基于神经网络模型的健康老龄化指数构建及应用

Construction and application of healthy aging index based on neural network model

中文关键词: 老年  机器学习  健康老龄化  神经网络模型  指数  效度

英文关键词: elderly  machine learning  healthy aging  neural network model  index  validity

中文摘要:

      目的 健康老龄化具有多维度的特性,因此科学全面地评估健康老龄化水平具有重要意义.本研究旨在基于世界卫生组织关于健康老龄化的概念框架,建立健康老龄化评价指标体系,并利用机器学习技术构建健康老龄化指数.方法 首先,以健康老龄化概念为基础,通过文献复习和专家咨询的方法,构建健康老龄化评价指标体系;然后在2023年6月-9月纳入上海市松江区和四川省攀枝花市3 696名年龄≥50岁的社区居民为研究对象,通过问卷调查收集健康老龄化评价指标体系相关数据.最后,应用机器学习并结合健康老龄化的概念对指标进行筛选后,以衰弱为响应变量,使用神经网络模型构建健康老龄化指数.通过分析健康老龄化指数与自评健康的相关性评估该指数的有效性,并使用多元线性回归分析健康老龄化指数与社会人口特征和健康行为的关系.结果 本研究最终确定用于评价健康老龄化水平的指标为16个,基于神经网络模型构建的人群健康老龄化指数平均为(70.15±14.02).相关分析结果显示,健康老龄化指数与自评健康呈正相关(r=0.4053,P＜0.001).多元线性回归结果显示,年龄与健康老龄化指数呈负相关,尤其是年龄在80岁以上组比年龄50～54岁组低13.74(95％CI:-15.63,-11.85,P＜0.001),同时受教育程度与健康老龄化指数呈正相关,其中高中及以上学历组回归系数为6.77(95％CI:5.20,8.34,P＜0.001).每日摄入水果≥200 g(3.69,95％CI:2.37,5.01,P＜0.001)和规律进行身体活动(17.25,95％CI:15.23,19.27,P＜0.001)的调查对象相对拥有更高的健康老龄化指数.结论 基于神经网络模型构建的健康老龄化指数具有良好的信效度,可用于监测健康老龄化变化轨迹和评价健康老龄化干预措施的效果.

英文摘要:

      Objective Healthy aging has multi-dimensional characteristics,so it is of great significance to evaluate the level of healthy aging scientifically and comprehensively.This study aimed to establish a healthy aging evaluation index system based on the World Health Organization's conceptual framework on healthy aging,and construct a healthy aging index(HAI)by means of machine learning technology.Methods Firstly,based on the concept of healthy aging,an evaluation index sys-tem for healthy aging was constructed through literature review and expert consultation.Then,from June to September 2023,3 696 community residents aged ≥50 years in Songjiang District of Shanghai City and Panzhihua City of Sichuan Province were included as research subjects.The relevant data of healthy aging evaluation index system were collected through questionnaire survey.Finally,the indicators were screened by means of machine learning combined with the concept of healthy aging.With frailty as the response variable,a neural network model was used to construct the HAI.The validity of the HAI was evaluated by analyzing its correlation with self-rated health.Multiple linear regression was used to analyze the relationship between HAI and sociodemographic characteristics and health behaviors.Results In this study,16 indicators were finally determined to evaluate the level of healthy aging,and the average HAI based on the neural network model was(70.15±14.02).The results of correlation analysis showed that HAI was positively correlated with self-rated health(r=0.4053,P＜0.001).The results of multiple linear regression showed that age was negatively correlated with the HAI,especially the HAI of the group aged over 80 years was 13.74 lower than that of the group aged 50-54 years(95％CI:-15.63,-11.85,P＜0.001).The education level was positively correlated with the HAI,and the regression coefficient of the group with high school education or above was 6.77(95％CI:5.20,8.34,P＜0.001).The subjects who consumed≥200 g of fruit daily(3.69,95％CI:2.37,5.01,P＜0.001)and engaged in regular physical activity(17.25,95％CI:15.23,19.27,P＜0.001)had relatively higher HAI.Conclusion The HAI constructed based on neural network model has good reliability and validity.It can be used to monitor the trajectory of changes in healthy aging and evaluate the effect of healthy aging intervention.