老年认知障碍风险简易预测模型

A simple prediction model for the risk of cognitive impairment in the elderly

目的建立老年认知障碍风险简易预测模型。方法于2020年8月—12月,在上海交通大学附属第一人民医院、上海交通大学医学院附属瑞金医院、首都医科大学宣武医院、首都医科大学附属复兴医院及协作的社区卫生服务中心和养老院,选取就诊的546例年龄≥60岁的老年人为研究对象。采集受试者的社会人口学资料、疾病史、家族史、衰弱表型和相关实验室指标等。采用LASSO回归模型,结合Logistic回归分析选取预测因素,绘制包含最佳预测因素的认知障碍风险预测模型诺谟图,通过校准曲线、C指数和Bootstrapping算法对模型进行验证。结果通过纳入变量分析,筛选出年龄、文化程度、直系亲属痴呆史、主观认知下降、衰弱表型、代谢综合征和低白蛋白血症共7个独立预测因素(P<0.05)。预测模型的C指数为0.875 (95%CI:0.847~0.903),bootstrapping算法相对校正后的C指数为0.858。校准曲线与理想曲线接近重合,模型具有良好的预测能力,且可增加受试者的临床获益。结论对于认知障碍高危人群,应考虑年龄、文化程度、直系亲属痴呆史、主观认知下降、衰弱状态、代谢综合征和低白蛋白血症等因素的综合作用。基于上述因素建立的预测模型,可以快速、简便评估个体认知障碍风险。

Objective To establish a simple prediction model for the risk of cognitive impairment in the elderly.Methods From August to December2020, 546 elderly people aged≥60 were selected as the research subjects in Shanghai General Hospital,Ruijin Hospital Affiliated to Shanghai Jiaotong University,Xuanwu Hospital Affiliated to Capital Medical University,Fuxing Hospital Affiliated to Capital Medical University,and their cooperative community health service centers and nursing homes. The sociodemographic data,disease history,family history,frailty phenotype and related laboratory indexes of these subjects were collected. LASSO regression model combined with Logistic regression analysis was used to select predictive factors. The Nomogram of the cognitive impairment risk prediction model including the optimal predictive factors was drawn,and the model was verified by calibration curve,C index and bootstrapping algorithm. Results Through the analysis of included variables,seven independent predictors were screened out(P< 0.05),including age,educational level,dementia history of immediate family members,subjective cognitive decline,frailty phenotype,metabolic syndrome and hypoalbuminemia. The C index of the prediction model was0.875(95% CI: 0.847-0.903),and the relative corrected C-index of the bootstrapping algorithm was0.858. The calibration curve was nearly coincident with the ideal curve,and the model had good predictive ability and could increase the clinical benefit of the subjects. Conclusion For people at high risk of cognitive impairment,the combined effects of age,education level,history of dementia of immediate family members,subjec-tive cognitive decline,frail status,metabolic syndrome and hypoalbuminemia should be considered. The prediction model based on the above factors can quickly and simply assess the risk of individual cognitive impairment.