Nilesh Saini, Jyoti Jaju, Rakesh Ps, Manu E Mathew, Kamalini Mukhopadhyaya, Nikita Srivastava, Midhun Rajiv, Awele Thurr, Azameera Balaji Nayak, Siddavalam Chandra Mouli, Pramod Tayal, Ph Sarojkumar Sharma, Umesh Kumar, Sundeep Saini
Community-wide AI-assisted chest X-ray screening was feasible to implement at scale and achieved high completion across the diagnostic cascade. Strengthening community participation may enhance the public health impact of similar strategies in high-burden urban settings.
BACKGROUND: Community-wide systematic screening is increasingly recognized as an important strategy for accelerating tuberculosis (TB) case detection in high-burden settings. In 2025, a community-wide TB screening intervention using ultra-portable chest X-ray devices and artificial intelligence (AI)-assisted interpretation was implemented in urban slums of Delhi, India.
METHODS: A concurrent mixed-methods evaluation was conducted in April 2026 across the urban slum clusters, using the Reach, Effectiveness, Adoption, Implementation, and Maintenance (RE-AIM) framework. Quantitative data were obtained from a real-time programme management information system, an independent cluster household survey, and programme records. Qualitative data were collected through document review, field observations, and key informant interviews. Descriptive statistics and thematic analysis were used.
RESULTS: Between July 2025 and March 2026, 3566 screening camps were conducted and 168,568 individuals aged ≥15 years underwent chest X-ray screening. Among the 18,377 individuals with presumed TB, 16,866 (91.7%) completed molecular testing. A total of 877 bacteriologically confirmed TB cases were identified, yielding 520 cases per 100,000 individuals screened and a number needed to screen of 192. Household survey findings estimated screening coverage at 30.5% (95% CI: 24.3-36.8%). Key barriers to participation included low perceived need for screening among asymptomatic individuals and competing work responsibilities, while engagement of ASHAs and doorstep service delivery facilitated uptake. Equipment downtime was low, accounting for 3.5% of planned operational days.
CONCLUSIONS: Community-wide AI-assisted chest X-ray screening was feasible to implement at scale and achieved high completion across the diagnostic cascade. Strengthening community participation may enhance the public health impact of similar strategies in high-burden urban settings.