Trends in behaviour change techniques for implementing and de-implementing healthcare practices using audit and feedback

<title>Abstract</title> Background Audit and feedback (A&amp;F) is a widely used quality improvement strategy to modify healthcare professionals’ practice. However, there is considerable variation in how A&amp;F is applied and in its effectiveness. Investigating this variation, in terms of differences in the behaviour change techniques (BCTs) interventions employ, and why it occurs may provide insights for optimising intervention design. This study, therefore, explored associations between which BCT are used in A&amp;F interventions, behaviour change direction (implementation vs. de-implementation), and target behaviour type. Methods An exploratory secondary analysis was conducted on data from 261 randomized trials of A&amp;F interventions, originally extracted as part of a Cochrane systematic review. Regression analyses investigated whether different BCTs were used for implementation versus de-implementation. These analyses were repeated in subgroups of different behaviour types (e.g., prescribing, testing/examinations). Parallel analyses aggregated data at the study level to assess the robustness of findings. Results Analyses on the whole sample demonstrated that the same BCTs were used to implement and de-implement target behaviours in A&amp;F interventions. Subgroup analyses identified potential associations between specific BCTs and implementation direction within certain behaviour types: social comparison with implementation of treatment decisions/actions; education (unspecified) with implementation of testing/examinations; social support (unspecified) with implementation of prescribing; and feedback on outcome of behaviour with de-implementation of treatment decisions/actions. However, these associations were not consistently replicated across the main and parallel analyses, and may reflect prevailing design practices or methodological artefacts rather than genuine differences in BCT selection. Conclusions This study demonstrated that, overall, A&amp;F interventions utilise similar BCTs regardless of whether behaviours are being implemented or de-implemented. However, exploratory subgroup analyses suggest that tailoring interventions through selectively using certain BCTs for either implementation or de-implementation, depending on the type of behaviour being acted on, may warrant further investigation. Future research should test these hypotheses using theory-informed intervention designs and robust methods, to determine whether current patterns of BCT use reflect true differences in intervention effectiveness.