Laura Guzman (University of Warwick)

Detection of outbreaks using epidemiological and genetic data: case-study of Campylobacter infections in England

Health authorities consider the identification and investigation of outbreaks a top priority. Timely detection of infectious disease outbreaks enables effective interventions to prevent further transmission. The whole-genome sequencing of pathogens provides a promising source of information that can enhance outbreak detection methods. However, incorporating the complex and rich nature of genomics data into Bayesian models poses a challenge. We propose a new statistical method that leverages spatiotemporal and genetic data to detect outbreaks and demonstrate its use in analysing Campylobacter infections reported in two regions in England.

In our approach, we employ a Bayesian hierarchical model to classify reported infections as either sporadic or outbreak cases. We integrate a Gaussian process into the model to capture similarities in genome sequences and employ Markov Chain Monte Carlo (MCMC) methods for inference. To address the challenge of Bayesian inference in a high-dimensional Gaussian process, we utilise a block sampling algorithm that enhances the MCMC performance.

We applied our method to various subsets of the Campylobacter dataset, including spatial-temporal, spatial-genetic and temporal-genetic subsets. This approach enabled us to identify potential outbreaks with different characteristics. Particularly, by analysing temporal-genetic data, our method identified cases that could potentially be linked with nationwide outbreaks, which would have gone unnoticed using only spatial-temporal methods.

Integrating both epidemiological and genetic data is vital for the detection of otherwise unnoticed outbreaks. Early identification of infectious disease outbreaks plays a crucial role in enabling health institutions and policymakers to promptly pinpoint the sources of the outbreak and implement effective interventions.