A total of 1095 adolescent psychiatric in-patients were followed up 15-33 years after hospitalization by record linkage to the National Registry of Causes of Death. On the basis of hospital records all patients were rediagnosed according to DSM-IV and scored on data postulated to be predictors of early death. The factors were investigated by Kaplan-Meyer survival analysis and Cox regression. Cox regression showed that male sex (relative risk (RR)=2.7, 95% confidence interval (CI)=1.9-3.9), psychoactive substance use disorder (RR=2.9, CI=2.0-4.0), short hospital stay (RR=3.1, CI=1.9-5.5) and poor impulse control (RR=1.6, CI=1.1-2.2) remained strong and independent predictors of death. Patients with psychoactive substance use disorder had different predictors of death (male sex, short hospital stay, poor impulse control, poor parental relationship, and low socio-economic status) than did patients without this disorder (male sex, somatic disorder at hospitalization, more than one hospitalization). In our patient population it seemed possible to identify a subgroup of adolescent psychiatric in-patients with an extremely high mortality (about 40%), namely males with psychoactive substance use disorder and poor impulse control.