目的 探讨不同代际居住模式老年人社交活跃度的现状及影响因素，为科学提升老年人社交活跃度提供依据。 方法 对2018年的中国健康与养老追踪调查数据建立随机森林预测模型，甄别共居和独居老人社交活跃度的影响因素，并预测各影响因素的重要性。 结果 所有老人的高社交活跃度比例为26.88%；独居老人的高社交活跃度（18.8%）所占比重高于共居老人（8.1%）。年龄、身体功能障碍、养老金等变量对共居和独居老人的社交活跃度均具有预测作用，疼痛困扰和生命期望仅对独居老人的社交活跃度有预测作用，其预测模型效果良好（共居：AUC=0.86；独居:AUC=0.82）。 结论 老年人高社交活跃度比例较低。随机森林的预测模型结果验证了老年人社交活动的影响交互理论框架的适用性。

Objective To explore the status quo among the social interaction activity of the elderly in different intergenerational living patterns and its influencing factors,and to provide scientific basis for improving social interaction activity of the elderly. Methods Based on the 2018 China health and retirement longitudinal survey data,a random forest predictive model was established to systematically identify the influencing factors of social interaction activity of the elderly in cohabiting and living alone,and predict the importance of each influencing factor. Results The study found that the proportion of all elderly people with high social interaction activity was 26.88%;the proportion of high social activity (18.8%) of the elderly who lived alone was higher than that of the elderly who lived with their children (8.1%).The results of random forest analysis showed that variables such as age,physical dysfunction,pension and other variables had predictive effects on the social interaction activity of cohabiting and living alone elderly,while pain distress and life expectancy only had predictive effect on the social interaction activity of the living alone elderly.The prediction model worked well (cohabiting:AUC=0.86;living alone:AUC=0.82). Conclusions This study proves that the proportion of high social activity among the elderly in my country is relatively low.The predictive model results of random forests validate the applicability of the interactive theoretical framework of the influence of social interaction activities in the elderly.

R473.59

河北省社会科学基金项目（HB21SH020）