Inference of population-level disease transmissibility from household-structured symptom onset data

TL;DR

This paper introduces a Bayesian method to estimate population-level disease transmissibility from household-structured symptom data, addressing a key gap in early pandemic analysis.

Contribution

The study develops a novel Bayesian approach that infers population-level transmissibility from household data, incorporating separate within- and between-household transmission rates.

Findings

Method performs well on simulated data

Accurately estimates transmissibility with small sample sizes

Removes bias in growing or decaying outbreaks

Abstract

First Few X (FFX) studies collect household-stratified data in the early stages of a pandemic, in order to infer severity and transmissibility of an emerging disease. We present a Bayesian method to approximately infer population-level transmissibility for the first time from such data; previous studies have only inferred household-level transmissibility. To do this we perform the inference at two levels, assuming one transmission rate parameter for within-household infection, and another transmission rate parameter for infection between different households. We use a simplifying assumption: that between-household infections always occur in naive households; while still performing full joint inference on the within-household infection parameters. In addition, a novel technique is used to remove systematic bias when the number of new infections per day is growing or decaying, as is common in real outbreaks, without the need for contemporaneous estimates of the serial interval. The method is validated on simulated data and is shown to perform well, even when the number of infected households is relatively small.