The research, led by Emelie Yonally Phillips and colleagues, examined the implementation of computer-aided detection software, commonly known as CAD, integrated with digital chest X-ray during a community-based tuberculosis active case finding campaign run by Médecins Sans Frontières. Rather than measuring diagnostic accuracy or yield, the study asked a different and often neglected question: how do the people who operate, oversee, and experience these technologies actually perceive them? Between January and April 2023, the team conducted in-depth interviews, focus group discussions, and direct observations with MSF staff, physicians at local health centers, national tuberculosis program personnel, community leaders, and residents who participated in the screening campaign itself.

The technological premise behind the intervention is straightforward but powerful. Digital chest X-rays can be processed by CAD algorithms that flag images suggestive of tuberculosis, prioritizing which patients need confirmatory testing and reducing the dependence on scarce radiologist expertise. In high-burden, resource-constrained settings, this matters enormously. Active case finding campaigns aim to reach people who would never walk into a clinic on their own, and the sheer volume of X-rays generated in such campaigns can overwhelm human reading capacity. CAD promises to widen the funnel, allowing thousands of residents to be screened rapidly while reserving scarce clinical attention for those whose images suggest disease.

Through thematic analysis of the interview and observation data, the researchers identified four major themes that together map the terrain of CAD implementation. The first concerned operational functionality and efficiency, and here the picture was largely positive. Staff reported that the software increased screening capacity, supported clinical decision-making, and streamlined workloads that had previously been bottlenecks. Participants in the active case finding campaign described the process as convenient, an important finding because community screening only works if people are willing to move through it. Speed and simplicity, in other words, are not cosmetic features but determinants of whether a campaign reaches enough people to make a dent in transmission.

The second theme, human resources and role adaptation, revealed a more complicated reality. Staff perceived the redistribution of roles that accompanied CAD adoption as initially challenging. When an algorithm takes over part of the image-reading process, the work of nurses, radiographers, physicians, and data handlers shifts in ways that can unsettle established routines and professional identities. The study’s authors highlight the need for adequate capacity building and technical support during this transition, suggesting that the success of digital health tools depends as much on workforce preparation as on algorithmic performance. Training, in this account, is not a one-off orientation but an ongoing process of helping staff understand what the software does, what it does not do, and how their own judgment fits around it.

Technical feasibility and system design formed the third theme, and it is here that the study offers some of its most practical lessons. Project managers and staff emphasized that understanding CAD’s capabilities and limitations was essential for planning infrastructure, setting thresholds, and compensating for the technology’s inherent weaknesses. CAD systems operate on probability scores, and the threshold at which an image is flagged as suspicious involves a trade-off: lower thresholds catch more true cases but generate more false positives and more confirmatory testing; higher thresholds reduce workload but risk missing disease. Getting this calibration right requires local knowledge of prevalence, health system capacity, and the consequences of error, none of which can be read off a software manual.

The fourth theme addressed community acceptability and ethical considerations, and it complicates any assumption that high-tech solutions are automatically welcomed. Community participants expressed mixed views on the reliability of novel tools, with some skeptical of what a machine could tell them about their health. The researchers stress the importance of aligning the technology with local healthcare needs, a point that carries real weight in a setting like Tondo, where residents may have long histories of interacting with health services that have not always served them well. Trust, the study suggests, is built when communities see that the technology is embedded in care they can actually access, rather than parachuted in as a demonstration of innovation.

Taken together, the findings demonstrate that implementing CAD in tuberculosis active case finding is both feasible and acceptable, but only when supported by careful planning and adaptation to local contexts. The phrase adaptation to local contexts risks sounding like a platitude, but the study gives it concrete content: infrastructure must be planned around the software’s requirements, thresholds must be tuned to local epidemiology, staff roles must be renegotiated with training and support, and communities must be engaged as participants rather than passive subjects of screening. Where any of these elements is neglected, the technology’s promise erodes.

