多水平零膨胀Poisson模型在分析居民两周就诊次数影响因素中的应用  

Identifying Factors Associated with Visits to Doctors using a Multilevel Zero-inflated Poisson Model

作　　者：李迪[1] 熊华英[1] 董婷[1] 刘俊阳 张菊英[1] 

机构地区：[1]四川大学华西公共卫生学院流行病与卫生统计学系,成都610041 [2]精鼎医药研究开发有限公司,成都610041

出　　处：《四川大学学报（医学版）》2015年第4期624-627,共4页Journal of Sichuan University(Medical Sciences)

摘　　要：目的分析农村居民两周就诊次数的影响因素,探讨多水平零膨胀Poisson模型在零过多计数资料中的应用。方法利用第五次国家卫生服务调查四川省农村居民数据,采用多水平随机截距零膨胀Poisson回归模型分析四川省农村地区居民两周就诊次数的影响因素。结果 Logistic回归分析结果表明性别、年龄和慢性病患病是影响农村居民两周患病与否的因素,Poisson回归分析结果表明到最近医疗机构的时间、民族、婚姻状态和受教育程度是农村居民两周患病后就诊次数的影响因素。结论文化程度高、到最近医疗机构时间短以及未婚的居民两周就诊次数多。同时,多水平零膨胀Poisson回归模型能较好地解决数据中过多零的问题。Objective To determine factors influencing patient visits to doctors using a multilevel zeroinflated Poisson model with count data with excessive zero.Methods The data in relation to rural health services collected in Sichuan Province as part of the Fifth National Health Service Survey（2013）were used in this study.A multilevel random intercept zero-inflated Poisson model was established to identify factors associated with visits to doctors in two weeks.Results The logistic regression showed that gender,age and presence of chronic diseases were the main factors associated with the likelihood of visiting doctors in two weeks.The Poisson regression revealed that travel time to the nearest medical facility,ethnicity,marital status and educational attainment were associated with numbers of visits to doctors in two weeks.Conclusion Rural residents with a high level of education,unmarried,and living close to medical facility are more likely to visit doctors than others.The multilevel zero-inflated Poisson model can be used for analyzing data with zero extra.

关 键 词：计数资料 两周就诊次数 影响因素 多水平零膨胀泊松模型 

分 类 号：R195[医药卫生—卫生统计学]