Predicting adolescent mental health service use in a prospective record-linkage study.

OBJECTIVE To test to what extent demographic variables and parent-reported problems and competencies in early adolescence predict incidence of mental health service use across a 5-year period in the general population. METHOD Data on parent-reported problem behavior (Child Behavior Checklist [CBCL]), gathered on 2,496 Dutch young adolescents, aged 10 to 12 years in 1989-1990, were linked to psychiatric case register data over the years 1990-1994. Cox proportional hazards models were used to predict the incidence of using mental health services from problem behavior, competencies, and demographic covariates. RESULTS Almost all CBCL problem scales predicted the incidence of mental health service use significantly. The effect was constant over time for most scales, but it decreased over time for others. CBCL Activities, gender, and one-parent family were significant predictors after accounting for the problem levels. All effects were similar for boys and girls. CONCLUSIONS In many cases, there is a remarkable delay between the awareness of the adolescent's problems and seeking and/or receiving professional help. Longitudinal studies that use continuous information on service use yield a more comprehensive picture of the utilization of mental health services than studies that aggregate the information over the follow-up period.