Chinese General Practice ›› 2025, Vol. 28 ›› Issue (22): 2776-2783.DOI: 10.12114/j.issn.1007-9572.2024.0451

• Original Research • Previous Articles    

Construction and Validation of a Risk Prediction Model for Cognitive Impairment in Community-dwelling Older Adults

  

  1. 1. School of Public Health, Shandong Second Medical University, Weifang 261053, China
    2. Collaborative Innovation Center for Prediction and Governance of Major Social Risks of "Healthy Shandong", Weifang 261053, China
    3. China Institute of Rehabilitation and Health Research, Weifang 261053, China
    4. School of Management, Shandong Second Medical University, Weifang 261053, China
    5. China ICF Research Institute, Weifang 261053, China
    6. Oxford Institute for Population Ageing, University of Oxford, Oxford OX26PR, UK
  • Received:2024-10-10 Revised:2025-01-09 Published:2025-08-05 Online:2025-06-30
  • Contact: ZHENG Wengui, JING Qi

社区老年人认知障碍风险预测模型的构建与验证研究

  

  1. 1.261053 山东省潍坊市,山东第二医科大学公共卫生学院
    2.261053 山东省潍坊市,"健康山东"重大社会风险预测与治理协同创新中心
    3.261053 山东省潍坊市,中国康复健康研究院
    4.261053 山东省潍坊市,山东第二医科大学管理学院
    5.261053 山东省潍坊市,中国ICF研究院
    6.OX26PR英国牛津郡,牛津大学牛津人口老龄研究所
  • 通讯作者: 郑文贵, 井淇
  • 作者简介:

    作者贡献:

    赵晓晴负责数据整理与分析,撰写论文初稿;郭桐桐、张欣怡负责研究的实施与可行性分析;李林虹、蔡伟芹负责数据的统计分析;张亚、高倩倩负责图表绘制;嵇丽红、董志伟负责论文的质量控制及审校;郑文贵、井淇负责论文设计与指导,对文章整体负责。

  • 基金资助:
    国家自然科学基金资助项目(72004165,72374156); 山东省高等学校"青创团队计划"团队研究课题(2022RW075); 山东省政府公派出国留学项目

Abstract:

Background

With the further aging of the population, the incidence of cognitive impairment is increasing, and there is a lack of effective treatments. The construction of an accurate risk prediction model can be used to help community healthcare workers to identify, warn and intervene with potential patients at an early stage, and to reduce the pressure on social healthcare.

Objective

This study aims to construct a prediction model for the risk of cognitive impairment in older adults in the community, analyse the influencing factors of cognitive impairment in older adults, and provide empirical references for the development of targeted interventions.

Methods

In April 2024, elderly people aged ≥60 years were selected from the China Health and Retirement Longitudinal Survey (CHARLS) 2020 database (n=7 334) , and their socio-demographic characteristics and data on their health status and behaviours, activities of daily living (ADL) , depression, and cognitive abilities were collected. They were randomly divided into a training set (n=5 133) and a validation set (n=2 201) in a ratio of 7∶3. The best predictor variables were screened using LASSO regression ten-fold cross-validation, the factors influencing cognitive impairment in older adults were analysed using Logistic regression, and nomagram were constructed, and the performance of the predicion model was assessed using the area under the curve of the subject work characteristics (ROC) curves and the analysis of the calibration curves.

