The NIHR Global Health Research Professorship (GHRP) scheme is open to all professions and all Higher Education Institutions (HEI), based in UK and low- and-middle-income-countries (LMICs), to nominate health researchers and methodologists with an outstanding research record of clinical and applied health research and its effective translation for improved health. Global Health Research Professors are required to have existing strong collaborations or links with collaborators or partners in institutions in countries on the OECD DAC list and the award should plan to strengthen these/support training and capacity development/mentorship in these partners. Tuberculosis (TB) is now the leading infectious killer globally, with unacceptably slow progress being made towards elimination targets. Southern Africa has been the centre of the global TB pandemic for the last 30 years, driven by generalised HIV (human immunodeficiency virus) epidemics. In 2022, more than one third of people with TB were not diagnosed or treated, resulting in sustained ongoing transmission, and a high risk of severe illness and death. A major barrier to efforts to control TB is the suboptimal diagnostic tests available. The most-commonly used test is sputum examination by microscopy, culture, or molecular testing. But many people with TB are not able to produce sputum because of severe illness or weakness. Also, the importance of subclinical TB (where sputum tests are positive, but the patient has no symptoms) has recently been recognised. People with subclinical TB have lung changes on chest X-ray, meaning it is possible to detect disease earlier; however, doctors able to interpret chest X-rays for TB are very limited in most high TB-burden countries. Recently we have shown in a randomised trial in Malawi that highly-accurate computer-aided detection of TB by portable and affordable digital chest X-ray using artificial intelligence software (DCXR-CAD) can increase the number of people diagnosed with TB and speed up diagnosis from a median of 11 days to 1 day. DCXR-CAD uses deep-learning algorithms trained on hundreds-of-thousands of X-rays to identify abnormality patterns consistent with TB. A positive DCXR-CAD needs to be confirmed with a sputum test before treatment is started, with sputum testing being by far the largest cost component of this approach. However, we have recently shown that setting a single DCXR-CAD abnormality threshold for all patients is likely to be highly inefficient, resulting in considerable over-referral for sputum testing. This is because individual patient characteristics – including older age, severity of symptoms, HIV status, and history of previous TB – are strongly associated with the presence of TB. I hypothesise that by setting “adaptive thresholds” accounting for these characteristics, we could dramatically reduce the number of confirmatory sputum tests that are required to diagnose TB, without impacting detection or accuracy. If correct, this would have massive implications for cost-saving for public health programmes in the Global South now rolling-out DCXR-CAD. Through knowledge exchange with a high TB/HIV prevalence country in southern Africa (Malawi), we will firstly undertake an individual participant meta-analysis and epidemiological modelling of existing datasets identified through systematic review to identify optimal adaptive thresholds based on patient characteristics, and their potential impact and cost-benefit on TB diagnosis. Subsequently, we will undertake a trial, randomising adults to receive sputum TB testing based on either the fixed DCXRCAD threshold, or the adaptive threshold, with the main outcome being the number of sputum tests required to confirm one TB case. We will embed community needs and voices throughout the project, and work with national, regional, and global policymakers and DCXR-CAD software developers to direct policy and guidelines to include adaptive chest X-ray screening for TB.

The key objectives in this project are to: 1. Use past research and data to figure out the costs and health benefits of a new adaptive Digital Chest X-ray with Computer-Aided Detection (DCXR-CAD) algorithms and follow-up sputum tests. This will help us find the best ways to screen for Tuberculosis (TB) in high burden countries. 2. Test how well this new artificial intellegence-powered X-ray and follow-up testing works in real life in Malawi, looking at how accurate, fast, and affordable it is. We'll do this with adults who have TB symptoms at local primary care clinics in a randomised controlled trial. 3. Work with affected communities, national, regional, and global policymakers and software developers to direct policy and guidelines to include adaptive artificial intellegence chest X-ray screening for TB (Policy and knowledge translation).