Scholarlink

<p>Against the backdrop of accelerating population ageing, accurately predicting the risk of disability among older adults and analyzing its underlying drivers represent a critical breakthrough for achieving healthy ageing. Existing research predominantly focuses on the impact of disease and health variables on disability, while insufficient attention is paid to the combined effects of socioeconomic, behavioral, and psychological factors. This study utilizes data from the 2014 and 2018 waves of the Chinese Longitudinal Healthy Longevity Survey. Integrating 38 variables encompassing demographic characteristics, health status, chronic disease profiles, socioeconomic status, social behaviors, and psychological dimensions, it employs a combined approach of LASSO for feature selection and machine learning for prediction to construct a dynamic predictive model capable of distinguishing varying degrees of disability. In both three-class and two-class prediction scenarios, the random forest model demonstrated robust predictive performance. Using the SHAP interpretability framework, we found that instrumental activities of daily living (IADL), age, baseline disability level, and place of residence (rural/urban) were high-contribution features in both the secondary and tertiary classification prediction tasks for disability. The presence of IADL impairment, advanced age, and high levels of loneliness and isolation emerged as core associated factors for moderate-to-severe disability risk. This study provides scientific evidence for tiered interventions and optimized resource allocation.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Explainable ML Predicts Disability Risk in China’s Graying Giants: Uncovering Socioeconomic and Health Determinants

  • 期刊论文

  • 发表日期：

    2026年4月28日

查看全文
  • Yunshu Tang,
  • Jiaqi Li

摘要

Against the backdrop of accelerating population ageing, accurately predicting the risk of disability among older adults and analyzing its underlying drivers represent a critical breakthrough for achieving healthy ageing. Existing research predominantly focuses on the impact of disease and health variables on disability, while insufficient attention is paid to the combined effects of socioeconomic, behavioral, and psychological factors. This study utilizes data from the 2014 and 2018 waves of the Chinese Longitudinal Healthy Longevity Survey. Integrating 38 variables encompassing demographic characteristics, health status, chronic disease profiles, socioeconomic status, social behaviors, and psychological dimensions, it employs a combined approach of LASSO for feature selection and machine learning for prediction to construct a dynamic predictive model capable of distinguishing varying degrees of disability. In both three-class and two-class prediction scenarios, the random forest model demonstrated robust predictive performance. Using the SHAP interpretability framework, we found that instrumental activities of daily living (IADL), age, baseline disability level, and place of residence (rural/urban) were high-contribution features in both the secondary and tertiary classification prediction tasks for disability. The presence of IADL impairment, advanced age, and high levels of loneliness and isolation emerged as core associated factors for moderate-to-severe disability risk. This study provides scientific evidence for tiered interventions and optimized resource allocation.