决策树与回归技术在居民就诊影响因素中的应用研究

目的了解影响重庆地区居民就诊的主要影响因素,满足更多居民卫生服务需求,提高卫生服务利用率。方法针对重庆地区不同人群的特征,不同的卫生政策,构建影响居民就诊率的决策树和logistic回归模型。结果调查的11570名居民中,合计就诊人次为2447次,两周就诊率为21.15%(城市为12.58%、农村为29.19%)。决策树与回归分析的错误分类率分别为0.147与0.267,决策树共有6层、13个节点,树的根节点为年龄,说明此变量对就诊率的影响最大,且模型中所选出的影响因素对不同人群的影响不同;logistic回归模型共筛选出10个变量,年龄依然是就诊最重要的影响因素,两者比较,最终选择决策树模型作为居民就诊影响因素分析依据。结论重庆地区居民就诊影响因素较多,且不同人群的影响因素不同,因此,在制订卫生服务规划时,要从各因素考虑,根据不同人群提出针对性的卫生政策。

The Application of Decision Tree and Logistic Regression in the Study of Factors Affecting Residential Health Treatment

Objective To better know main factors affecting the treatment of residents.With different affecting factors in different groups,different health policies shall be taken to meet the demands of health service of more residents in accordance Therefore,health service utilization shall be improved.Methods Aiming at the influent factors of different populations in Chongqing and different health polices,then construct decision tree and logistic regression models.Results Of 11570 residents receiving investigation,there are 2 447 person seeing the doctors in total;what is more,the treatment rate for two weeks is 21.15%(the city is 12.58% and 29.19% in rural areas).The misclassification rate of decision tree and regression models are 0.147 and 0.267.There are 6 floors and 13 nodal points in the decision tree.The nodal points is age,indicating that the variant affects most on treatment rate,what’s more,the selected factors are different for different groups.Seeing from the logistic Regression model,we seek 10variables all totally,and age is also the important factor of the treatment rate.Compared with each another,we choose decision tree as the basis to analysis the factors on the treatment rate.Conclusion There are many factors on the health service treatment of the residents in Chongqing.And the influent factors are different in different groups.Therefore,while formulating health service treatment plan,corresponding health policies should be put forward aiming at different groups.