Risk Prediction of Mild Cognitive Impairment Conversion based on Multivariate Bayesian Joint Model
10.11783/j.issn.1002-3674.2025.05.014
- VernacularTitle:基于多变量贝叶斯联合模型的轻度认知障碍转归风险预测
- Author:
Durong CHEN
1
;
Hongjuan HAN
;
Shuting XUE
Author Information
1. 山西医科大学公共卫生学院卫生统计教研室(030001)
- Publication Type:Journal Article
- Keywords:
Bayesian joint model;
Longitudinal data;
Survival data;
Mild cognitive impairment
- From:
Chinese Journal of Health Statistics
2025;42(5):721-726
- CountryChina
- Language:Chinese
-
Abstract:
Objective The Bayesian joint model was applied to predict conversion from mild cognitive impairment(MCI)to Alzheimer's disease(AD),and to compare time effect in longitudinal outcomes,and functional forms for the longitudinal outcomes that are included in the survival model.Methods The longitudinal sub-model used generalized linear mixed models to model the trajectory of longitudinal neuropsychological tests changes pattern,and the survival sub-model adopted a Cox proportional risk model.The longitudinal sub-model utilized two forms of effects,t and t2,and the link structures of longitudinal sub-model and survival sub-model adopted three functional forms(updated value,slope,and area).The data were divided into training and validation sets according to 7∶3,the model parameters were estimated based on Bayesian algorithm,and the dynamic AUC was used to evaluate model prediction performance.Results There were 374(54.20％)of 690 patients with MCI progressed to AD during the follow-up period.In training set,the results of Bayesian joint model with a time effect of t and area effect of three longitudinal neuropsychological tests were optimal,with a mean dynamic AUC of 0.8336.The APOEε4 gene and a low functional activities questionary were risk factors for MCI conversion to AD.The dynamic AUCs of the joint model with the t time effects and area effect for longitudinal data were above 0.75 in the validation sets.Conclusion The application of multivariate Bayesian joint model to MCI risk prediction may provide a theoretical basis for individualized interventions for cognitive impairment and inform statistical methods for survival modeling of chronic diseases.
