The unrecognized role of fidelity in effectiveness-implementation hybrid trials: simulation study and guidance for implementation researchers

Abstract

Abstract Background Effectiveness-implementation hybrid designs are a relatively new approach to evaluate efficacious interventions in real-world settings while concurrently gathering information on the implementation. Intervention fidelity can significantly influence the effectiveness of an intervention during implementation. However little guidance exists for applied researchers conducting effectiveness-implementation hybrid trials regarding the impact of fidelity on intervention effects and power. Methods We conducted a simulation study based on parameters from a clinical example study. For the simulation, we explored parallel and stepped-wedge cluster randomized trials (CRTs) and hypothetical patterns of fidelity increase during implementation: slow, linear, and fast. Based on fixed design parameters, i.e., the number of clusters (C = 6), time points (T = 7), and patients per cluster (n = 10) we used linear mixed models to estimate the intervention effect and calculated the power for different fidelity patterns. Further, we conducted a sensitivity analysis to compare outcomes based on different assumptions for the intracluster-correlation coefficient and the cluster size. Results Ensuring high fidelity from the beginning is central to achieve accurate intervention effect estimates in stepped-wedge and parallel CRTs. The importance of high fidelity in the earlier stages is more emphasized in stepped-wedge designs than in parallel CRTs. In contrast, if the increase of fidelity is too slow despite relatively high starting levels, the study will likely be underpowered and the intervention effect estimates will also be biased. This effect is more accentuated in parallel CRTs, here reaching 100% fidelity within the next measurement points is crucial. Conclusions This study discusses the importance of intervention fidelity for the study`s power and highlights different recommendations to deal with low fidelity in parallel and stepped-wedge CRTs from a design perspective. Applied researchers should consider the detrimental effect of low fidelity in their evaluation design. Overall, there are fewer options to adjust the trial design after the fact in parallel CRT as compared to stepped-wedge CRTs. Particular emphasis should be placed on the selection of contextually relevant implementation strategies.