The entry of technologies into tuberculosis and integrated lung health

India’s tuberculosis (TB) programme has achieved a 21% reduction in TB incidence over the past decade, but the country continues to carry the world’s largest TB burden and has yet to meet TB elimination targets. Over this period, technology solutions have entered the public health domain in increasingly ambitious ways, integrating into the programmatic layer of how health systems organise and monitor care. Digital adherence technologies such as video-observed therapy, smart pillboxes, and missed-call confirmation systems were among the first widely deployed solutions to remotely monitor daily patient adherence to drug regimens. The National TB Elimination Programme  (NTEP) was an early adopter of these tools as a more patient-centred alternative to directly observed therapy (DOT), which required a family or community member to physically observe daily medication intake.

Early evidence from a Max Institute of Healthcare Management (MIHM) study, however, points to important limitations of these technologies. Learnings from the first randomised controlled trial on digital adherence technologies in India indicate that system-level interventions such as structured care coordination with treating physicians, access to referrals for advanced care, and clinical monitoring for modifications of drug regimens are necessary to mediate the effects of these technologies on outcomes. These gaps pre-date the introduction of technologies and reflect a public health approach that assumes intensive surveillance is the primary methodological approach to patient support. Technologies that inherit these assumptions tend to encode them further and are not designed to surface the gap between what is measured and what matters clinically.

Artificial intelligence (AI) applications are now entering the healthcare domain with the potential to carry the same risks at a greater scale and with less visibility. A promising new domain is to leverage the strength of the TB program to integrate workflows that can diagnose and manage multiple lung health conditions, such as  COPD,  asthma, lung cancer,  post-TB lung disease, and other chronic lung conditions. These conditions share common risk factors and a common presentation of symptoms, creating diagnostic ambiguity and uncoordinated care pathways that delay treatment. AI-assisted clinical decision support tools can theoretically support more integrated diagnostic and management workflows within this framework.

Realising the potential of AI solutions again depends on how these technologies become embedded into health systems to monitor care. A simple technological integration into existing surveillance pathways is unlikely to yield results, and if these solutions fail, they will not fail visibly if the metrics used to evaluate them are not calibrated to meaningful outcomes. Effective deployment of AI solutions, therefore, requires rigorous clinical and behavioural evaluation prior to scale; and a deliberate evidence generation agenda that ensures implementation does not outpace understanding of effect.