Bayesian estimation of the aetiology of fever using big case-control data

Fever is one of the most common symptoms leading to hospital admission in sub-Saharan Africa and Asia. However, little is currently known about the causes of fever in this setting and many febrile illnesses present with non-specific symptoms. In the past, it had often been assumed that fever without specific symptoms was caused by malaria; but this assumption is no longer tenable with a global reduction in malaria combined with better diagnostics.

This project will use data from a newly available big case control data set to investigate the causes of fever in sub-Saharan Africa and Asia. The Febrile Illness Evaluation in a Broad Range of Endemicities (FIEBRE) study is a large international collaboration being conducted at several sites, resulting in around 15000 fever cases and controls with data on a large number of potential causes of fever. There exist standard, methods available to calculate "attributable fractions" from case control data. However, sophisticated new methods based on Bayesian techniques have recently been proposed. This project will use and compare the different methods through application to the FIEBRE data.

The potential for impact from this project is high. Current guidance to clinicians often results in treatable diseases being left untreated or treated with inappropriate antimicrobials, while over-treatment has important consequences for the development of antimicrobial resistance. By improving what is known about the causes of fever, treatments will be able to be more appropriately utilised.