Validation of CBCL depression scores of adolescents in three independent datasets

Abstract

Background

Depression is common, burdensome, and is frequently first diagnosed in adolescents. The popular Adolescent Brain Cognitive Development dataset (ABCD) provides an attractive opportunity to research depression in adolescents. The only continuous measure of depression, as defined by DSM-5, in ABCD is the Child Behavior Checklist's DSM-5-Oriented Affective Problems scale (CBCL-Aff). We validated CBCL-Aff in the ABCD data and confirmed our results on two independent datasets: the Healthy Brain Network (HBN) and the Brazilian High Risk Cohort Study (BHRC).

Methods

We tested Sensitivity, Specificity and Strict Specificity. Our participants were aged 8–11 years (ABCD) or 8–12 years (HBN, BHRC). Sample size was 183–1189 participants, depending on the analysis. Sample sizes and positive case ratios were established from power estimations. We evaluated goodness of prediction with AUCROCs.

Results

In ABCD with parent-report diagnoses, CBCL-Aff had the AUCROC value of 0.95 in the Sensitivity and 0.87 in the Specificity analysis. In ABCD with child-report diagnoses, CBCL-Aff had the AUCROC of 0.62 (Sensitivity), 0.48 (Specificity) and 0.46 (Strict Specificity). In HBN and BHRC, CBCL-Aff successfully predicted clinician-report diagnoses (HBN Sensitivity: AUCROC = 0.86, Specificity: AUCROC = 0.71; BHRC Sensitivity: AUCROC = 0.90, Specificity: AUCROC = 0.80, Strict Specificity: AUCROC = 0.78).

Conclusions

We validated CBCL-Aff as a measure of depression in adolescents aged 8–11 years and we recommend its use with the following limitation: as parents and children disagreed on the child's symptoms, we discuss implications of using a parent-report only measure of child depression.