Abstract
PURPOSE: To investigate the predictive value of Traditional Chinese Medicine Constitution (TCMC) in cognitive frailty (CF) among older adults and explore its potential influencing pathways. PATIENTS AND METHODS: From 2021 to 2023, a total of 905 older participants were recruited from three geographic regions in China: Southwest (Sichuan), East (Shanghai), and North (Beijing). A multidimensional survey (including TCMC assessment) was conducted. Prediction models were developed using nomogram and C5.0 decision tree algorithms. Internal and external validations were performed. The KHB method was applied for mediation analysis. RESULTS: Logistic regression identified Qi-stagnation constitution (QSC) and Qi-deficiency constitution (QDC) as important risk factors for CF (P < 0.01). Both the C5.0 decision tree model and Nomogram model based on TCMC demonstrated strong predictive performance (AUC=0.824 and 0.812, respectively). External validation indicated superior extrapolability of the C5.0 model (AUC=0.810 vs 0.772). Mediation analysis revealed that sleep quality partially mediated the association between QSC and CF (P < 0.05), with a mediation proportion of 22.7%. CONCLUSION: QSC and QDC were identified as modifiable risk factors for CF. Prediction models based on TCMC demonstrated strong predictive performance and generalizability. Furthermore, QSC may worsen CF progression through its detrimental effects on sleep quality, identifying its clinical applicability as both a risk stratification factor and a prevention focus for CF.