Background: Adolescence is a key life period for the development of depression. Predicting the development of depression in adolescence through detecting specific early symptoms may aid in the development of timely screening and intervention programmes. Methods: We administered the Short Mood and Feelings Questionnaire (SMFQ) to 5769 American and Australian students aged 10 to 15 years, at two time points, separated by 12 months. We attempted to predict high levels of depression symptoms at 12 months from symptoms at baseline, using statistical approaches based upon the quality, as well as the quantity, of depression symptoms present. These approaches included classification and regression trees (CART) and logistic regression. Results: A classification tree employing four SMFQ items, such as feelings of self-hatred and of being unloved, performed almost as well as all 13 SMFQ items at predicting subsequent depression symptomatology. Limitations: Depression was measured using a self-report instrument, rather than a criterion standard diagnostic interview. Conclusion: Further validation on other populations of adolescents is required: however the results suggest that several symptoms of depression, especially feelings of self-hatred, and being unloved, are associated with increased levels of self-reported depression at 12 months post baseline. Although screening for depression can be problematic, symptoms such as the ones above should be considered for inclusion in screening tests for adolescents.
- Classification trees