Learning exposes limits of predicting new running injuries
Source check: Source checked. Not independent scientific confirmation.
Finding: An Explainable Boosting Machine analyzed surveys and running logs from 643 marathon trainees, holding out all observations from test runners. Prediction was much weaker among runners not already modifying training: AUROC was 0.673 and precision-recall area 0.077, compared with 0.873 and 0.521 across all runner-weeks. Outcomes were self-reported training-limiting injuries; the study did not demonstrate prevention or real-time clinical benefit.
Technology: Explainable Boosting Machine; boosted generalized additive model
Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- Boosted shallow trees learn nonlinear additive relationships between prior training/symptom measures and subsequent weekly injury status.
- Test and result
- Runner-grouped 80:20 train/test evaluation of upcoming-week self-reported injury using only earlier surveys/logs.
- Limitation / still unknown
- External replication, medically confirmed injury outcomes and prevention benefit remain unestablished; persistent existing injury accounts for much overall predictive performance.
Evidence sources and annotation context
Evidence context (descriptions follow the cited sources; test type is a catalog classification):
Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13193517/
Source checked · editorial details
The source page was checked against this entry. Independent confirmation is not assessed by this label.
Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.
Review date: 21 Nov 2026. Review scope: Primary relevant model Methods, Results and limitations; exact primary identity and current primary PubMed correction/retraction relationships checked..
No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.
Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.