A scoping review of statistical approaches to the analysis of multiple health-related behaviours

• A scoping review identified two main approaches in analyses of multiple behaviours.
• Co-occurrence is concerned with multiple concurrent but independent behaviours.
• Analyses of clustering investigate associations among concurrent behaviours.
• There are inconsistencies in the use of terminology.
• Clustering should be reserved for analyses of inter-related behavioural combination.

BackgroundSmoking, diet, exercise, and alcohol are leading causes of chronic disease and premature death, many engage in two or more of these behaviours concurrently. The paper identified statistical approaches used to investigate multiple behavioural risk factors.MethodA scoping review of papers published in English from 2000 to 2011 was conducted; papers are related to concurrent participation in at least two of the behaviours. Statistical approaches were recorded and categorised.ResultsAcross 50 papers, two distinct approaches were identified. Co-occurrence analyses focused on concurrent but independent behaviours, represented by prevalence of behavioural combinations and/or by the summing behaviours into risk indexes. Clustering analyses investigated underlying associations between the concurrent behaviours, with clustering identified by divergences in observed and expected prevalence of combinations or through identification of latent or unobservable clusters. Co-occurrence was more frequently reported, but the use of clustering techniques and, in particular, cluster analytic and latent variable techniques increased across the study period.DiscussionThe two approaches investigate concurrent participation in multiple health behaviours but differ in conceptualisation and analysis. Despite differences, inconsistency in the terminology describing the study of multiple health behaviours was apparent, with potential to influence understandings of concurrent health behaviours in policy and practice.