Results

The detection rate of cognitive impairment in older adults was 14.48% (1 062/7 334) . LASSO regression screened nine potential predictor variables, which were age, type of residence, marital status, gender, education, exercise, society, activity of daily living, and depression. The results of multifactorial Logistic regression analysis showed that age [OR (95%CI) =1.238 (1.109-1.504) for 70-79 years old and OR (95%CI) =2.231 (1.546-3.222) for ≥80 years old using 60-69 years old as a reference] , type of residence [OR (95%CI) =2.144 (1.617-2.842) for rural using urban as a reference] , marital status [OR (95%CI) =0.691 (0.562-0.851) for no spouse, using spousal as a reference] , education [OR (95%CI) =0.209 (0.173-0.254) for primary school and below, using illiteracy as a reference, and for junior high school OR (95%CI) =0.059 (0.038-0.090) , OR (95%CI) for high school/vocational high school=0.043 (0.021-0.089) , and OR (95%CI) for college and above=0.038 (0.005-0.280) ] , and society [with no society as a reference, and OR (95%CI) with society=0.746 (0.624-0.892) ] , ability to perform ADL [OR (95%CI) =1.529 (1.171-1.997) with no impairment as a reference and OR (95%CI) =1.580 (1.319-1.891) with impairment] , and depression [OR (95%CI) =1.580 (1.319-1.891) with no depression as a reference and OR (95%CI) =1.580 (1.319-1.891) with depression] were the influencing factors of cognitive impairment (P<0.05) . Based on the seven predictor variables screened by multifactor Logistic regression analysis, a prediction model was established. The areas under the ROC curves of the prediction model in the training and validation sets were 0.821 (95%CI=0.805-0.836) and 0.839 (95%CI=0.817-0.861) , respectively; the Hosmer-Lemeshow test χ2=5.022 (P=0.755) and χ2=3.963 (P=0.860) ; calibration curves showed significant agreement between predicted and actual values.

Conclusion

In this study, a prediction model for the risk of cognitive impairment in community-dwelling older adults containing a total of seven indicators, including age, residence, and so on, was established, and the prediction model had good accuracy and differentiation, which can be used to identify the risk of developing cognitive impairment in older adults.

Key words: Cognition disorders, Aged, Prediction model, Nomograms, Root cause analysis

摘要:

背景

随着人口老龄化程度的持续加深,认知障碍发病率越来越高,但目前尚缺乏有效治疗方法,构建精准的风险预测模型可以帮助社区医护人员早期识别、预警与干预潜在患者,减轻社会医疗压力。

目的

构建社区老年人认知障碍风险的预测模型,分析老年人认知障碍的影响因素,为制定针对性的干预措施提供实证参考。

方法

于2024年4月,在中国健康与养老追踪调查(CHARLS)2020年数据库中选取≥60岁老年人为研究对象(n=7 334),收集其社会人口学特征及健康状况和行为、日常生活活动能力(ADL)、抑郁、认知功能数据。以7∶3的比例随机分为训练集(n=5 133)和验证集(n=2 201),采用LASSO回归十折交叉验证法筛选最佳预测变量,采用Logistic回归分析老年人认知障碍影响因素,并构建列线图,采用受试者工作特征(ROC)曲线下面积、校准曲线等评估预测模型的性能。

结果

老年人的认知障碍检出率为14.48%(1 062/7 334)。LASSO回归筛选出9个潜在预测变量,分别为性别、年龄、受教育程度、婚姻状况、居住地类型、运动、社交、ADL、抑郁。多因素Logistic回归分析结果显示,年龄[以60~69岁为参照,70~79岁的OR(95%CI)=1.238(1.109~1.504),≥80岁的OR(95%CI)=2.231(1.546~3.222)],受教育程度[以文盲为参照,小学及以下的OR(95%CI)=0.209(0.173~0.254),初中的OR(95%CI)=0.059(0.038~0.090),高中/职高的OR(95%CI)=0.043(0.021~0.089),大专及以上的OR(95%CI)=0.038(0.005~0.280)],婚姻状况[以有配偶为参照,无配偶的OR(95%CI)=0.691(0.562~0.851)],居住地类型[以城市为参照,农村的OR(95%CI)=2.144(1.617~2.842)],社交[以无社交为参照,有社交的OR(95%CI)=0.746(0.624~0.892)],ADL[以无障碍为参照,有障碍的OR(95%CI)=1.529(1.171~1.997)],抑郁[以无抑郁为参照,抑郁的OR(95%CI)=1.580(1.319~1.891)]是认知障碍的影响因素(P<0.05)。依据多因素Logistic回归分析筛选出的7个预测变量,建立预测模型。预测模型在训练集和验证集的ROC曲线下面积分别为0.821(95%CI=0.805~0.836)和0.839(95%CI=0.817~0.861);Hosmer-Lemeshow检验χ2=5.022(P=0.755)和χ2=3.963(P=0.860);校准曲线显示预测值和实际值之间存在显著一致性。

结论

本研究建立了包含年龄、居住地类型等共7个指标的社区老年人认知障碍风险预测模型,预测模型准确度和区分度均较好,可用于预测老年人认知障碍的发生风险。

关键词: 认知障碍, 老年人, 预测模型, 列线图, 影响因素分析