Clustering of health behaviours in Canadians : A multiple behaviour analysis of data from the Canadian Longitudinal Study on Aging

Van Allen, Zak and Bacon, S and Bernard, P and Brown, Heather and Desroches, Sophie and Kastner, M and Lavoie, Kim and Marques, M.M. and McCleary, N and Straus, S and Taljaard, M and Thavorn, K and Tomasone, Jennifer and Presseau, Justin (2023) Clustering of health behaviours in Canadians : A multiple behaviour analysis of data from the Canadian Longitudinal Study on Aging. Annals of Behavioral Medicine. ISSN 0883-6612

Text (CLSA_Clusters_Manuscript_Feb_24_Final)
CLSA_Clusters_Manuscript_Feb_24_Final.pdf - Accepted Version
Available under License Creative Commons Attribution.
Download (1MB)

Abstract

Background: Health behaviours such as physical inactivity, unhealthy eating, smoking tobacco, and alcohol use are each leading risk factors for non-communicable chronic disease. Better understanding which behaviours tend to co-occur (i.e., cluster together) and co-vary (i.e., are correlated) may provide novel opportunities to develop more comprehensive interventions to promote multiple health behaviour change. However, whether co-occurrence or co-variation based approaches are better suited for this task remains relatively unknown. Purpose: To compare the utility of co-occurrence vs co-variation based approaches for understanding the interconnectedness between multiple health impacting behaviours. Methods: Using baseline and follow-up data (N=40,268) from the Canadian Longitudinal Study of Aging, we examined the co-occurrence and co-variation of health behaviours. We used cluster analysis to group individuals based on their behavioural tendencies across multiple behaviours and to examine how these clusters are associated with demographic characteristics and health indicators. We compared outputs from cluster analysis to behavioural correlations and compared regression analyses of clusters and individual behaviours predicting future health outcomes. Results: Seven clusters were identified, with clusters differentiated by six of the seven health behaviours included in the analysis. Sociodemographic characteristics varied across several clusters. Correlations between behaviours were generally small. In regression analyses individual behaviours accounted for more variance in health outcomes than clusters.