Machine learning-based prediction of substance use in adolescents in 3 independent worldwide cohorts: Algorithm development and validation study (Preprint)

Journal of Medical Internet Research, Год журнала: 2025, Номер 27, С. e62805 - e62805

Опубликована: Янв. 16, 2025

Background To address gaps in global understanding of cultural and social variations, this study used a high-performance machine learning (ML) model to predict adolescent substance use across three national datasets. Objective This aims develop generalizable predictive for using multinational datasets ML. Methods The the Korea Youth Risk Behavior Web-Based Survey (KYRBS) from South (n=1,098,641) train ML models. For external validation, we (YRBS) United States (n=2,511,916) Norwegian nationwide Ungdata surveys (Ungdata) Norway (n=700,660). After developing various models, evaluated final model’s performance multiple metrics. We also assessed feature importance traditional methods further analyzed variable contributions through SHapley Additive exPlanation values. Results development analyzing data 1,098,641 KYRBS adolescents, 2,511,916 YRBS participants, 700,660 Ungdata. XGBoost was top performer on KYRBS, achieving an area under receiver operating characteristic curve (AUROC) score 80.61% (95% CI 79.63-81.59) precision 30.42 28.65-32.16) with detailed analysis sensitivity 31.30 29.47-33.20), specificity 99.16 99.12-99.20), accuracy 98.36 98.31-98.42), balanced 65.23 64.31-66.17), F1-score 30.85 29.25-32.51), precision-recall 32.14 30.34-33.95). achieved AUROC 79.30% 68.37% dataset, while validation it recorded 76.39% 12.74%. Feature value analyses identified smoking status, BMI, suicidal ideation, alcohol consumption, feelings sadness despair as key contributors risk use, status emerging most influential factor. Conclusions Based Korea, States, Norway, shows potential particularly model, predicting use. These findings provide solid basis future research exploring additional influencing factors or targeted intervention strategies.

Язык: Английский

Artificial intelligence models predicting abnormal uterine bleeding after COVID-19 vaccination

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Фев. 27, 2025

Язык: Английский