Of 1969 previous adolescent psychiatric inpatients, 1792 (91%) were traced after a mean follow-up period of 15 years. Thirty-five patients, 1.7% of the females and 2.2% of the males, had committed suicide, corresponding to a yearly suicide rate of 145/100,000 for males and 110/100,000 for females. This represents a 6-fold increase for males and a 19-fold increase for females compared with the suicide rate for 15- to 29-year-old males and females in the general population. There was seasonality in violent but not in nonviolent suicides. The patients who had committed suicide were compared with matched patients from the same sample who stayed alive. The suicide group had more depressive symptoms, more learning difficulties, poorer self esteem, were more help-rejecting, and had more immature defense mechanisms. They lacked parental support and were more often verbally abused by their parents. They had more frequently experienced serious losses in early childhood and had a higher score on enduring stressors on Axis IV in DSM-III-R. They more often came from urban areas and received poorer follow-up after discharge from hospital. Eight of these discriminating factors were combined into a predictive model for the lifetime risk of suicide in adolescent psychiatric inpatients. The model had strong predictive power, classifying 84% of the population correctly.