Dementia Rehabilitation Needs and Risk Prediction in Chinese Middle-Aged and Older Adults: Insights From GBD-WHO and CHARLS

AI Summary

Dementia rehabilitation needs in China rose 3.6-fold to 15.19 million cases and 3.7-fold YLDs to 3.33 million, trend projected to 2030.
Distributed Random Forest and gradient boosting machine models yielded best prediction (test AUCs 0.728 and 0.712), enabling individual-level risk stratification.
Top predictors include older age, lower education, pain, self-rated health expectations, and blood biomarkers, supporting integrated population and individual rehabilitation planning.

TL;DR (AI Generated)

Quantifying population-level rehabilitation needs and developing individual-level risk stratification tools are essential for targeted policy formulation and efficient resource allocation.

Using the GBD-WHO Rehabilitation Database (1990-2019), we estimated and forecasted dementia rehabilitation needs among Chinese adults aged 45 years and older.

The developed ML stratification model provides a practical tool for population-level risk assessment, highlighting the urgent need to integrate population-level forecasting with individual-level stratification in national dementia rehabilitation strategies.

Basic summary

Am J Geriatr Psychiatry. 2026 Jul 31:S1064-7481(26)00484-7. doi: 10.1016/j.jagp.2026.07.016. Online ahead of print.

ABSTRACT

OBJECTIVE: Dementia poses a growing public health challenge in China, driving increasing demand for rehabilitation. Quantifying population-level rehabilitation needs and developing individual-level risk stratification tools are essential for targeted policy formulation and efficient resource allocation.

METHODS: This two-part study integrated population- and individual-level analyses. Using the GBD-WHO Rehabilitation Database (1990-2019), we estimated and forecasted dementia rehabilitation needs among Chinese adults aged 45 years and older. Using the China Health and Retirement Longitudinal Study cohort (n = 10,146), we developed and validated machine learning (ML) models to stratify individual dementia risk and conducted variable importance analysis to identify key predictors.

RESULTS: From 1990 to 2019, the number of prevalent dementia cases requiring rehabilitation in China increased 3.6-fold, reaching 15.19 million; while years lived with disability rose 3.7-fold, reaching 3.33 million. The age-standardized prevalence rate increased from 2,447.48 to 2,942.26 per 100,000 [AAPC: 0.63 (95% CI: 0.61-0.66)], and the age-standardized YLDs rate rose from 538.77 to 654.25 per 100,000 [AAPC: 0.67 (95% CI: 0.64-0.69)]. This upward trajectory is projected to continue through 2030. Among the developed ML models, Distributed Random Forest (DRF; test AUC = 0.728) and the gradient boosting machine (GBM; test AUC = 0.712) demonstrated the best overall predictive performance. Key predictors identified included older age, lower educational attainment, pain, self-rated health expectations, and several blood biomarkers.

CONCLUSION: Dementia rehabilitation needs in China are both substantial and rapidly increasing. The developed ML stratification model provides a practical tool for population-level risk assessment, highlighting the urgent need to integrate population-level forecasting with individual-level stratification in national dementia rehabilitation strategies.

PMID:42624699 | DOI:10.1016/j.jagp.2026.07.016

Document this CPD