The risk of suicide associated with different psychiatric diagnoses was estimated in 80,970 inpatients in Stockholm County (population 1.6 million). All patients discharged with at least one psychiatric diagnosis between 1973 and 1986 were followed by linkage with the cause-of-death registry through 1987. There were 1,115 definite suicides and 467 undetermined suicides among these during the 15-year follow-up. When 12 diagnostic categories were entered in a proportional hazards model, the highest relative risk (RR) of definite suicide, controlling for sex and age, was noted for affective disorders (RR 2.82), followed by unspecified psychoses (RR 2.69), paranoid psychoses (RR 2.60), addiction to prescription drugs (RR 2.38), neuroses and reactive psychoses (RR 1.96), and schizophrenia (RR 1.64). Alcoholism, personality disorders, organic psychoses, and street drug addiction did not have significantly increased risks of suicide. Male sex increased the risk for definite suicide by 1.56, while the risk was somewhat higher among the young. Having more than one diagnosis increased the relative risk by 1.42. When undetermined suicides were included in the analysis, to alcoholism and street drug abuse were attributed significantly increased risks of suicide, probably owing to the greater difficulty of verifying such cases. We conclude that several psychiatric disorders were conductive to suicide, but that the risk did not vary much with the type of diagnosis. Further studies of confounders are needed, such as the reasons for being admitted to inpatient care, and the impact of somatic and psychiatric comorbidity.