The Use of Nudge Strategies in Improving Physicians' Prescribing Behavior: A Systematic Review and Meta-analysis

Background Nudges have been proposed as a method of influencing prescribing decisions.Purpose The purpose of this article is to 1) investigate associations between nudges' characteristics and effectiveness, 2) assess the quality of the literature, 3) assess cost-effectiveness, and 4) create a synthesis with policy recommendations.Methods We searched health and social science databases. We included studies that targeted prescribing decisions, included a nudge, and used prescribing behavior as the outcome. We recorded study characteristics, effect size of the primary outcomes, and information on cost-effectiveness. We performed a meta-analysis on the standardized mean difference of the studies' primary outcomes, tested for associations between effect size and key intervention characteristics, and created a funnel plot evaluating publication bias.Synthesis We identified 21 studies containing 25 nudges. In total, 62 of 85 (73%) outcomes showed a statistically significant effect. The average effect size was -0.22 standardized mean difference. No studies included heterogeneity analyses. We found no associations between effects and selected study characteristics. Study quality varied and correlated with study design. A total of 7 of 21 (33%) studies included an evaluation of costs. These studies suggested that the interventions were cost-effective but considered only direct effects. We found evidence of publication bias.Limitations Heterogeneity and few studies limit the possibilities of statistical inference about effectiveness.Conclusions Nudges may be effective at directing prescribing decisions, but effects are small and health effects and cost-effectiveness are unclear. Future nudge studies should contain a rationale for the chosen nudge, prioritize the use of high-quality study designs, and include evaluations of heterogeneity, cost-effectiveness, and health outcomes to inform decision makers. Moreover, preregistration of the protocol is warranted to limit publication bias. Graphical Abstract This is a visual representation of the abstract.