Measuring vaccine effectiveness from limited public health datasets: Framework and estimates from India's second COVID wave

Despite an urgent need, authorities in many countries are struggling to track COVID vaccine effectiveness (VE) because standard VE measures cannot be calculated from their public health data. Here, we use regression discontinuity design (RDD) to estimate VE, motivated by such limitations in public health records from West Bengal, India. These data cover 8,755,414 COVID vaccinations (90% ChAdOx1 NCov-19, almost all first doses, until May 2021), 8,179,635 tests, and 141,800 hospitalizations. The standard RDD exploits age-based vaccine eligibility; we also introduce a new RDD-based VE measure that improves on the standard one when better data are available. Applying these measures, we find a VE of 55.2% (95% confidence interval: 44.5 to 65.0%) against symptomatic disease, 80.1% (63.3 to 88.8%) against hospitalizations, and 85.5% (24.8 to 99.2%) against intensive care/critical care/high dependency admissions or deaths. Other data-deficient countries with age-based eligibility for any vaccine—and not just COVID vaccines—can also use these easy-to-implement measures to inform their own immunization policies. VE can be calculated by exploiting age-based eligibility, even when data limitations preclude using standard methods.