摘要
目的 运用随机森林模型对河南省居民心理健康素养影响因素进行研究,为探索提升居民心理健康素养水平的新干预模式奠定理论基础。方法 于2022年10月—2023年2月采用多阶段分层随机抽样方法抽取河南省18岁及以上常住居民6 207人为研究对象,问卷调查基本情况、生活状态和健康状况,采用国民心理健康素养问卷调查居民心理健康素养水平。利用随机森林模型和logistic回归模型对居民心理健康素养水平影响因素进行分析,采用受试者工作特征(ROC)曲线下面积(AUC)、灵敏度、特异度、精确率、准确度对两模型拟合效能进行分析。统计软件为SPSS 19.0、R 3.6.1和MedCalc 19.2。结果 河南省居民心理健康素养水平为12.32%(年龄标化后为15.35%),心理健康知识判断题得分≥80分者占13.45%,自我评估得分≥24分者占96.61%,案例分析题得分≥28分者占52.52%。随机森林模型的平均精度下降显示前5位重要因素分别为近1年家庭月收入、年龄、文化程度、近1年个人月收入和健康状况,平均基尼系数下降显示前5位影响因素分别为近1年家庭月收入、年龄、锻炼频率、近1年个人月收入和文化程度。随机森林模型和logistic回归模型AUC分别为0.859、0.735,两模型ROC曲线比较差异有统计学意义(Z=14.466,P<0.01),随机森林模型拟合效能优于logistic回归模型。结论 随机森林模型影响因素重要性排序对于探索提升居民心理健康素养水平分类干预模式具有一定指导意义,尤其应关注经济欠发达地区、低家庭月收入和高年龄段人群心理健康状况。
Objective To study the influence factors of residents'mental health literacy in Henan Province based on random forest(RF)model,and lay a theoretical foundation for exploring new intervention models to improve residents'mental health literacy.Methods From October 2022 to February 2023,the multi-stage stratified random sampling method was used to select 6207 permanent residents(≥18 years old)in Henan Province as the subjects.The investigation was performed with the questionnaire(including the basic information,living conditions and health status);and the National Mental Health Literacy Questionnaire was utilized to assess their levels of mental health literacy.The random forest model and logistic regression model were used to analyze the influencing factors of mental health literacy levels.The area under the receiver operator characteristic(ROC)curve(AUC),sensitivity,specificity,precision and accuracy were used to analyze the fitting efficiency of two models.The used software included SPSS 19.0,R version 3.6.1 and MedCalc 19.2.Results The mental health literacy level among residents in Henan Province was 12.32%(15.35%after age standardization).The proportion of subjects with score≥80 on the mental health knowledge judgment questions was 13.45%,the proportion of subjects with score≥24 on the self-assessment was 96.61%,the proportion of subjects with score≥28 on the case analysis questions was 52.52%.The mean accuracy decrease of the random forest tree model showed that the top five important factors were family monthly income in recent year,age,education level,personal monthly income in recent year and health status.The mean decrease of gini coefficient indicated that the top five important factors were family monthly income in recent year,age,exercise frequency,personal monthly income in recent year and education level.The AUC values of the random forest model and the logistic regression model were 0.859 and 0.735,respectively.The ROC curves of the two models showed a statistically significant difference(Z=14.466,P<0.01).The fitting efficiency of the random forest model was better than that of the logistic regression model.Conclusion The ranking of the importance of influencing factors in the random forest model provides valuable guidance for exploring targeted intervention models aimed at enhancing residents'mental health literacy levels.It should pay attention to the mental health for economically underdeveloped areas,low family monthly income residents and residents with higher age.
作者
郭正军
陈怡然
王世林
王玉杰
董娇
王文华
王传升
GUO Zhengjun;CHEN Yiran;WANG Shilin;WANG Yujie;DONG Jiao;WANG Wenhua;WANG Chuansheng(Department of Prevention and Control,the Second Affiliated Hospital of Xinxiang Medical University,Xinxiang,Henan Province 453002,China;不详)
出处
《中国慢性病预防与控制》
北大核心
2025年第10期738-743,748,共7页
Chinese Journal of Prevention and Control of Chronic Diseases
基金
河南省医学科技攻关计划联合共建项目(LHGJ20230537)。
关键词
随机森林模型
心理健康素养
影响因素
拟合效能
Random forest model
Mental health literacy Influencing factor
Fitting efficiency
