目的 基于Logistic回归和人工神经网络构建老年脑卒中患者衰弱的风险预测模型，并评价模型预测效能，为早期识别并预防老年脑卒中患者衰弱的发生提供依据。 方法 2021年3月至2022年5月，采用便利抽样方法选取锦州市某医院就诊的老年脑卒中患者532名为研究对象。通过问卷调查收集资料，筛选患者发生衰弱的独立影响因素，用R软件绘制多因素Logistic回归模型的列线图，借助神经网络中的多层感知器构建神经网络预测模型，采用受试者工作特征曲线（receiver operating characteristic curve,ROC）评价模型预测效能。 结果 老年脑卒中患者在年龄、独居、吸烟、体育锻炼、高血压、糖尿病、首发脑卒中、睡眠障碍、跌倒史、工具性日常生活能力上的差异均有统计学意义（均P＜0.05）。年龄≥80岁、睡眠障碍、工具性日常生活能力受损、跌倒史、独居为老年脑卒中患者发生衰弱的独立风险因素，体育锻炼为保护因素；建模组列线图和神经网络预测模型ROC曲线下面积（area under curve，AUC）分别为 0.908、0.904。 结论 构建的老年脑卒中患者衰弱风险预测模型预测效能较好，有利于医护人员早期发现发生衰弱的高风险人群。

Objective To establish a prediction model for frailty in elderly stroke patients based on logistic regression and artificial neural network,to evaluate the predictive efficacy,and to provide basis for early identification and prevention of fraility in senile stroke patients. Methods Using the convenience sampling method,532 elderly stroke patients attending the hospital of Jinzhou were selected as the study subjects from March 2021 to May 2022.Data were collected through questionnaire survey,the independent risk factors for patient frailty were screened,the nomogram of multi-factor Logistic regression model with R software were drawn,and the neural network prediction model with the help of multi-layer perceptron in the neural network was constructed.The receiver operating characteristic curve(ROC) was used to verify the predictive effect of the models. Results There were statistically significant differences in age,living alone,smoking,physical exercise,hypertension,diabetes,first stroke,sleep disorders,fall history and instrumental daily living ability in elderly stroke patients (all P＜0.05).Age ≥80 years old,sleep disorder,impaired instrumental daily living ability,history of falls,and living alone were independent risk factors for asthenia in senile stroke patients,and physical exercise was a protective factor.The area under curve (AUC) of the modeling group line diagram and the neural network prediction model were 0.908 and 0.904,respectively. Conclusions The established model for predicting frailty risk in elderly stroke patients has good predictive efficacy,which is conducive to early detection of high risk groups of frailty by medical staff .

R473.74

辽宁省自然科学基金（2019-ZD-0802）