Author: Nathan Poultney
Poultney, Nathan, 2022 Characterising habitual health behaviour patterns for physical activities in constrained settings, Flinders University, College of Nursing and Health Sciences
Terms of Use: This electronic version is (or will be) made publicly available by Flinders University in accordance with its open access policy for student theses. Copyright in this thesis remains with the author. You may use this material for uses permitted under the Copyright Act 1968. If you are the owner of any included third party copyright material and/or you believe that any material has been made available without permission of the copyright owner please contact copyright@flinders.edu.au with the details.
Successful outcomes for health behaviour change interventions rely in part on the engagement of the subjects in the context of their everyday life. When the context in which the intervention is to be implemented is relatively structured, better engagement might be achieved if the intervention is delivered in harmony with the context. In the design of health behaviour change interventions it would be desirable to instigate the subject’s expected interaction with interventions at a "point-in-time" when an instance of repeated behaviour is occurring or about to occur, rather than inserting disconnected and disruptive new activities. Such structured contextual behaviours that are being practiced repeatedly are deemed “habits”. For example, if the pattern is a short walk to reach some endpoint and then return, the subject may be nudged to extend the length of the walk by varying the return route. Typically, structured settings where this approach may be applicable exist in several free-living situations, such as home, workplace, daily routines, to name a few.
The motivation behind this research was to devise a method for identifying and characterising repeated habitual patterns of health behaviours, specifically for the class of walking and associated sedentary activity in workplace settings, as determined by the collection and analysis of step count data and periods of inactivity from commercially available mobile or wearable consumer fitness devices. This work was confined to the type of information typically provided by these basic sources to provide a utilitarian solution. A few different actual settings, with associated variations in structure and health habits, were selected for application of the research to enable the stability and reliability of the method to be tested.
Generally, consumer fitness devices do not identify habitual patterns of behaviour but provide only aggregate step counts at predetermined time intervals, without fine-grained information on step-to-step variations such as data on speed or stride. They may provide some adjunct physiological information such as heart rate and environmental information such as vertical displacement, which may be useful in broad terms for recognising patterns without the need for precise information on other more nuanced details of the physical activity or environment.
The main contribution of this work is the specification of a staged template procedure that has been devised using the Design Science Research methodology for characterising and identifying habitual patterns of health behaviour. Initially, the Relevance Cycle identifies the data that is of interest by describing the tasks to identify and removing noise. The Design Cycle then characterises the tasks by defining boundaries to focus the scope. Then the Rigor Cycle refines the characterisations in an iterative process to increase the accuracy of the health habit detection. Once the habitual behaviour patterns have been identified, statistical models for the patterns can be constructed so that their subsequent effect on behaviour change interventions can be quantified. The primary effort in the Design Science Research approach was concentrated on the development of the procedural pipeline to provide a universal template for step count activity data analysis. The initial prototype was refined through feedback from a trial application on simulated activity data, as detailed in the first case study.
This research has been trialled in three different constrained environment situations. Case study 1 (Simulated Workplace Tasks) was carried out in an open-plan multi-story workplace setting by most of the participants at Flinders University Tonsley campus and one participant at Western Sydney University Werrington South campus. Case study 2 (Open-Plan Workplace Tasks) was carried out in the same two environments as the Simulated Workplace Tasks case study. Case study 3 (Working from Home Versus Office) was undertaken in both environments from the first two case studies for the working from office part of the study, and in a modern residential apartment environment for the working from home part, during the SARS-CoV-2 pandemic lockdowns in Australia.
In each case study, multiple habitual behavioural patterns were identified and characterised using the proposed method. The resulting characterisation provided a baseline for further experimental work involving health behaviour change.
Through the understanding of these patterns in daily activities (including sedentary behaviour), specific points in the day could subsequently be chosen as appropriate for interventions to increase physical activity. For instance, if points of interest associated with walking could be identified in real-time, subjects could be notified of an opportunity for them to increase their step count immediately and unobtrusively, so that they could maximise their overall daily step count without feeling coerced or requiring them to consciously change to new health habits. In the three case studies reported, a total of 9 different types of patterns which could be used for such nudge type behaviour change interventions were identified and characterised.
Keywords: heath habits, characterising health habits, health behaviour patterns, physical activity, sedentary behaviour, step count
Subject: Public Health thesis
Thesis type: Doctor of Philosophy
Completed: 2022
School: College of Nursing and Health Sciences
Supervisor: Chris Barr