Geographical targeting of active case finding for tuberculosis in Pakistan using artificial intelligence software (SPOT-TB): a pragmatic stepped wedge cluster randomized control trial
Abstract
Background
Community-wide active case-finding (ACF) is being increasingly implemented as a tuberculosis (TB) elimination intervention. However, conventional site selection strategies may result in low yields from screening. We evaluated whether an artificial intelligence (AI) software guided targeting strategy could improve detection of TB during screening activities (called camps) relative to routine approaches to site selection in the programmatic setting in Pakistan.
Methods
We conducted a stepped-wedge cluster-randomised trial embedded within Global Fund supported ACF activities implemented by Pakistan’s National TB Program and private sector partners. Thirty mobile X-ray van teams operating in 68 districts were randomly assigned to transition from routine site selection approaches (based on field-staff experience and historical data) to an AI-guided targeting strategy, using the software MATCH-AI. We assessed the effect of the intervention on the primary outcome, Camp Positivity Yield, defined as the number of individuals diagnosed with bacteriologically confirmed TB per camp, using generalised linear mixed models. The primary analysis was by intention to treat. Camps conducted within a 5-km radius of the AI selected locations were included in a “validated” per-protocol analysis. We conducted several district-level subgroup analyses. This trial is registered, number <ext-link xmlns:xlink="http://www.w3.org/1999/xlink" ext-link-type="clintrialgov" xlink:href="NCT06017843">NCT06017843</ext-link> .
Findings
Between August 2023 and September 2024, 3,936 screening camps were conducted (2,046 control, 1,890 intervention), screening 269,254 individuals. In the intention-to-treat analysis, Camp Positivity Yield was 7% higher in the intervention group relative to the control group, however this difference was not statistically significant (adjusted risk ratio [RR] 1·07, 95% CI: 0·94–1·22). In the validated per-protocol analysis, Camp Positivity Yield was 32% higher in the intervention group relative to the control group (adjusted RR 1·32, 95% CI: 1·12–1·54). Yields were highest in districts that had moderate baseline yields of 0.5-1% per population screened prior to the trial (adjusted RR: 1.57, 95% CI: 1.13 – 2.18) and in rural districts (adjusted RR 1.43, 95% CI: 1.23 – 1.65).
Interpretation
The use of an AI-guided targeting strategy significantly increased detection of bacteriologically confirmed TB during active case-finding in the validated per-protocol analysis, relative to conventional site-selection approaches employed by field-staff. This software may be considered as a supportive tool to improve the efficiency of community-based TB case-finding interventions in other high burden countries.
Funding
This study was funded by the Bill & Melinda Gates Foundation grant number INV-037454. The study was embedded within ongoing active case finding (ACF) activities supported operationally by the Global Fund through the National TB Control Program Pakistan.
Research in context
Evidence before this study
We searched PubMed for articles published between Jan 1, 2010, and Dec 31, 2024, using the terms “tuberculosis”, “surveillance”, “active case finding”, “community-based screening”, “hotspot targeting”, “predictive modelling”, and “artificial intelligence”. A previously conducted systematic review by Cudhay et al . in 2018 identified three prospective studies using spatially targeted interventions for TB from low incidence settings but none from high incidence settings. One study reported from Uganda in 2025 identified a moderate benefit of conducting door-to-door screening in a 50-100m radius around households of tuberculosis (TB) cases diagnosed from passive case-finding in health facilities, relative to the general population. We identified several retrospective analyses of screening interventions identifying spatial variation in TB case-detection. However, no prospective studies or randomized evaluations were identified that compared methods for selecting geographic areas for screening, including through data-driven or artificial intelligence (AI) based approaches.
Added value of this study
This stepped-wedge cluster randomized trial is the first prospective evaluation of a geographical targeting strategy for TB screening from a high incidence setting. The trial compared screening in hotspots predicted by an artificial intelligence software (MATCH-AI) to the standard of care that largely relied on staff-led approaches and historical experience for site-selection in Pakistan. We found that use of AI software increased detection of bacteriologically confirmed TB relative to conventional site selection methods in the “validated” per-protocol analysis in which screening locations were verified independently to be within a 5-km radius of hotspots identified by the software. These findings provide further evidence for spatial variation in TB burden and support the strategy of targeted screening in geographic hotspots. This trial was conducted on a national scale and under real-world programmatic conditions. Our results provide empirical evidence in support of a commercially available software that programs in other high-burden, low-resource settings can utilize to improve the efficiency of community-based screening through systematic and data-driven approaches.
Implications of all the available evidence
Globally between 1.9 – 3.5 million incident TB cases are not diagnosed each year resulting in significant morbidity and mortality that may be preventable. Early diagnosis and treatment of these individuals may also prevent community transmission and support TB elimination. There is significant evidence that distribution of TB is spatially heterogeneous, however robust, quantitative approaches towards identification of hotspots using routine surveillance data had been challenging. Our study demonstrates that AI can enhance the efficiency of active case-finding interventions within routine programmatic activities without requiring expansion of screening coverage. For programs operating under resource constraints such as in Pakistan, integrating such tools into routine operational planning may allow screening activities to be more strategically deployed to populations at highest risk. This may improve early diagnosis and treatment of individuals with TB and help accelerate TB elimination efforts in high burden countries with substantial gaps in case detection.
Related articles
Related articles are currently not available for this article.