摘要
目的 了解2009-2018年中国老年人关节炎发生的影响因素并基于机器学习构建关节炎发生风险的预测模型,为老年人关节炎的防治提供科学依据。方法 选取中国老年健康与家庭幸福调查(CLHLS-HF)中2009年未患关节炎且能连续随访至2018年的65岁及以上老年人为研究对象,收集其一般情况、健康状况、生活习惯及心理状况等资料。采用SPSS 25.0进行统计分析,使用多因素logistic回归模型分析老年人关节炎发生的影响因素。采用logistic回归(LR)、支持向量机(SVM)、随机森林(RF)、极限梯度提升(XGBoost)、深度神经网络(DNN)和卷积神经网络(CNN)6种机器学习方法构建关节炎发生风险的预测模型,对模型性能进行评价。结果 共纳入1 955名老年人,平均年龄为(75.1±8.3)岁,其中女性占51.2%(1 001/1 955)。2009-2018年,老年人关节炎发病率为16.9%(331/1 955)。多因素logistic回归分析结果显示,女性(OR=1.681,95%CI:1.314~2.151)、65~79岁(与≥80岁相比,OR=1.429,95%CI:1.075~1.898)、医疗服务不可及(OR=1.811,95%CI:1.167~2.813)、高血压(OR=1.379,95%CI:1.041~1.825)、糖尿病(OR=1.904,95%CI:1.061~3.418)、基础性日常生活自理能力差(OR=2.112,95%CI:1.096~4.070)与老年人关节炎高风险相关(P<0.05,P<0.01)。关节炎风险预测模型中,XGBoost具有最高的灵敏度(0.737)、较好的阴性预测值(0.880)以及较佳的校准度(0.165),展现了较好的综合性能。结论 应重点关注女性、年龄65~79岁、医疗服务不可及、高血压、糖尿病及基础性日常生活自理能力差的老年人关节炎的防治。机器学习模型在关节炎早期筛查中有着一定的应用价值。
Objective To understand the influencing factors of arthritis occurrence,build a predictive model for the risk of arthritis development based on machine learning from 2009 to 2018,and provide the scientific basis for the prevention and treatment of arthritis in elderly.Methods Elderly(≥65 years old)without arthritis following up continuously from 2009 to 2018 in the Chinese Longitudinal Healthy Longevity and Happy Family Study(CLHLS-HF)were selected as the subjects.The investigation was performed with collecting the data of general condition,health,custom habit and psychological status.Multivariate logistic regression model was used to analyze the influencing factors of arthritis occurrence in elderly.Six machine learning methods:logistic regression(LR),support vector machine(SVM),random forest(RF),extreme gradient boosting(XGBoost),deep neural network(DNN)and convolutional neural network(CNN)were used to build the prediction models of arthritis risk,and model performance was evaluated.The used software was SPSS 25.0.Results A total of 1955 elderly with mean age(75.1±8.3)years old(females were 1001/1955,51.2%)were included.From 2009 to 2018,the incidence rate of arthritis in elderly was 16.9%(331/1955).Logistic regression analysis showed that female(OR=1.681,95%CI:1.314-2.151),aged 65-79 years old(compared with≥80 years old,OR=1.429,95%CI:1.075-1.898),inaccessibility of medical services(OR=1.811,95%CI:1.167-2.813),hypertension(OR=1.379,95%CI:1.041-1.825),diabetes(OR=1.904,95%CI:1.061-3.418),and poor basic activities of daily living(BADL,OR=2.112,95%CI:1.096-4.070)were associated with a high risk of arthritis in elderly(P<0.05).Among the arthritis risk prediction models,XGBoost model demonstrated the highest sensitivity(0.737),better negative predictive value(0.880),and better calibration(0.615),demonstrating the best overall performance.Conclusion Female,lower age(65-79 years old),inaccessibility of medical services,hypertension,diabetes,and poor BADL should be focused on.Machine learning models have applications in early arthritis screening.
作者
黎晨
陈晓东
何凌骁
LI Chen;CHEN Xiaodong;HE Lingxiao(School of Public Health,Xiamen University,Xiamen,Fujian Province 361000,China)
出处
《中国慢性病预防与控制》
北大核心
2025年第4期285-289,295,共6页
Chinese Journal of Prevention and Control of Chronic Diseases
基金
中央高校基本科研业务费(20720220061)
厦门市自然科学基金项目(3502Z202471019)。
关键词
关节炎
影响因素
机器学习
Arthritis
Risk factors
Machine learning
