RESUMO
BACKGROUND: Better screening and testing approaches are needed to improve TB case finding, particularly in health facilities where many people with TB seek care but are not diagnosed using the existing approaches. OBJECTIVE: We aimed to evaluate the performance of various TB screening and testing approaches among hospital outpatients in a setting with a high prevalence of HIV/TB. METHODS: We screened outpatients at a large hospital in Cameroon using both chest X-ray and a symptom questionnaire including current cough, fever, night sweats and/or weight loss. Participants with a positive screen were tested for TB using smear microscopy, the Xpert MTB/RIF assay, and culture. RESULTS: Among 2051 people screened, 1137 (55%) reported one or more TB symptom and 389 (19%) had an abnormal chest X-ray. In total, 1255 people (61%) had a positive screen and 31 of those screened (1.5%) had bacteriologically confirmed TB. To detect TB, screening with cough >2 weeks had a sensitivity of 61% (95% CI, 44-78%). Screening for a combination of cough >2 -weeks and/or abnormal chest X-ray had a sensitivity of 81% (95% CI, 67-95%) and specificity of 71% (95% CI, 69-73%), while screening for a combination of cough >2 weeks or any of 2 or more symptoms had a similar performance. Smear microscopy and Xpert MTB/RIF detected 32% (10/31) and 55% (17/31), respectively, of people who had bacteriologically-confirmed TB. CONCLUSIONS: Screening hospital outpatients for cough >2 weeks or for at least 2 of current cough, fever, night sweats or weight loss is a feasible strategy that had a high relative yield to detect bacteriologically-confirmed TB in this population. Clinical diagnosis of TB is still an important need, even where Xpert MTB/RIF testing is available.
RESUMO
Deep learning (DL) neural networks have only recently been employed to interpret chest radiography (CXR) to screen and triage people for pulmonary tuberculosis (TB). No published studies have compared multiple DL systems and populations. We conducted a retrospective evaluation of three DL systems (CAD4TB, Lunit INSIGHT, and qXR) for detecting TB-associated abnormalities in chest radiographs from outpatients in Nepal and Cameroon. All 1196 individuals received a Xpert MTB/RIF assay and a CXR read by two groups of radiologists and the DL systems. Xpert was used as the reference standard. The area under the curve of the three systems was similar: Lunit (0.94, 95% CI: 0.93-0.96), qXR (0.94, 95% CI: 0.92-0.97) and CAD4TB (0.92, 95% CI: 0.90-0.95). When matching the sensitivity of the radiologists, the specificities of the DL systems were significantly higher except for one. Using DL systems to read CXRs could reduce the number of Xpert MTB/RIF tests needed by 66% while maintaining sensitivity at 95% or better. Using a universal cutoff score resulted different performance in each site, highlighting the need to select scores based on the population screened. These DL systems should be considered by TB programs where human resources are constrained, and automated technology is available.
