Background:
Leading risk factors for chronic disease – smoking, alcohol consumption, poor nutrition and physical inactivity (SNAP behaviours) – cluster together (i.e., appear in specific combinations in distinct subgroups). Longitudinal clustering and its association with health outcomes are less well understood.

Objective:
This is the first study to identify longitudinal clusters of SNAP behaviours and to relate them to health outcomes in older adults.

Methods:
Using data from Waves 4-8 of the English Longitudinal Study of Ageing (n=3787), we identified longitudinal clusters of SNAP behaviours using latent class analysis. Health outcomes (from Wave 9) included multimorbidity and complex multimorbidity, along with eight body system disorders defined according to the International Classification of Diseases 10th Revision system. To examine how clusters are associated with socio-demographic characteristics and health outcomes, we used multinomial and binomial logistic regressions, respectively.

Results:
Six clusters with stable within-cluster behaviour trajectories were identified: Low-risk (20.9%), Low-risk but heavy drinkers (11.1%), Low-risk but inactive (22.2%), Do nothing (17.2%), Inactive, heavy drinkers (18.1%), and High-risk smokers (10.5%). Health-risk dominant clusters had lower levels of education and wealth. Women dominated the Low-risk but inactive cluster, whereas men dominated the heavy drinking clusters. Low-risk and Low-risk but heavy drinkers had a lower prevalence of all adverse health outcomes compared to other clusters. In contrast, the Low-risk but inactive cluster had the most ‘negative’ outcomes: highest prevalence of multimorbidity, complex multimorbidity, circulatory disorders, and endocrine, nutritional and metabolic disorders. High-risk smokers were most likely to suffer respiratory disorders, while the least physically active clusters were most likely to suffer endocrine, nutritional, and metabolic problems.

Conclusions:
Health behaviour clusters were strongly but differentially associated with health outcomes, suggesting a complex relationship. Identified clusters can be compared with similar analyses in other countries and used to tailor interventions to specific sub-populations and socio-demographic profiles.