved in the program (HR=0.310; CI=0.196–0.490) contributed significantly to predicting disengagement from treatment. An examination of those who dropped out at different times revealed that those who dropped out prior to 6months had significantly greater cannabis (p<0.05) and other drug use (p<0.01).Conclusions: Engagement in early services may be helped by attending carefully to substance use to prevent early dropout, to those who may have had a short duration of untreated psychosis and to working with families to engage families in the program.