Development and Initial Evaluation of an Adolescent Depression Screening Tool: Integrating Factor Analytic, Network, and Clinical Evidence

Abstract Background Adolescent depression screening requires brief instruments that capture clinically relevant symptoms and acknowledge inconclusive results. This study developed and initially evaluated an ICD-11-anchored screening instrument comprising a full form and an abbreviated short form, integrating structural, network, and clinical evidence. Methods The development sample included 828 adolescents aged 13–19: 661 community respondents, 87 other clinical respondents, and 80 depression-symptom respondents. Item evaluation combined exploratory factor analysis of polychoric correlations, regularized symptom-network analysis, response distributions, symptom overlap, and clinical content review. Unweighted, provisionally weighted, and ridge-regression scoring were compared, including nested cross-validation for regression evaluation. Discrimination between the depression-symptom and each comparison group was assessed using receiver operating characteristic areas under the curve (AUCs) and paired bootstrap intervals. Uncertain intervals and common raw-score boundaries were explored. Internal consistency, classification agreement, subgroup differences, and WHO-5 correlations were examined. Results The retained 23-item full form showed four related symptom domains. A 13-item short form was selected, although its one- and two-factor solutions retained substantial misfit. McDonald’s omega was .960 for the full form and .935 for the short form. Short- and full-form AUCs were .848 and .846 against the other clinical group and .825 and .824 against the community group; paired difference intervals included zero. Weighting provided no consistent advantage across contrasts. Revised short-form bands were 0–18, 19–22, and 23–52; full-form bands were 0–31, 32–44, and 45–92. At the short-form lower boundary, sensitivity was 83.8% and specificity was 72.4% and 72.5% against the other clinical and community groups, respectively. Classification agreement was 90.2%. Sensitivity was lower among boys and particularly adolescents aged 13–14 in both forms. WHO-5 correlations were − .631 and − .649 for the short and full forms. Conclusions The proposed Adolescent Depressive Symptoms Screener offers a brief, simply scored form with promising initial discrimination. Its provisional bands support further evaluation of symptom enquiry and follow-up decisions rather than diagnosis. Development-sample optimism and subgroup limitations require independent validation before routine screening recommendations.