Can the Strength and Difficulties Questionnaire predict depression among Australian adolescents? A Markov Blanket structure learning approach
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Introduction: The poor mental health of Australian adolescents is considered a public healthcrisis. Development of brief screening tools to identify adolescents at high risk of Major DepressiveDisorder (MDD) is consequently a high research and clinical priority. To addressthis gap, the current study aims to investigate the extent to which the self-report Strengthsand Difficulties Questionnaire (SDQ) accurately identifies 11-17-year-olds in the communitywho meet diagnostic criteria for MDD. Methods: Data was from the Second AustralianChild and Adolescent Survey of Mental Health and Wellbeing. The sample included 2,967adolescents aged 11 to 17 years who answered the SDQ and for whom MDD was evaluatedwith the Diagnostic Interview Schedule for Children Version IV (DISC-IV). To develop thescreening tool, seven prediction models were examined: (1) logistic regression (total score);(2) logistic regression (total score including prosocial behaviours); (3) logistic regression(subscale scores); (4) logistic regression (item scores); (5) logistic LASSO regression (itemscores); (6) Markov Blanket-based logistic regression (item scores); and (7) XGBoost (itemscores). Predictive performance was evaluated with the Area Under the Receiver OperatingCharacteristic Curve (AUROC), Area Under the Precision-Recall Curve (AUPRC),calibration, sensitivity, specificity and positive predictive value (PPV). Results: A 3-itemscreening tool (identified by the Markov Blanket-based logistic regression) performed thebest and achieved appropriate predictive performance (AUROC=0.73; AUPRC=0.16; sensitivity=67.5%; specificity=70.9%; and PPV=10.8%). Conclusion: A 3-item screening tooldisplayed appropriate predictive performance to identify MDD among Australian adolescentsaged 11 to 17 years in the community. Future studies should externally validate theproposed tool so it can be implemented in clinical practice.