中国药物经济学2025，Vol.20Issue(6)：32-37,6.DOI:10.12010/j.issn.1673-5846.2025.06.005
慢性病共病风险预测和健康管理——基于体检队列
Chronic Comorbidity Risk Prediction and Health Management:based on A Physical Examination Cohort
摘要
Abstract
Objective This study aims to predict chronic disease and comorbidity risk factors using physical examination cohort data and to compare the performance of different models,so as to provide evidence for health management of prevention of chronic comorbidity.Methods A total of 15 007 people were included in the 2018-2021 medical examination cohort of Shenzhen Luohu,and the risk prediction of hypertension,diabetes,dyslipidemia and the four combined forms of the above diseases was conducted by constructing Logistic regression and decision tree.The predictive performance for seven disease groups—hypertension,diabetes,dyslipidemia,"hypertension+diabetes,""hypertension+dyslipidemia,""diabetes+dyslipidemia,"and"hypertension+diabetes+dyslipidemia"—was comprehensively evaluated by plotting receiver operating characteristic(ROC)curves and conducting Hosmer-Lemeshow tests.Results In the health examination cohort,significant disparities were observed in the prevalence of chronic diseases,with dyslipidemia demonstrating the highest prevalence rate(29.25%),followed by hypertension(16.50%)and diabetes mellitus(6.10%).Regarding multimorbidity patterns,the"hypertension+dyslipidemia"combination exhibited the highest co-prevalence(7.10%).Both"age"and"BMI"showed consistently positive associations across all disease groups,while hepatic/renal function indicators including ALP,ALT,and BUN were positively correlated with the onset of chronic disease multimorbidity.Logistic regression model showed better prediction performance than decision tree model in most scenarios.Taking"hypertension"as an example,AUCLogistic=0.758,AUCDecision tree=0.745,and"Hypertension&Diabetes&Dyslipidemia"comorbidity combination as an example,AUCLogistic=0.811,AUCDecision tree=0.651.Conclusion Compared with a single chronic disease,The hepatic and renal function biomarkers(including ALP,ALT,and BUN)are more closely related to the incidence of chronic disease comorbidity.It is suggested that in the prevention and management of chronic disease comorbidity,it is also necessary to pay attention to the health status of liver and kidney.In addition,compared with the decision tree model,Logistic regression model has better performance in predicting chronic comorbidity,but it still has certain limitations.It is suggested to combine the regression model with other machine learning algorithms such as support vector machine and random forest to improve the prediction accuracy in the future.关键词
慢性病共病/风险预测/Logistic回归模型/决策树模型Key words
Chronic comorbidity/Risk prediction/Logistic regression model/Decision tree model分类
医药卫生引用本文复制引用
崔秀娟,汤雅倩,赵明烨,宫芳芳,唐文熙..慢性病共病风险预测和健康管理——基于体检队列[J].中国药物经济学,2025,20(6):32-37,6.基金项目
国家自然科学基金面上项目(72174207) （72174207）
国家自然科学基金面上项目(72374214) （72374214）
国家医疗保障局2023年度课题(2023001) （2023001）
