Physical activity, sleep duration, and multimorbidity: a cross-sectional study of prevalence associations and association rule mining of multimorbidity patterns

Multimorbidity has become a major public health challenge in aging societies worldwide. Sleep duration and physical activity (PA) are widely recognized as modifiable behavioral determinants of multimorbidity; however, their associations with specific multimorbidity patterns remain unclear. This study aimed to examine how multimorbidity patterns vary across subgroups defined by sleep duration and PA. Using the 2020 wave of the China Health and Retirement Longitudinal Study (CHARLS), 18,664 participants aged ≥ 45 years were analyzed. Multimorbidity was defined as the presence of two or more of the 15 chronic diseases assessed. Multivariable logistic regression was used to examine associations with multimorbidity prevalence, adjusting for sociodemographic and behavioral covariates. Association rule mining (ARM) was applied to identify multimorbidity patterns and strong association rules across subgroups stratified by sleep duration, PA, and the combined subgroups. In fully adjusted models, short sleep (OR = 1.640, 95% CI: 1.505–1.788) and insufficient PA (OR = 1.187, 95% CI: 1.060–1.329) were independently associated with higher odds of multimorbidity; their combination was associated with the highest odds (OR = 2.148, 95% CI: 1.818–2.539). In exploratory, unadjusted ARM analyses, the highest-support multimorbidity patterns consistently clustered around cardiometabolic and osteoarticular-digestive diseases across all subgroups, with hypertension as the central hub. Comorbidity networks differed across subgroups: The group with both insufficient PA and short sleep exhibited the highest network density and generated the most association rules (54 rules), whereas the group with both sufficient PA and optimal sleep duration exhibited the lowest network density and generated the fewest association rules (8 rules). Strong association rules showed pronounced subgroup specificity: {asthma} → {chronic lung diseases} appeared exclusively in subgroups with short sleep or insufficient PA; {hypertension, diabetes} → {dyslipidemia} was absent in the long sleep subgroup; and rules with stomach or other digestive diseases as the consequent were identified only in short sleep subgroup. Multimorbidity prevalence and patterns differed across subgroups defined by PA and sleep duration. ARM was useful for identifying multimorbidity patterns across these behavioral subgroups. These subgroup differences may help generate hypotheses and provide directions for research on behavior-stratified prevention strategies for multimorbidity.