慢性病共病中老年人居民医疗保险参与情况的影响因素研究——基于二元逻辑回归和决策树模型的量化分析

Analysis of Influencing Factors of Medical Insurance Participation Willingness of Middle-Aged and Elderly Residents with Chronic Disease Comorbidities—A Quantitative Analysis Based on Binary Logistic Regression and Decision Tree Model

摘要: 目的：探讨慢性病共病中老年人居民医疗保险参与情况的影响因素，为有关部门施策以提高参保率提供现实依据。方法：基于计划行为理论与中国健康与养老追踪调查数据(2020CHARLS)，构建影响指标体系并运用Kruskal-Wallis检验、二元逻辑回归、决策树分析变量对医保参与影响。结果：共纳入8743名调查对象，参保率78.6%；逻辑回归显示年龄、教育、户口、退休情况、婚姻与同居状况、养老保险参与、个人年收入、子女提供经济支持有显著影响(P < 0.05)；决策树显示养老保险参与为主要影响因素，其次是个人年收入、退休情况、户口、婚姻状况；决策树预测效果(AUC = 0.904)稍优于逻辑回归(AUC = 0.883)。结论：年龄<80岁、农业户口、未退休、已婚、参与居民养老保险、收入较高、子女提供较多经济支持是关键影响因素。逻辑回归和决策树分别展示自变量与因变量间依存关系、变量间交互关系对因变量的影响，可结合二者进行医保参与相关分析。

Abstract: Objective: To explore the influencing factors of medical insurance participation of middle-aged and elderly residents with chronic diseases, and to provide practical basis for relevant departments to improve the insurance participation rate. Methods: Based on the theory of planned behavior and the data of China Health and Pension Tracking Survey (2020CHARLS), an impact indicator system was constructed and Kruskal-Wallis test, binary logistic regression and decision tree were used to analyze the impact of variables on medical insurance participation. Results: A total of 8743 respondents were included, and the insurance rate was 78.6%; Logistic regression showed that age, education, household registration, retirement, marital status, pension insurance participation, personal annual income and financial support provided by children had significant effects (P < 0.05). The decision tree shows that pension insurance participation is the main influencing factor, followed by personal annual income, retirement, household registration and marital status. The prediction effect of decision tree (AUC = 0.904) was slightly better than logistic regression (AUC = 0.883). Conclusion: Age < 80 years old, agricultural household registration, not retired, married, participating in residents’ pension insurance, higher income, more financial support provided by children are the key influencing factors. Logistic regression and decision tree respectively show the interdependence between independent variables and dependent variables, and the influence of interaction between variables on dependent variables, the correlation analysis of medical insurance participation can be combined with the two.