青岛大学医学院学报2017，Vol.53Issue(3)：253-256,4.DOI:10.13361/j.qdyxy.201703001
血管性认知障碍发病危险因素预测模型研究
A PREDICTIVE MODEL FOR RISK FACTORS FOR VASCULAR COGNITIVE IMPAIRMENT
摘要
Abstract
Objective To establish a predictive model for the risk factors for vascular cognitive impairment (VCI) using Bayesian network.Methods A total of 465 patients with cerebrovascular diseases were selected.A questionnaire survey was performed to collect the data on lifestyle and clinical factors,as well as the demographic data.The patients were randomly divided into training set (421 patients) and test set (44 patients) at a ratio of 10 ∶ 1,and the patients in the training set were further divided into non-VCI group (225 patients) and VCI group (196 patients).The tabu search algorithm was used to establish a Bayesian network model for the data of the training set,and the influencing factors for the development of VCI in patients with cerebrovascular diseases were analyzed.The data of the test set were used to assess the prediction accuracy of this model,and these results were compared with the results of the conventional logistic regression model.Results The Bayesian network model established for the risk factors for VCI had a prediction accuracy of 67.70％,and the prediction accuracy in the test set was 75.00％.Educational level,hobbies,and diabetes were directly associated with the development of VCI.Lifestyle such as eating,drinking tea,drinking,and smoking had indirect influence on the development of VCI via their influence on diabetes.Hobbies including physical exercise not only directly affected the development of VCI,but also had indirect influence via their influence on diabetes.According to the conditional probability tables,low educational level,no hobbies,and diabetes were risk factors for VCI in patients with cerebrovascular diseases.The logistic regression analysis showed that this model had a prediction accuracy of 66.98％,and educational level (P =0.005),drinking (P =0.001),physical exercise (P =0.027),and diabetes (P =0.012) were independent risk factors for VCI.According to the receiver operating characteristic (ROC) curve of the Bayesian network model and the logistic regression model,the Bayesian network model had an area under the ROC curve (AUC) of 0.718 (95％ CI=0.669-0.768),while the logistic regression model had an AUC of 0.664 (95％ CI=0.612-0.717).Conclusion Bayesian network can be used to establish a predictive model for the risk factors for VCI,and the Bayesian network model is superior to the logistic regression model in the accuracy of predicting VCI in patients with cerebrovascular diseases.关键词
贝叶斯网络/认知障碍/危险因素/预测Key words
Bayesian network/cognition disorders/risk factors/forecasting分类
医药卫生引用本文复制引用
王潇,郭宗君,季晓云,王晓林,张敏,王志宏..血管性认知障碍发病危险因素预测模型研究[J].青岛大学医学院学报,2017,53(3):253-256,4.基金项目
青岛市科技局基金资助项目(Kzd-03 （）
09-1-1-33-nsh （）
KZJ-28 （）
15-9-2-74-nsh) （）
青岛市黄岛区科技局计划项目(2014-1-73) （2014-1-73）
