Persuasive COVID-19 vaccination campaigns on Facebook and nationwide vaccination coverage in Ukraine, India, and Pakistan.

Social media platforms have a wide and influential reach, and as such provide an opportunity to increase vaccine uptake. To date, there is no large-scale, robust evidence on the offline effects of online messaging campaigns. We aimed to test whether pre-tested, persuasive messaging campaigns from UNICEF, disseminated on Facebook, influenced COVID-19 vaccine uptake in Ukraine, India, and Pakistan. In Ukraine, we deployed a stepped-wedge randomized controlled trial (RCT). Half of the 24 oblasts (provinces) received five weeks of the intervention, the other half ten weeks of the intervention. In India, an RCT with an augmented synthetic control was conducted in five states (Bihar, Chhattisgarh, Jharkhand, Madhya Pradesh, Rajasthan), whereby 40 out of 174 districts were randomized to receive six weeks of intervention. In Pakistan we deployed a pre-post design, whereby 25 city districts received six weeks of the intervention. Weekly COVID-19 vaccination data was sourced through government databases. Using Poisson regression models, the association between the intervention and vaccine uptake was estimated. In Ukraine we conducted a survey among Facebook users at three time points during the RCT, to ascertain vaccination intentions and trust in vaccines. The campaigns reached more than 110 million Facebook users and garnered 2.9 million clicks. In Ukraine, we found that the intervention did not affect oblast-level vaccination coverage (Relative Risk (RR): 0.93, 95% Confidence Interval (CI) 0.86-1.01). Similarly, in India and Pakistan we found no effect of our intervention (India: RR 0.85, 95% CI 0.70-1.04; Pakistan: RR 0.64, 95% CI 0.01-29.9). The survey among Facebook users in Ukraine showed that trust in vaccines and information sources was an important predictor of vaccination status and intention to get vaccinated. Our campaigns on Facebook had a wide reach, which did not translate in shifting behaviours. Timing and external events may have limited the effectiveness of our interventions.