Testing Multiple Behavioral Science Strategies to Increase Flu-Shot Rates

This research aims to identify which behavioral science strategies are most effective at increasing flu vaccination rates overall and based on patients' individual characteristics. Past behavioral science interventions have shown promise in increasing flu vaccinations. For example, successful interventions have encouraged people to make concrete plans for when they will get a flu vaccination (Milkman et al. 2011), sent automated calls or text messages reminding patients to get a flu vaccination (Cutrona et al. 2018; Regan et al. 2017), or provided financial incentives for getting vaccinated (Nowalk et al. 2010). Although these results are promising, these studies have been conducted in isolation on different populations, which makes it difficult to compare their interventions' effectiveness or to have enough power to reliably detect differing responses to interventions based on individual characteristics.

This research will simultaneously test 19 different SMS interventions to increase flu vaccinations in a "mega-study" and apply machine learning to identify which interventions work best for whom. The interventions are designed by behavioral science experts from the Behavior Change for Good Initiative (BCFG), Penn Medicine Nudge Unit (PMNU), and Geisinger Behavioral Insights Team (BIT). We expect to include at least 80,000 participants.

The specific aims of this research are to identify (1) which behavioral science strategies effectively increase flu vaccination rates overall, and (2) which strategies are most effective for different subgroups (e.g., based on age, gender, race).