Adapting Behavioral Interventions for a Changing Public Health Context: A Worked Example of Implementing a Digital...

Background: A rigorous approach is needed to inform rapid adaptation and optimisation of behavioral interventions in evolving public health contexts, such as the Covid-19 pandemic. This helps ensure that interventions are relevant, persuasive, and feasible while remaining evidence-based. This paper provides a set of iterative methods to rapidly adapt and optimize an intervention during implementation. These methods are demonstrated through the example of optimizing an effective online handwashing intervention called Germ Defense. Methods: Three revised versions of the intervention were rapidly optimized and launched within short timeframes of 1-2 months. Optimisations were informed by: regular stakeholder engagement; emerging scientific evidence, and changing government guidance; rapid qualitative research (telephone think-aloud interviews and open-text surveys), and analyses of usage data. All feedback was rapidly collated, using the Table of Changes method from the Person-Based Approach to prioritize potential optimisations in terms of their likely impact on behavior change. Written feedback from stakeholders on each new iteration of the intervention also informed specific optimisations of the content. Results: Working closely with clinical stakeholders ensured that the intervention was clinically accurate, for example, confirming that information about transmission and exposure was consistent with evidence. Patient and Public Involvement (PPI) contributors identified important clarifications to intervention content, such as whether Covid-19 can be transmitted via air as well as surfaces, and ensured that information about difficult behaviors (such as self-isolation) was supportive and feasible. Iterative updates were made in line with emerging evidence, including changes to the information about face-coverings and opening windows. Qualitative research provided insights into barriers to engaging with the intervention and target behaviors, with open-text surveys providing a useful supplement to detailed think-aloud interviews. Usage data helped identify common points of disengagement, which guided decisions about optimisations. The Table of Changes was modified to facilitate rapid collation and prioritization of multiple sources of feedback to inform optimisations. Engagement with PPI informed the optimisation process. Conclusions: Rapid optimisation methods of this kind may in future be used to help improve the speed and efficiency of adaptation, optimization, and implementation of interventions, in line with calls for more rapid, pragmatic health research methods.