Trade-offs in diagnostic algorithm, population coverage, and duration of community screening for tuberculosis: A modelling study

BACKGROUND: Current tuberculosis (TB) prevention and care strategies have failed to reduce disease burden at the pace required to meet global targets. Community screening may enable more rapid declines in TB burden, but evidence is limited. We used mathematical modelling to evaluate approaches using different diagnostic algorithms, population coverage, and duration of screening. METHODS AND FINDINGS: We used a deterministic compartmental TB model, which recognised symptomatic and asymptomatic infectious TB (defined by whether an individual reported symptoms at screening), as well as noninfectious TB. We simulated diagnostic algorithms targeting symptomatic infectious TB (prolonged cough with confirmatory Xpert Ultra), infectious TB (Xpert Ultra), or all TB (chest X-ray), and we varied population coverage and duration of screening. Main outcomes were estimated reduction in symptomatic TB incidence and TB mortality over a 10-year horizon. Maximum coverage (100%) and duration (five annual rounds) was projected to reduce symptomatic TB incidence by 26.9% (UI 22.8, 31.5%) with the algorithm targeting symptomatic TB and 74.0% (UI 68.5, 79.1%) with the algorithm targeting infectious TB. However, incidence rebounded at the end of screening, erasing 9.8% and 15.9%, respectively, of those reductions within 5 years. The algorithm targeting all TB showed higher potential for rapid reductions-over 98%-with negligible rebound; however, low diagnostic accuracy of current tools led to prohibitive overdiagnosis, with 7.2 false positives per true positive in a single round of screening. Screening algorithms targeting broader disease definitions (infectious or all TB) generally achieved greater impact with lower population coverage and/or duration. Findings were broadly similar for mortality. As a modelling study, our approach assumed a homogeneous population for simplicity and required assumptions where data were lacking around algorithm performance, treatment acceptance, and treatment completion across disease states. CONCLUSIONS: We show that substantial reductions in TB morbidity and mortality can be achieved by community screening, highlighting the importance of symptom-agnostic algorithms and the need to balance population coverage and duration. To maximise and sustain epidemiological impact, accurate diagnostic tools and appropriate treatment regimens for individuals with noninfectious TB are needed.