Wearable-data models evaluate near-term binge-eating risk
Source check: Source checked. Not independent scientific confirmation.
Finding: In a four-week observational study of 30 adults with recurrent binge eating, random-forest models used smartwatch physiology or self-reported affect to evaluate episode risk. The first three weeks trained models and the final week tested them within the same participants. Sensor-only AUROC was 0.78, with sensitivity 0.92 but specificity 0.35; macro-averaged accuracy did not significantly exceed self-report. No model met both prespecified sensitivity and specificity thresholds, and benefits from adaptive intervention remain untested.
Technology: Random forest; elastic-net feature selection
Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- Elastic-net selects sensor features on training data; RF combines person/group models with tuning inside first-three-week data and tests on last week. Chronological retrospective evaluation is not prospective model-guided intervention. Source discussion has a sensitivity/specificity wording error; numerical Results and tables support 0.92 sensitivity and 0.35 specificity.
- Test and result
- Within-participant temporal hold-out comparison of sensor, self-report and combined models.
- Limitation / still unknown
- New-participant performance, deployable specificity and benefits of model-triggered therapy remain unestablished.
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/PMC13132026/
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 full article Methods, Results and limitations plus Europe PMC bibliographic metadata; PMC full article..
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.