论文部分内容阅读
目的利用logistic回归模型、决策树C 5.0和神经网络进行脑卒中高危筛查性能的比较,探讨不同模型在疾病中的应用效果。方法所有数据来自上海市心脑血管疾病监测系统和闵行区居民健康档案系统,采用病例对照研究方法,将上海市闵行区2014年1月1日至12月31日所有40岁以上、闵行区常住户籍的1 391例新发脑卒中患者纳入病例组,同时从居民电子档案系统和2014年闵行区居民危险因素调查中抽取≥40岁1 388名常住户籍居民作为对照组,用R统计软件包检测3种模型对脑卒中筛查的准确率。结果 logistic回归模型、决策树C 5.0和神经网络3种模型的精确度分别是71.3%、74.5%和73.5%,灵敏度分别为70.9%、74.8%和67.0%,特异度分别为71.6%、74.2%和80.0%。3种筛查模型的受试者工作特征曲线下面积(AUC)分别为0.801、0.822和0.805,在筛查效果上差异无统计学意义(P>0.05)。3种模型中均有统计学意义的影响因素为职业、是否患有高脂血症和高血压、是否有规律的体育锻炼、高血压家族史。结论 3种模型在筛查效果上无明显差异,筛查的危险因素有所差异,在研究筛查技术时,将不同的方法进行合理的比较,应该根据实际情况结合专业背景进行。
Objective To compare the performance of high-risk screening for stroke with logistic regression model, decision tree C 5.0 and neural network, and to explore the application effects of different models in the disease. Methods All data were from Shanghai Cardiovascular and Cerebrovascular Disease Surveillance System and Minhang District Resident Health Archives System. A case-control study was conducted to compare the data of Shanghai Minhang District from January 1 to December 31, A total of 1 391 newly registered stroke patients were enrolled in the case group. At the same time, 1 388 resident permanent residents aged 40 or above were selected as the control group from the resident electronic file system and the survey of risk factors of residents in Minhang District in 2014, Accuracy of three models for stroke screening. Results The accuracy of the logistic regression model, the decision tree C 5.0 and the neural network model were 71.3%, 74.5% and 73.5% respectively, the sensitivity was 70.9%, 74.8% and 67.0%, the specificity was 71.6% and 74.2% And 80.0%. The area under the working characteristic curve (AUC) of the three screening models were 0.801, 0.822 and 0.805, respectively. There was no significant difference in the screening results (P> 0.05). Among the three models, statistically significant influencing factors were occupation, whether or not suffering from hyperlipidemia and hypertension, regular physical exercise, and family history of hypertension. Conclusion The three models have no significant difference in screening effect and the risk factors of screening are different. When studying screening technology, different methods should be compared reasonably and should be based on the actual situation and professional background.