WS5: interdisciplinary synthesis

In this WS, we will integrate and synthesise the data collected in WS1–4 to explore how population and individual level behaviours and processes interact to contribute to the risk of ABR. Using the patient linked dataset, we will generate hypotheses about direct and indirect drivers of ABR using Bayesian network analysis.³¹ We will use Bayesian networks to identify latent factors in different data types and then connect them with each other and key outcome variables in a heterogeneous network across all data. The network structure will identify direct and indirect influences on ABR, and Bayesian networks' probabilistic inference will predict the probability of impact on ABR of change in different drivers. Multilevel regression will then be used to identify which of these direct and indirect drivers account for the most variance in outcome and provide numerical predictions of modifications. Information about AB sensitivity of urinary pathogens will be blinded from fieldworkers conducting household visits and questionnaires and from microbiologists collecting environmental samples from households to ensure fidelity.