Answers are grounded in this paper and cite their sources.
This review examined machine learning applications to adolescent sports injury prediction. Six databases were searched on 25 May 2026 for publications from January 2021 onward involving participants aged 10–19 years. Two reviewers independently screened studies. Methodological appraisal was conducted using PROBAST+AI, design-specific JBI checklists, or QualSyst, with one reviewer conducting the assessments and the other verifying them. Findings were synthesized narratively. Thirty reports representing 29 studies covered injury prediction, clinical and surrogate outcomes, biomechanical classification, and monitoring or exploratory applications. Thirteen reports provided ROC-AUC values of 0.58–0.94; sensitivity/recall for injury occurrence or recurrence ranged from 25.0% to 94.3%. Features included maturation, anthropometry, clinical history, neuromuscular performance, biomechanics, physiological and metabolic markers, training load, and psychological and sleep-related measures. Among 14 PROBAST+AI reports, performance-related risk of bias was high in 4 and unclear in 10. JBI assessments identified methodological concerns; nine QualSyst scores ranged from 0.50 to 0.80, with one report unscored owing to indeterminate criterion applicability. Evidence is insufficient to establish the generalizability of machine-learning performance because of high or unclear risk of bias and limited external validation. Addressing these limitations may support the future use of machine learning in sports injury risk prediction among adolescents. The review was prospectively registered with OSF (PYE97) and received no external funding.
No takes yet. Share an insight, caveat, or question.
Yu-Wei et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: