Model predicts functional and financial vulnerability in older adults
Penalized regression achieved an AUROC of 0.728 for dual vulnerability and 0.755 for death, against 0.629 for catastrophic health spending alone.
Research Square
In older Chinese adults, researchers evaluated transitions between functional impairment, catastrophic health spending, and death. Using 12,866 transitions from 7,353 participants aged 60 and older in the China Health and Retirement Longitudinal Study between 2011 and 2018, the authors developed a multistate penalized multinomial regression model to forecast joint vulnerability.
The model achieved a macro-AUROC of 0.722 and a Brier score of 0.570 in a 2015–2018 evaluation set of 5,175 participants. It showed an AUROC of 0.728 for dual vulnerability—defined as basic activity difficulty combined with catastrophic health spending—and 0.755 for death. An alternative XGBoost model did not improve macro-AUROC. In an exploratory 2018–2020 validation, the highest-risk 20% of participants accounted for 53.5% of deaths and 46.1% of combined dual-vulnerability or death events.
Why it matters
Functional decline during aging can increase care needs while high out-of-pocket costs limit rehabilitation and clinical care. Forecasting transitions into combined physical and financial vulnerability may help identify individuals who need coordinated social and medical interventions.
Caveats
The analysis relies on observational survey data, and isolated catastrophic spending remained difficult to predict without detailed reimbursement, local cost, or long-term care data. This work is a preprint and has not yet completed peer review.
The paper
Guangdong Work Injury Rehabilitation Hospital
Research Square · 29 Sep 2026 · Preprint, not peer-reviewed