Longitudinal clustering of health risk behaviours and their association with multimorbidity in the United Kingdom

Understanding the relationship between modifiable risk behaviours such as smoking, poor nutrition, alcohol consumption, and physical inactivity (termed ‘SNAP’ behaviours) and multimorbidity is crucial for disease prevention. Risk behaviours often cluster in specific combinations within distinct subpopulations, but little is known about how this clustering changes with age, and how it is associated with multimorbidity. Thus, the present research aimed to: i) explore whether and how the SNAP behaviours cluster over time, ii) investigate whether and how membership in different behavioural clusters varies by socio-demographic characteristics, and iii) examine which, if any, behavioural clusters are prospectively associated with multimorbidity over time. Using data from longitudinal surveys, a first study analysed data from the English Longitudinal Study of Ageing (ELSA) on older adults (aged 50+). A second study replicated the first using data from the UK Household Longitudinal Study (UKHLS) on a broader age group of adults (aged 16+). A repeated latent class analysis was conducted to identify clusters of SNAP behaviours. Logistic regressions were used to examine how the clusters were associated with socio-demographic characteristics and disease status. Further, a third study conducted latent class moderation to examine whether socio-demographic characteristics moderated the relationship between the clusters identified using the ELSA dataset and disease status. Seven clusters of individuals with distinct SNAP behaviour patterns and sociodemographic profiles were identified in older adults and the general adult sample. Behaviour patterns within these clusters remained fairly stable over time. Across both studies, broad similarities were found in the behaviour profiles and sociodemographic characteristics of clusters that had both the highest and the lowest prevalence of multimorbidity. Clusters characterised by sociodemographic profiles with significant socio-economic disadvantages had a higher prevalence of multimorbidity and other health conditions–despite engaging in fewer risky behaviours than some other clusters. The moderation analysis did not reveal any significant interactions. A strong social gradient was observed in the relationship between risk behaviour clusters and health outcomes such that the clusters characterised by sociodemographic profiles with significant social disadvantages had worse health outcomes–despite engaging in fewer risky behaviours.