Time-Dynamic AI Models to Predict Quality of Life in Patients With Breast Cancer: Development and Validation Study Using the EORTC BALANCE Cohort.
Source check: Source checked. Not independent scientific confirmation.
Finding: Time-dynamic ML models used earlier EORTC quality-of-life assessments and clinical data to predict later impairment in 6,316 patients across more than 70,000 assessment pairs. Independent NKI validation AUCs ranged 0.643–0.837, with substantial overconfidence and weaker calibration for some outcomes.
Technology: Gradient boosting and other classifiers used prior QLQ-C30 assessments, clinical data and time since assessment to predict later quality-of-life impairment.
Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- The models predicted later patient-reported quality-of-life impairment with independent-cohort discrimination, while calibration limits remained substantial.
- Test and result
- Across 6,316 patients and more than 70,000 assessment pairs, independent NKI validation AUC ranged from 0.643 to 0.837; the paper also reports substantial overconfidence and calibration weaknesses for low-prevalence symptoms.
- Limitation / still unknown
- Five cohorts were used for development and an independent NKI cohort for external validation. Calibration slopes were 0.089–0.286; the primary preprint posted 2025-08-18 reports the same core result as the 2026 journal version.
Evidence sources and annotation context
Evidence context (descriptions follow the cited sources; test type is a catalog classification):
Evidence sources: https://preprints.jmir.org/preprint/81424 · https://www.jmir.org/2026/1/e81424
Source checked · editorial details
The source page was checked against this entry. Independent confirmation is not assessed by this label.
Source type: paper. Paper: linked separately.
Review date: 22 Nov 2026. Review scope: JMIR v1 primary preprint and final full text compared for result-bearing version; Methods, Results, independent validation and calibration 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 4/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.