【摘要】 目的 分析脑卒中患病率和影响因素，使用人工神经网络（artificial neural network, ANN）构建高危人群的脑卒中预测模型，为预防脑卒中提供依据。方法 2023年3—8月收集脑卒中流行病学资料，采用χ²检验、logistic回归分析其流行特征；使用ANN构建预测模型，采用受试者操作特征曲线的曲线下面积（area under the curve, AUC）等指标评价预测效能。结果 >50岁人群脑卒中患病率为8.0%,患病率随人群的年龄增长而升高(χ_趋势²=17.763,P<0.001)。吸烟（OR=2.566,95%CI:1.306～5.041）、二手烟（OR=2.353,95%CI:1.224～4.523）、超重（OR=2.644,95%CI:1.222～5.724）、高血压病史（OR=2.211,95%CI:1.142～4.283）、脑卒中家族史（OR=5.221,95%CI:2.616～10.417）、CO中毒（OR=4.093,95%CI:1.312～12.772）、年龄增长（OR=1.056,95%CI:1.008～1.106）、亚硝酸盐阳性（OR=5.469,95%CI:1.892～15.806）、低密度脂蛋白升高（OR=4.942,95%CI:1.676～14.576）、阻塞性睡眠呼吸暂停低通气综合征（OR=3.162,95%CI:1.198～8.343）、房颤或瓣膜性心脏病（OR=7.538,95%CI:2.878～19.806）、摄入腌制或咸辣食品（OR=3.356,95%CI:1.429～7.879）为脑卒中发病的主要危险因素；大专/本科（OR=0.236,95%CI:0.080～0.694）和摄入红肉≤350g/W（OR=0.400,95%CI:0.185～0.863）为脑卒中发病的保护因素。ANN模型训练集的AUC为0.933,准确率为93.6%,灵敏度为74.5%,特异度为97.9%,约登指数为72.4%。结论 脑卒中发病危险因素较多，居民应养成良好生活习惯，相关部门应做好筛查及干预工作；ANN模型预测效能较好，利于早期发现脑卒中高风险人群。更多还原

【Abstract】 Objective To analyze the prevalence and influencing factors of stroke, and to construct a forecasting model of stroke in high-risk groups using the artificial neural network（ANN） to provide a basis for the prevention of stroke.Methods Epidemiological data of stroke were collected from March to August 2023,and epidemiological characteristics were analyzed using the Chi-square test and logistic regression.The forecasting model was constructed using ANN,and the predictive efficacy was evaluated using indicators of the area under the curve（AUC） of receiver operator characteristic curve etcetera.Results The prevalence of stroke in people older than 50 years old was 8.0%,and the prevalence increased with the age of the population(χ²_trend=17.763,P<0.001).Smoking（OR=2.566,95%CI: 1.306-5.041）,secondhand smoke（OR=2.353,95%CI:1.224-4.523）,overweight（OR=2.644,95%CI:1.222-5.724）,history of hypertension（OR=2.211,95%CI:1.142-4.283）,stroke family history（OR=5.221,95%CI:2.616-10.417）,CO poisoning（OR=4.093,95%CI:1.312-12.772）,increased age（OR=1.056,95%CI:1.008-1.106）,nitrite positivity（OR=5.469,95%CI:1.892-15.806）,elevated low-density lipoprotein（OR=4.942,95%CI:1.676-14.576）,obstructive sleep apnea hypopnea syndrome（OR=3.162,95%CI:1.198-8.343）,atrial fibrillation or valvular heart disease（OR=7.538,95%CI:2.878-19.806）,intake of curing or salty and spicy food（OR=3.356,95%CI:1.429-7.879） were the major risk factors for stroke onset.Junior college or undergraduate education（OR=0.236,95%CI:0.080-0.694） and intake of red meat ≤350g/W（OR=0.400,95%CI:0.185-0.863） were protective factors for stroke onset.The AUC of the training set of the ANN model was 0.933,with the accuracy of 93.6%,the sensitivity of 74.5%,the specificity of 97.9%,and the Youden index of 72.4%.Conclusion There are many risk factors for stroke onset.Residents should develop good life habits.The relevant departments should do a good job in screening and intervention.The ANN model has good predictive efficacy of forecasting and facilitates early detection of people at high risk of stroke.更多还原