Influence of setting and diagnostic algorithm on disease severity among people diagnosed with symptomatic and asymptomatic tuberculosis in South Africa

Background There is growing interest in tuberculosis (TB) community screening and detection of asymptomatic TB (aTB). We explored how setting and screening approach influence the relationship between reported symptoms and underlying disease severity and infectiousness. Methods We compared markers of TB severity and infectiousness (computer-aided detection [CAD, CAD4TBv5] scores derived from chest radiographs and trace vs exceeding trace Xpert MTB/RIF Ultra results) among people diagnosed with aTB or symptomatic TB (sTB) through a community survey, and those diagnosed with aTB or sTB in clinics in the same South African community. We evaluated how variation in screening algorithms influence the relative severity of community-diagnosed aTB vs sTB, and estimated the TB prevalence and severity among people not eligible for testing in the survey (CAD score <25 and no reported symptoms). Results People with clinic-diagnosed sTB had higher CAD scores and a greater proportion of Xpert results exceeding trace than those with community-diagnosed sTB, whereas differences between community-diagnosed aTB and sTB were minimal. Under a hypothetical community universal Xpert testing strategy, people detected with sTB may have more severe disease on average than people detected with aTB. In contrast, restricting testing to people with CAD scores [&ge;]50 and/or reported symptoms would have resulted in higher CAD scores among those diagnosed with aTB than sTB. Conclusions Screening algorithm design and context influence the spectrum of TB disease detected. Community screening approaches where testing is based on symptoms and/or a threshold CAD score may identify aTB that is on average more severe than sTB.