Background: Health behaviours play a central role in health status and quality of life, and engaging in behaviours such as physical inactivity, unhealthy eating, smoking tobacco, and alcohol use are leading risk factors for chronic disease. However, most literature in health psychology focuses on each health behaviour in isolation, whilst everyday life experience is characterized by engaging in multiple different behaviours. The proportions of Canadians engaging in multiple sub-optimal health impacting behaviours concurrently is not well understood, nor are the interactive relationships between multiple health behaviours and health outcomes. Moving from a single behaviour to a multiple behaviour paradigm can enable a new set of questions to be answered about which health behavioural combinations people tend to engage in, and what are the strengths and directions of associations between health behaviours, questions for which we do not yet have robust answers. This dissertation aimed to advance the basic science of 'multiple health behaviours' by examining the co-occurrence and covariation of health impacting behaviours.
Methods: The thesis presents four studies that draw upon two large datasets: Studies 1, 2, and 3 use cross-sectional and longitudinal data (n = 40,268) from the Canadian Longitudinal Study of Aging (CLSA) while Study 3 and 4 use cross-sectional and longitudinal data from the international COVID-19 awareness, responses, and evaluation (iCARE) study (n = 85,861). Study 1 examines the co-occurrence of health impacting behaviours assessed with unsupervised machine learning methods, while Study 2 investigates the predictive utility of cluster analysis using multiple supervised machine learning methods. Study 3 investigates the interconnectedness of health behaviours, and their sociodemographic patterns, via network psychometrics (i.e., recursive partitioning-based network trees and network comparison tests) using cross-sectional data. Finally, Study 4 models the temporal associations between traditionally studied health behaviours and COVID-19 pandemic protective behaviours using temporal, contemporaneous, and between-subject network analysis.
Results: Cluster analysis performed with data from the Canadian Longitudinal Study of Aging revealed seven groups of people based on similarity of behaviours (Study 1). These groups demonstrated sociodemographic variation but were not stronger predictors of health outcomes than individual behaviours. This pattern was consistent across several machine learning models (Study 2). Network psychometric analysis of national and international datasets explored correlations between health behaviours and revealed generally small associations with the exception of a larger relationship between physical activity and healthy diet, while the relationship between mask use and social distancing was stronger for males then women. (Study 3). The temporal dynamics of health behaviours (e.g., physical activity, alcohol consumption) and pandemic related health behaviours (e.g., hand washing, physical distancing) were modelled with items within the iCARE survey which identified bidirectional temporal effects between outdoor mask wearing and vaping behaviour as well as a temporal relationship between outdoor mask use and healthy eating (Study 4).
Discussion: This dissertation aimed to advance the basic science of multiple health behaviours through an examination of the co-occurrence and co-variation of health impacting behaviours. Using cross-sectional and longitudinal data from the CLSA and the iCARE study, I identified seven clusters of commonly co-occurring health behaviours and their sociodemographic characteristics (Study 1), compared these clusters against individual behaviours for classifying and prediction health outcomes using machine learning (Study 2), explored the interconnectedness of traditionally studied behaviours and pandemic specific behaviours and identified sociodemographic patterning (Study 3), and modelled the temporal relationships between health behaviours over time during the Covid-19 pandemic (Study 4). In the multiple health behaviour change literature, it is assumed that health behaviours covary; however, findings from this dissertation call into question this assumption. Additionally, the lack of alignment between covariation and co-occurrence approaches for modelling the interconnectedness of health behaviours call into question the validity of cluster analysis for determining which behavioural combinations co-occur in the population. Before behavioural science can explain and predict health behaviour change, we must establish the basic science of multiple health behaviours.
| Identifer | oai:union.ndltd.org:uottawa.ca/oai:ruor.uottawa.ca:10393/45483 |
| Date | 27 September 2023 |
| Creators | van Allen, Zachary |
| Contributors | Presseau, Justin |
| Publisher | Université d'Ottawa / University of Ottawa |
| Source Sets | Université d’Ottawa |
| Language | English |
| Detected Language | English |
| Type | Thesis |
| Format | application/pdf |
| Rights | Attribution-NonCommercial 4.0 International, http://creativecommons.org/licenses/by-nc/4.0/ |
Page generated in 0.0017 seconds