The implications extend well beyond Manila. The World Health Organization has endorsed computer-aided detection as a triage tool for tuberculosis, and donors and national programs are increasingly investing in digital X-ray and AI reading systems across high-burden countries. Yet the gap between a technology that works in principle and one that works in practice is precisely where many implementation efforts fail. The Tondo experience offers a template for closing that gap, emphasizing that sustainable implementation hinges on continuous technical support and on the meaningful involvement of healthcare workers and communities, whose insights help ensure that digital health innovation is ethical as well as effective.

For a disease that kills more people than any other single infection, the stakes of getting this right could hardly be higher. The study’s authors argue that the experience in Tondo offers transferable lessons for deploying CAD in other high-burden, resource-constrained settings, from informal settlements in megacities to remote districts with minimal radiology capacity. What emerges is a portrait of artificial intelligence in global health that is neither utopian nor dismissive: a tool that genuinely expanded screening capacity and eased workloads, but one whose success was earned through negotiation with human systems, professional habits, technical constraints, and community trust. In the fight against tuberculosis, the algorithm is only as effective as the health system and the community that surround it.

Subject of Research: Feasibility and acceptability of computer-aided detection with digital chest X-ray for tuberculosis active case finding in Tondo, Manila

Article Title: Implementing computer‑aided detection for TB active case finding: A qualitative study of feasibility and acceptability in Tondo, Manila

Article References: Yonally Phillips, E., Galvan, M. D. K. P., Min, J., Hewison, C., Duyala, C., Palmado, J. C., Peral, S. N., Camelique, O., Roxas, M. R. C., Duka, M. D., Pardilla, G. F., Recidoro, M. J. C., Castro, R. H., Hossain, F., Huerga, H., & Carnimeo, V. (2026). Implementing computer‑aided detection for TB active case finding: A qualitative study of feasibility and acceptability in Tondo, Manila. PLOS Global Public Health, 6(10), e0007429. https://doi.org/10.1371/journal.pgph.0007429

Image Credits: AI Generated

DOI: 10.1371/journal.pgph.0007429

Keywords: tuberculosis, computer-aided detection, digital chest X-ray, active case finding, artificial intelligence, Manila, Médecins Sans Frontières, global health, qualitative research, screening, digital health, Philippines

The research team, led by investigators working with Viet Nam’s national tuberculosis program and international partners, evaluated a large-scale public–private mix, or PPM, intervention that ran from 2020 to 2023 across 15 provinces covering roughly 40.6 million people—about 40 percent of the country’s population. Rather than building new clinics, the intervention recruited 3,154 healthcare providers who operated outside the national TB program, including private physicians and other out-of-network facilities, and linked them to the public system through systematic screening, referral, and reporting mechanisms. The model leaned on several reinforcing components: active case finding among people with suggestive symptoms, financial incentives for providers who completed referrals, patient support to keep people moving through the care cascade, and integrated data systems that made private-sector activity visible to program managers in near real time.

The central analytical challenge was that the intervention was not rolled out uniformly. Provinces joined at different times, and the intensity of activity—measured by the volume of verbal symptom assessments conducted—varied widely from place to place and quarter to quarter. Classic evaluation methods, such as a simple before-and-after comparison or a basic difference-in-differences model, can be badly biased in this situation, particularly when treatment effects emerge gradually and vary across units. The researchers therefore used a heterogeneity-robust extension of the two-way fixed-effects framework, a family of estimators designed to handle staggered adoption and dynamic treatment effects without the negative-weighting problems that have plagued conventional approaches. Their estimand was the average treatment effect on the treated: how much the intervention changed notifications in the provinces that actually received it.

The data backbone was a province-level quarterly panel of TB notifications spanning 2016 to 2025, giving the team nine years of pre-intervention history and two years of post-intervention follow-up. All models adjusted for province fixed effects and calendar-quarter fixed effects, absorbing both time-invariant local differences and national seasonal and secular trends. Because the COVID-19 pandemic disrupted TB services worldwide—Viet Nam saw sharp notification drops during lockdown periods—the analysis explicitly accounted for these disruptions rather than letting them contaminate the intervention effect estimates. The primary outcome was the total number of TB cases notified per province per quarter.

