Does problem behaviour affect attrition from a cohort study on adolescent mental health?

BACKGROUND: Most surveys are plagued by significant numbers of non-respondents. In psychiatric epidemiology, differences in levels of psychopathology are often assumed between respondents and non-respondents. However, studies on attrition are rare, especially in child and adolescent populations. The present study sought to estimate the association of depression, anti-social behaviour with attrition from a cohort study among adolescents. METHODS: The adolescent mental health cohort study (AMHC) is a prospective follow-up study on prevalence and determinants of mental health problems and risk behaviour among Finnish adolescents at two study sites. Subjects of the final baseline sample (n = 3278) were reached for a 2-year follow-up. Differences in data collecting methods between the study sites yielded different response rates (78% vs. 51%). Chi-square statistics and logistic regression models were computed to estimate the effect of two types of problem behaviour on attrition. RESULTS: Depression at baseline was associated with higher probability of attrition at follow-up. School performance was a stronger predictor of attrition than problem behaviour. The models predicted non-response in the study site with mainly school-based surveys but not in the study site with mainly postal surveys. CONCLUSION: Internalizing problem behaviour may be underrepresented in adolescent mental health surveys. School performance is strongly associated with attrition. Hence, special attention should be paid to designing questionnaires targeting adolescents.