A view over the horizon: cohort incidence model
After spending a lot of time with the primary literature on typhoid vaccine efficacy, correlates of protection, natural immunity, dose response from human challenge studies, seroepidemiology studies, and a lot of thinking, my manager Edward Wenger encouraged me to just do the thing. This script is the central artifact of that push.
What? Implement the new intrahost immunity model in the context of a constant force-of-infection cohort model. Show that I can reproduce medium, high, and very high incidence archetypes being considered by a WHO-organized modeling consortium exercise. Other than the waning model, which was calibrated to organized and specific data with a simple log-least-squares approach all default parameters are currently hand-tuned by me, triangulating across various observations I’ll try to summarize below.
What did I learn? It works! The model can accomodate a bunch of non-trivial observations across a diversity of sources and settings. And, having the model prototype with a bunch of outputs has enabled a lot of rich back and forth with Kyra Grantz that’s already making this model better and asking important questions of this new model, their current model, and other sources of data and inferences about how typhoid works.
What’s next? Put the model into Typhoidsim. Get serious about calibration and applying it to vaccine policy questions. Iterate on many science questions and additional model features around shedding.