Estimating Vaccine Coverage from Serial Trivariate Serologic Data in the Presence of Waning Immunity

Abstract

Introduction: Vaccine coverage data are typically collected through vaccine registers and retrospective surveys. Alternatively, cross-sectional serosurveys enable direct estimation of vaccine coverage from antibody prevalence by exploiting correlated seropositivity for multi-antigen vaccines. Here, we extend previous methods by accounting for temporal antibody decline in estimating vaccine coverage for measles-mumps-rubella vaccine using serial serosurvey data. Methods: We introduce a Markovian cohort model of antibody waning and boosting applied to dichotomous seropositivity data for measles, mumps, and rubella. Simulation studies are used to test model identifiability and to explore bias induced by previous methods that ignore waning. The cohort model is then fitted to three Australian serosurveys, entailing estimates of vaccine coverage from routine and catch-up vaccination as well as waning rates for each antigen. Results: The simulation results show that the cohort model is identifiable and qualitatively captures the decline in seropositivity observed in older children. When fitted to all three Australian surveys, the estimated seroconversion and waning parameters are similar to estimates based on recent meta-analyses, whereas the coverage estimates appear consistent with previous Australian survey-based estimates. Discussion: We show that previous methods of estimating coverage from serological data can be improved by fitting a cohort model with waning and boosting processes to serial serosurvey data, furthermore yielding estimates of more parameters of interest such as rates of waning. In settings where serial serosurvey data is available, our method could be duplicated or applied to related questions such as coverage in routine two-dose schedules or from other combination vaccines.