Utilization of population-wide digital contact tracing to estimate real-world vaccine effectiveness in a pandemic setting.

Objectives: Observational cohort methods for evaluating real-world vaccine effectiveness can introduce biases and yield negative vaccine-effectiveness estimates. We evaluated if utilization of digital contact tracing (DCT) data could yield more realistic estimates of vaccine effectiveness.

Methods: Vaccine effectiveness against SARS-CoV-2 infection was estimated using an observational population-based cohort of older Singaporeans (≥60 years) and DCT data of contacts (≥60 years) significantly exposed to COVID-19 cases, during Delta/Omicron predominance. Person-day generalized Poisson regressions adjusted for sociodemographic characteristics were performed to estimate adjusted incidence rate ratios of infection at different time intervals from second/third vaccine doses up to 5 months post vaccination, with unvaccinated/partially vaccinated person-time as the reference group. Vaccine effectiveness against infection (VE-I) was calculated as 1 minus incidence rate ratio.

Results: In total, 883 227 and 853 435 older Singaporeans were included in the observational cohort during Delta and Omicron predominance. Also, 102 208 and 347 817 case-contact pairs were identified from national DCT data during Delta and Omicron-predominant transmission. During Delta, estimates derived using the observational population-based cohort method mirrored DCT-based estimates (e.g. boosting, 60-69 years: VE-I = 0.60, 95% CI: 0.57-0.63 [observational cohort]; VE-I = 0.76, 95% CI: 0.72-0.78) [DCT-based estimates]). However, during the Omicron surge, observational cohort methods yielded negative vaccine-effectiveness estimates (e.g. boosting <2 months post vaccination, 60-69 years: VE-I = -1.12, 95% CI: -1.29 to -0.96). When DCT data were utilized to estimate vaccine effectiveness, boosting restored protection (e.g. boosting <2 months post-vaccination, 60-69 years: VE-I = 0.32, 95% CI: 0.23-0.39), with subsequent waning 2-month post-booster (e.g. boosting >5 months post-vaccination, 60-69 years: VE-I = 0.14, 95% CI: 0.04-0.23).

Discussion: During an Omicron COVID-19 surge, observational cohort methods yielded negative vaccine-effectiveness estimates, whereas the use of national DCT data yielded more realistic estimates. DCT data can augment real-world estimates of vaccine effectiveness in the population at large and convey a more complete picture of vaccine-derived protection, especially in the context of a complex immunization landscape and an evolving pandemic situation.