Making inferences with incomplete epidemiological data: a proof-of-concept estimating measles vaccine coverage across Canada

Current measles outbreaks around the world have highlighted the evolving immunity landscape of some vaccine-preventable diseases (VPDs). In Canada, there have been approximately 1,100 reported measles cases between January and June 2026 alone. This far exceeds the previous average annual case counts of <200. This significant outbreak has underscored the need to better understand the present state of population immunity to measles in Canada, an essential input parameter for outbreak response models. Most population immunity in this setting is derived from routine childhood vaccination, but vaccine coverage data is available heterogeneously across the country and is typically collected cross-sectionally to monitor population adherence to immunization schedules rather than to inform population-level susceptibility. This study developed statistical methods to adapt available measles vaccine coverage estimates into more complete estimates of present-day immunity in Canada by age and province/territory (PT).

First, a standardized dataset for vaccine coverage by PT was collated from existing publicly-available datasets and reports. A significant number of vaccine coverage estimates were missing by PT, dose, and year. We used Gaussian Process models to impute current coverage with at least one dose of a measles vaccine, to create a complete 'modelled dataset' by PT and age. Our modelling framework was tested against data for England, which was much more complete, to validate our approach.

We developed methodology for estimating vaccine coverage in the absence of detailed population immunity data, to support ongoing measles outbreak modelling response work. While this project focused on measles, we built the associated code/tools such that the methodology can be applied to other vaccine-preventable diseases in future outbreak settings.