Phone features complement self-reports for questionnaire-defined risk labels
Source check: Source checked. Not independent scientific confirmation.
Finding: In selected Korean smartphone participants, random-forest models combining passive features with daily self-reports achieved cross-validated ROC areas of0.82 for depression-risk labels and0.86 for anxiety-risk labels. These are weekly questionnaire-defined states, with overlapping symptom-report windows, not clinical diagnoses or future-onset forecasts.
Technology: Random forest; gradient boosting; XGBoost
Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- Fitted tree ensembles combine mobility/activity summaries and brief symptom reports to classify weekly PHQ-9/GAD-7 threshold labels.
- Test and result
- Of455 recruited adults,277 depression and275 anxiety participants met coverage requirements. Repeated observations remained with their participant in fivefold evaluation; fourteen-day observation windows preceded weekly questionnaire dates.
- Limitation / still unknown
- Coverage thresholds and window length were selected using the same cross-validation framework, without an established untouched configuration test. Active daily mood/anxiety reports overlap conceptually and temporally with questionnaire references; weekly PHQ-9 repeats a two-week recall window. Many participants were excluded for coverage, and there was no external or temporal validation. Sleep features are phone-motion proxies. Average-precision numbers are not used here without independently resolving positive-class prevalence/encoding details. Prospective collection does not establish a prospectively frozen model or benefit.
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/PMC13068600/ · https://doi.org/10.1016/j.invent.2026.100934
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: Actual original relevant clinical/computational Methods, reference labels, evaluation Results and limitations read.
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.