Impact Evaluation of Use of MATCH AI Predictive Modelling for Identification of Hotspots for TB Active Case Finding in Pakistan: a Pragmatic Stepped Wedge Cluster Randomized Trial

Despite significant progress over the past decades, an estimated 10.6 million individuals fell ill with tuberculosis (TB) in 2021 and the disease caused 1.6 million deaths globally. Pakistan is ranked as the 5th highest TB burden country in the world and TB causes 42,000 deaths annually in the country. A key challenge in the Pakistan's response to TB is ensuring diagnosis and treatment of all individuals with TB. In 2020, out of the 573,000 cases, a total of 276,736 (48%) were notified. Bridging this case-detection gap is a critical objective for the National TB Program (NTP). Active case-finding (ACF), is a potential strategy to increase case-detection by systematic screening of communities for TB. Recent evidence, indicates that ACF can also reduce population-level TB incidence and prevalence through early detection. While ACF interventions have demonstrated effectiveness in community-trials and are now being conducted at scale in Pakistan, concerns remain regarding their yields and cost-effectiveness in programmatic settings.

The primary aim of this study is to investigate whether a targeted approach towards community-based screening using MATCH-AI, an artificial intelligence software that models sub-district TB prevalence, can improve the yield of ACF interventions in Pakistan. In the intervention arm, field-team will conduct community-based ACF activities (called chest camps) primarily in locations predicted by MATCH-AI to have a higher prevalence of TB. In the control arm, field-teams will continue to utilize existing approaches towards camp site-selection. The trial will be conducted in 65 districts of Pakistan in collaboration with implementation partners of the NTP.