Development and validation of a risk stratification model for mild cognitive impairment among older adults with possible sarcopenia integrating traditional Chinese medicine constitution: a multicenter cross-sectional study
Abstract
BackgroundPossible sarcopenia is associated with cognitive impairment among older adults, while mild cognitive impairment (MCI) represents an important window for early identification. Existing MCI assessment approaches may depend on neuroimaging, genetic testing, or biochemical biomarkers, limiting their use in large-scale screening. Traditional Chinese medicine (TCM) constitution may provide accessible information on individual physiological and functional characteristics, but it has rarely been incorporated into MCI risk stratification.ObjectiveTo develop and validate a machine learning-based risk stratification model integrating TCM constitution and conventional health-related indicators for MCI among older adults with possible sarcopenia.MethodsThis multicenter cross-sectional study included 3,310 older adults with possible sarcopenia from ten tertiary hospitals in China and an independent single-institution external validation sample of 314 participants. The concurrent outcome was screening-defined probable MCI based on a prespecified Montreal Cognitive Assessment (MoCA) criterion. After a stratified 7:3 training–internal-validation split, data-dependent preprocessing parameters and least absolute shrinkage and selection operator (LASSO) feature selection were estimated using the training data only. Multilayer perceptron (MLP), gradient boosting decision tree (GBDT), support vector machine, K-nearest neighbors, logistic regression, and AdaBoost models were fitted using prespecified settings and evaluated using discrimination, classification, calibration, and decision-curve measures. SHapley Additive exPlanations (SHAP) were used only to describe contributions to model output.ResultsSeven features were retained: educational attainment, sex, smartphone usage, Pittsburgh Sleep Quality Index score, Patient Health Questionnaire-9 score, qi-deficiency constitution, and damp-heat constitution. The MLP showed the highest internal discrimination, with an AUC of 0.853 (95% CI: 0.827–0.878); its AUC was 0.799 (95% CI: 0.773–0.823) in the external sample. The internal and external Brier scores were 0.168 and 0.151, calibration intercepts were −0.07 and 0.18, and calibration slopes were 0.80 and 1.07, respectively. SHAP analysis ranked feature contributions to model output and was not interpreted causally.ConclusionThis cross-sectional screening framework integrated TCM constitution with conventional indicators to stratify the probability of current screening-defined probable MCI among older adults with possible sarcopenia. The single-institution external evaluation supports only limited transportability; prospective multicenter evaluation and, where needed, recalibration are required before clinical implementation.
Related
- CALLY is not associated with colorectal polyps risk: a systematic evaluation and mechanism exploration controlling for traditional factors and Traditional Chinese Medicine constitution
- Traditional Chinese medicine constitution-related body-composition phenotypes and 1-year metabolic changes: a repeated health-examination study
- Feature-token clinical transformer for 24-h landmark mortality risk stratification in critically ill patients with heart failure: development and multicenter external validation
- Development and validation of an interpretable machine learning model for risk stratification among individuals with latent tuberculosis infection: a multicenter case-control study
- Toward Precision Colorectal Cancer Screening: Integrating Clinical Risk Stratification, Molecular Biomarkers, and Artificial Intelligence
BibTeX
@article{wang2026development,
title = {{Development and validation of a risk stratification model for mild cognitive impairment among older adults with possible sarcopenia integrating traditional Chinese medicine constitution: a multicenter cross-sectional study}},
author = {Ruoli Wang and Nanping Mao and Yanyun He and Yi Wang and Wenbin Wu},
journal = {Frontiers in Public Health},
year = {2026},
month = oct,
doi = {10.3389/fpubh.2026.1980675},
url = {https://www.frontiersin.org/articles/10.3389/fpubh.2026.1980675},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Development and validation of a risk stratification model for mild cognitive impairment among older adults with possible sarcopenia integrating traditional Chinese medicine constitution: a multicenter cross-sectional study
AU - Ruoli Wang
AU - Nanping Mao
AU - Yanyun He
AU - Yi Wang
AU - Wenbin Wu
PY - 2026
DA - 2026/10/30
JO - Frontiers in Public Health
DO - 10.3389/fpubh.2026.1980675
UR - https://www.frontiersin.org/articles/10.3389/fpubh.2026.1980675
AB - BackgroundPossible sarcopenia is associated with cognitive impairment among older adults, while mild cognitive impairment (MCI) represents an important window for early identification. Existing MCI assessment approaches may depend on neuroimaging, genetic testing, or biochemical biomarkers, limiting their use in large-scale screening. Traditional Chinese medicine (TCM) constitution may provide accessible information on individual physiological and functional characteristics, but it has rarely been incorporated into MCI risk stratification.ObjectiveTo develop and validate a machine learning-based risk stratification model integrating TCM constitution and conventional health-related indicators for MCI among older adults with possible sarcopenia.MethodsThis multicenter cross-sectional study included 3,310 older adults with possible sarcopenia from ten tertiary hospitals in China and an independent single-institution external validation sample of 314 participants. The concurrent outcome was screening-defined probable MCI based on a prespecified Montreal Cognitive Assessment (MoCA) criterion. After a stratified 7:3 training–internal-validation split, data-dependent preprocessing parameters and least absolute shrinkage and selection operator (LASSO) feature selection were estimated using the training data only. Multilayer perceptron (MLP), gradient boosting decision tree (GBDT), support vector machine, K-nearest neighbors, logistic regression, and AdaBoost models were fitted using prespecified settings and evaluated using discrimination, classification, calibration, and decision-curve measures. SHapley Additive exPlanations (SHAP) were used only to describe contributions to model output.ResultsSeven features were retained: educational attainment, sex, smartphone usage, Pittsburgh Sleep Quality Index score, Patient Health Questionnaire-9 score, qi-deficiency constitution, and damp-heat constitution. The MLP showed the highest internal discrimination, with an AUC of 0.853 (95% CI: 0.827–0.878); its AUC was 0.799 (95% CI: 0.773–0.823) in the external sample. The internal and external Brier scores were 0.168 and 0.151, calibration intercepts were −0.07 and 0.18, and calibration slopes were 0.80 and 1.07, respectively. SHAP analysis ranked feature contributions to model output and was not interpreted causally.ConclusionThis cross-sectional screening framework integrated TCM constitution with conventional indicators to stratify the probability of current screening-defined probable MCI among older adults with possible sarcopenia. The single-institution external evaluation supports only limited transportability; prospective multicenter evaluation and, where needed, recalibration are required before clinical implementation.
SN - 2296-2565
ER -