Factors Associated With Self-Rated Health Among Older Adults in Japan: A Decision Tree Analysis

Abstract

Background Self-rated health (SRH) is a widely used single-item measure that predicts morbidity, mortality, and healthcare use. In aging societies, such as Japan, SRH serves as a vital public health indicator. Although many factors influence SRH, their relative importance and interactions remain unclear, particularly among older adults. Prior studies have mostly used linear models, which are limited in their ability to capture interactions and non-linear relationships. Such complexities are often present in multifactorial outcomes such as SRH. This study aimed to identify the key determinants of SRH using decision tree analysis in a large sample of community-dwelling older adults in Japan to inform targeted strategies for promoting healthy aging. Method We analyzed cross-sectional data from 1,821 older adults in Ayase City, Japan, corresponding to a response rate of 62.1% from 3,058 individuals invited by mail. SRH was dichotomized into high and low categories. Missing data were addressed using multiple imputations. Decision tree analysis using the classification and regression tree (CART) algorithm identified the key determinants of SRH, focusing on modifiable factors. The predictors included age, sex, Geriatric Depression Scale (GDS) score, Motor Fitness Scale (MFS) score, instrumental activities of daily living (IADL) assessed by the Tokyo Metropolitan Institute of Gerontology Index of Competence (TMIG-IC), and the frequency of going out and exercising. The model performance was evaluated using 10-fold cross-validation. Results Among the 1,821 older adults, 73.5% were classified as belonging to the high SRH group. Higher MFS scores, lower GDS scores, greater TMIG-IC scores, and more frequent going out and exercise were significantly associated with a high SRH (all p < 0.001). Decision tree analysis identified MFS as the most important discriminator, followed by GDS and activity frequency. The model achieved an accuracy of 80.3%, with a specificity of 90.8% and a sensitivity of 51.5%. Conclusions Using decision tree analysis, this study identified MFS, GDS, and TMIG-IC as key determinants of SRH among older adults in Japan. These modifiable factors, including physical function, mental health, and daily competence, offer actionable targets for health promotion. The model's ability to stratify SRH based on practical variables supports its use in guiding individualized and population-level strategies. These findings highlight the importance of addressing motor fitness, depressive symptoms, and functional autonomy through community-based exercise programs, mental health screening, and IADL-enhancing services, in order to improve perceived health and quality of life in aging populations. However, due to its modest sensitivity, the model may be less effective in detecting individuals with low SRH and should be used alongside other screening tools when applied in population health settings.

Keywords: decision tree analysis; depressive symptoms; japanese older adults; physical activity; self-rated health.

Figure 1. Classification tree identifying hierarchical determinants of SRH in community-dwelling older adults

Classification tree identifying key determinants of self-rated health (SRH) among community-dwelling older adults. The model shows hierarchical splits by Motor Fitness Scale (MFS), Geriatric Depression Scale (GDS), and the frequency of going out. MFS ≥ 11 was the primary discriminator of high SRH, followed by GDS and going out frequency in lower-functioning subgroups

Figure 2. Variable importance in the decision tree model for predicting SRH

Relative importance scores of predictors in the classification and regression tree (CART) model for self-rated health (SRH) classification. The Motor Fitness Scale (MFS) was the most influential variable, followed by the Geriatric Depression Scale (GDS), the Tokyo Metropolitan Institute of Gerontology Index of Competence (TMIG-IC), age, the frequency of going out, and the frequency of exercise. Scores reflect each variable’s contribution to reducing Gini impurity during tree construction