脑卒中类型与复发的Logistic回归预测

Logistic discriminant analysis of stroke type and recurrence

目的本研究拟构建脑卒中缺血性或出血性分型回归模型、脑卒中复发回归模型。方法2012年5月-2013年9月在河北医科大学第二医院因脑卒中入院治疗244例患者的病例,对其出院后进行随访,平均随访时间为18个月。通过IBM SPSS进行Logistic回归分析,采用向前法,获得Logistic回归方程。并用Cox&Snell R Square检验计算拟合优度,获得预测准确率。使用SMOTE算法用R语言平衡数据。结果脑卒中分型的回归模型检验拟合优度为0.634,预测准确率为86.1%;缺血性脑卒中复发预测的回归模型检验拟合优度为0.236,其灵敏度为28.6%,出血性脑卒中检验拟合优度为0.272,其灵敏度为60%;SMOTE算法处理缺血性脑卒中复发预测数据,获得的回归模型Cox&Snell R Square检验拟合优度为0.488,灵敏度为89.3%。结论缺血性与出血性脑卒中类型可以使用Logistic回归进行辅助诊断,以提高诊断准确率;脑卒中患者的复发可以根据出院后对其健康检测的数据代入Logistic回归方程评估,继而针对性进行强化二级预防管理。

Objective To build a discriminant analysis model for stroke type diagnosis and prediction of stroke recurrence.Method We retrospectively collected clinical data of244acute stroke patients admitted to the Second Hospital of Hebei Medical University from May2012to September2013,with an average of18-months follow-up.Stroke type models and recurrence models were built with SPSS and R.Logistic regression by IBM SPSS.Cox&Snell R Square test was used to obtain the prediction accuracy.We used SMOTE algorithm when dealing with unbalanced data.Results The result of Cox&Snell R Square test in stroke type prediction was0.634.Prediction accuracy was86.1%.The result of Cox&Snell R Square test in ischemic stroke and hemorrhagic stroke recurrence prediction was0.236and0.272,respectively.Prediction sensitivity was28.6%in ischemic stroke and60%in hemorrhagic stroke.We used SMOTE algorithm when dealing with unbalanced data in the ischemic stroke recurrence prediction.Result of Cox&Snell R Square test was0.488.Prediction sensitivity is89.3%.Conclusion The classification model can be used for assisting diagnosis.Stroke recurrence can be predicted by using the Logistic prediction model in order to assist further secondary prevention.