Using an Analysis of Behavior Change to Inform Effective Digital Intervention Design: How Did the PRIMIT Website Change Hand Hygiene Behavior Across 8993 Users?

Abstract

BackgroundIn designing digital interventions for healthcare, it is important to understand not just whether interventions work but also how and for whom—including whether individual intervention components have different effects, whether a certain usage threshold is required to change behavior in each intervention and whether usage differs across population subgroups.PurposeWe investigated these questions using data from a large trial of the digital PRimary care trial of a website based Infection control intervention to Modify Influenza-like illness and respiratory tract infection Transmission) (PRIMIT) intervention, which aimed to reduce respiratory tract infections (RTIs) by increasing hand hygiene behavior.MethodBaseline and follow-up questionnaires measured behaviors, intentions and attitudes in hand hygiene. In conjunction with objective measures of usage of the four PRIMIT sessions, we analysed these observational data to examine mechanisms of behavior change in 8993 intervention users.ResultsWe found that the PRIMIT intervention changed behavior, intentions and attitudes, and this change was associated with reduced RTIs. The largest hand hygiene change occurred after the first session, with incrementally smaller changes after each subsequent session, suggesting that engagement with the core behavior change techniques included in the first session was necessary and sufficient for behavior change. The intervention was equally effective for men and women, older and younger people and was particularly effective for those with lower levels of education.ConclusionsOur well-powered analysis has implications for intervention development. We were able to determine a ‘minimum threshold’ of intervention engagement that is required for hand hygiene change, and we discuss the potential implications this (and other analyses of this type) may have for further intervention development. We also discuss the application of similar analyses to other interventions.

A Framework for Analyzing and Measuring Usage and Engagement Data (AMUsED) in Digital Interventions: Viewpoint (Preprint)

“…Through this process it is possible to identify effective engagement, finding the level of usage necessary for a specific intervention in order to change the target behavior [9]. This will lead to DBCIs being developed to be more concise, efficient and targeted, making them less arduous for the user and supporting higher rates of uptake and engagement.…”

“…how an intervention worked, when and for whom [6,7]. Recent interest in 'effective engagement' considers these individual patterns of usage and the minimum level of engagement necessary for changes in target behavior to occur, including variation across individuals [6,[8][9][10]. Effective engagement may be achieved through using the intervention a certain amount.…”

“…Effective engagement may be achieved through using the intervention a certain amount. Alternatively it may be context-dependent whereby viewing specific content, in a certain order, or at an appropriate time, provides the minimum threshold necessary for change [6,9,10].…”

“…BCTs) are used by subgroups with particular characteristics (i.e. behavioral determinants), and the extent to which there is a relationship to the target behavior [9,15]. Devising a plan of analysis to answer these questions using the fine-grained data often available from DBCIs enables us to examine the constructs of the logic model and further our understanding of the mechanisms of action underlying successful behavior change [16,17]..…”

“…Researchers have recently advocated using DBCI data in this way [9,10,18]. However, efficient management and systematic analysis of large data sets can be challenging in the absence of detailed guiding frameworks [8], particularly if specific testable hypotheses/questions about usage have not been considered in advance of data collection.…”