Fortifying the automated smartphone-based Cough Audio classification for rapid triaGE testing for tuberculosis project

Tuberculosis (TB) is the single biggest infectious cause of death globally, a situation worsened by COVID-19.

Revolutionary new TB triage tests are required, including at facilities where people are accessible and efficiently referable to confirmatory testing.

The Cough Audio triaGE for TB (CAGE-TB) study was EDCTP2-funded to 1) collect cough sounds at people’s point-of-entry into primary care facilities (South Africa) and derive an audio classifier, 2) validate diagnostic accuracy in independent cohorts (South Africa, Uganda), and 3) deploy mixed methods research (costing, implementation science, medical anthropology) to inform design and implementation so that this classifier, which will report people as “likely TB” or “unlikely TB” for confirmatory testing, is embedded within a user-friendly mHealth app with on-device offline computation.

Per the original call’s objective, CAGE-TB’s goal was to deliver an accurate validated mHealth app usable in trials assessing clinical outcomes (necessary for adoption).

Uniquely, this pure mHealth innovation mitigates barriers that jeopardise target product profile criteria (e.g., reagents, cold-chain, transport, infrastructure).

After CAGE-TB hired personnel, COVID-19 prevented and slowed recruitment (people had limited clinic access, recruitment interrupted by successive COVID-19 waves).

Throughout this, CAGE-TB paid personnel and trainees, resulting in severe budget overruns without participant recruitment, and limited trainee progress and site visits.

In 4-CAGE-TB we request essential support over two years to accomplish the original scope-of-work, ensure trainees can finish degrees and, due to the longer recruitment period, accommodate critical additional site visits.

COVID-19 has only reinforced our premise: it caused TB to increase for the first time in a decade, damaged already weak facility-based triage practices, and accelerated cough classification technologies.