Using Explainable AI Risk Predictions to Nudge Influenza Vaccine Uptake

We will run a randomized controlled trial (RCT) to evaluate the effects of informing patients they are high risk for serious flu complications. We will also study the effects of (a) providing an explanation of the personalized factors that contribute to risk and (b) indicating that risk was determined by an algorithm. Adult Geisinger patients will be included in the study if their flu complication risk score is in the 90th percentile or above. We will randomize approximately 47,000 high-risk patients among four treatment and one control arm. Subjects in treatment arms will receive communications via up to three modalities – postal letter, SMS text, and/or patient portal message – during the first six weeks of the 2021-22 flu season. Our primary outcome is vaccination at 2 weeks after the final outreach date. A secondary outcome is vaccination at 9 weeks after the final outreach date. Additional study outcomes will include flu diagnosis, flu complications, and healthcare utilization.