A quantitative world: should we stop binarizing our epidemiological data?
Infectious disease models are usually fitted to binary or count data to estimate key epidemiological parameters. However, many of these data sources are derived from quantitative measurements of continuous variables, such as antibody titres or Ct values. In this talk, I will argue that dichotomising data before fitting models needlessly throws away useful information, even when the data are noisy and the data-generating process is not well understood. I will present a framework for comparing the precision of epidemiological parameter estimates derived from binarized or quantitative data and several case studies (SARS-CoV-2 and West Nile Virus) illustrating the benefits, but also limitations, of using quantitative biomarker data in surveillance. With the rise of new, high throughput multiplex diagnostic and serological assays, there is a need to rethink how we use individual-level quantitative data in infectious disease modelling.
We look forward to having you attend the event in-person!
If you wish to join online, please register here.