卒中后残疾风险预测模型的构建与验证

Construction and Validation of A Risk Prediction Model for Disability after Stroke

目的:构建风险预测模型预测脑卒中患者残疾的发生。方法:于电子病历系统中调取499例出院诊断为脑卒中患者的病历资料,采用计算机软件随机抽取373例作为建模组,剩余126例作为验证组。建模组分为非残疾组(n=153)和残疾组(n=220),利用logistic回归分析构建风险预测模型。通过预测值的符合程度和辨别程度检验对模型进行验证。结果:居住地(X_(1))、脑卒中类型(X_(2)=缺血,X_(3)=出血)、卒中复发(X_(4))和GCS评分(X_(5))均是独立危险因素。风险预测模型为logit(P)=-2.124+0.590X_(1)+1.813X_(2)+2.372X_(3)+0.535X_(4)+0.703X_(5)。H-L检验P=0.894,C-统计量为0.716,约登指数为0.303,敏感度为0.486,特异度为0.817,实际应用的正确率为83.3%。结论:本研究构建的预测模型效果良好,可为临床筛选残疾的高危患者提供参考。

Objective:To construct a risk prediction model to predict the occurrence of of disability after stroke.Method:The medical records of 499 discharged patients diagnosed as stroke were obtained in the electronic medical record system,373 cases were randomly selected as modeling group and 126 cases were selected as verification group by computer software.The modeling group was divided into non disabled group(n=153)and disabled group(n=220),using logistic regression analysis to build risk prediction model.The model was verified by the consistency and discrimination of the predicted values.Result:Residence(X_(1)),stroke type(ischemia=X_(2),hemorrhage=X_(3)),stroke recurrence(X_(4)),and GCS score(X_(5))were independent risk factors.The risk prediction model constructed was logit(P)=-2.124+0.590X_(1)+1.813X_(2)+2.372X_(3)+0.535X_(4)+0.703X_(5).H-L test P=0.894,C-statistic was 0.716,Jorden index was 0.303,sensitivity was 0.486,specificity was 0.817,and the accuracy of practical application was 83.3%.Conclusion:The predictive model constructed in this study is effective and can provide reference for clinical screening of high-risk patients with post-stroke disability.