Uncovering multimorbidity patterns linked to disability in aging populations: A machine learning analysis

Multimorbidity and disability are prevalent in older adults, yet the relationship between specific multimorbidity patterns and disability is poorly understood. This study aimed to identify clinically significant multimorbidity patterns in older adults and assess their impact on disability levels.

We used data from a cross-sectional survey of individuals aged 60 and above in China. A machine learning approach combining the self-organizing maps and K-means clustering was employed to identify multimorbidity patterns. Ordered logistic regression was used to analyze the association between multimorbidity patterns and disability levels. Three-dimensional surface modeling was applied to visualize the heterogeneity of multimorbidity patterns and disability levels.

Considering all the respondents, a total of 10 multimorbidity patterns were identified, each named after the two diseases with the highest prevalence. Older adults with multimorbidity were at greater risk for higher disability levels compared with those without multimorbidity (odds ratio [OR] = 1.679, P < 0.001). Among the identified multimorbidity patterns, the Allomnesia-Arthritis pattern exhibited the highest risk of more severe disability (OR = 3.976, P < 0.001), followed by the Hypertension-Stroke pattern (OR = 3.745, P < 0.001). Heterogeneity analysis revealed that multimorbidity-disability combinations differed across demographic factors, including age, education, and income levels.

This study underscores the significant impact of specific multimorbidity patterns on disability levels in older adults. The application of machine learning techniques offers a more nuanced understanding of how various disease combinations contribute to disability, highlighting the importance of tailored interventions for managing multimorbidity in aging populations. Geriatr Gerontol Int 2025; 25: 1200-1208.