The headline result was statistically clear. The intervention was associated with an additional 316.8 notifications per province per quarter, with a 95 percent confidence interval of 23.4 to 610.2 and a p-value of 0.034. Aggregated across 15 provinces and multiple quarters, that per-province effect translates into a substantial national contribution. National-level models, triangulated through an interrupted time-series analysis of aggregate notifications, suggested approximately 3,206 additional notifications annually, with a confidence interval of 1,658 to 4,763 and a p-value below 0.001. In other words, the intervention plausibly accounted for thousands of TB cases each year that would otherwise have gone unreported to the national program—and, critically, untreated or treated without standard support.

One of the study’s most policy-relevant features is its dose–response analysis. Using Poisson pseudo-maximum-likelihood fixed-effects models, the researchers examined whether notification rates rose in proportion to the intensity of intervention activity. They found that every 100 verbal assessments per 100,000 population was associated with a 1.4 percent increase in notifications, expressed as an incidence rate ratio of 1.00014 (95 percent CI 1.00002 to 1.00025; p = 0.018). National models showed a steeper gradient, with a 3.5 percent intensity-associated increase (IRR = 1.00035; 95 percent CI 1.00018 to 1.00052; p < 0.001). This graded relationship matters because it suggests a biological-plausibility analogue for program design: more screening activity produced proportionally more detected cases, which is what one would expect if the intervention itself, rather than some unmeasured confounder, was driving the effect.

The authors went to considerable lengths to stress-test the finding. They repeated the entire analysis using PPM-specific notifications—the subset of cases reported through the intervention channel—as the outcome, and obtained concordant results, which is reassuring because this intermediate outcome is closer to the mechanism of action and less vulnerable to unrelated shifts in public-sector reporting. They also triangulated across the extended two-way fixed-effects estimates and an interrupted time-series analysis of national data, two methods with different assumptions and different vulnerability profiles. Convergence across these approaches substantially strengthens the causal interpretation compared with any single estimate.

Cost is where the findings become genuinely striking for funders. From a funder perspective, the marginal cost per additional notification ranged from US$166 to US$458, depending on the assumptions and estimation approach. In the global TB financing landscape, where donor budgets are flat or shrinking and countries are being pushed toward sustainability, that price point compares favorably with many alternative case-finding strategies. The economic logic is that the intervention does not purchase new infrastructure; it activates existing provider networks that patients already visit, converting latent private-sector demand into notified, program-supported treatment. Each additional notification represents a person connected to quality-assured diagnosis and care, with downstream transmission benefits that the cost figures do not even capture.

The study is not without limitations, and the authors are candid about the most important one: the design was non-randomized, and high-burden provinces were purposively selected as intervention areas. Provinces chosen precisely because they have more TB might have experienced different notification trajectories regardless of the intervention. The team mitigated this risk through conservative control selection and the multi-method triangulation described above, but some residual uncertainty is unavoidable in any quasi-experimental design. The staggered deployment, which complicated the analysis, also turned out to be an analytical asset, because it created natural variation that allowed the dose–response and dynamic-effect models to be estimated at all.

The implications extend well beyond Viet Nam. Intermediary agencies—organizations that sit between national TB programs and thousands of dispersed private providers—have long been recognized as a promising way to operationalize public–private mix, but the evidence base has mostly consisted of pilot projects, small geographies, or uncontrolled before-and-after studies. This study demonstrates that the model can be evaluated rigorously at population scale and that it delivers measurable, dose-responsive gains in notifications when implemented across an entire health system. For the dozens of high-burden countries with dominant private sectors, the message is that engaging private providers through a structured intermediary is not merely a pilot-worthy idea but an affordable, scalable strategy with quantified returns. In an era of constrained global health financing, leveraging networks that already exist may be one of the most efficient paths toward closing the TB treatment coverage gap that stands between the world and its end-TB targets.

Subject of Research: Population-level evaluation of an intermediary-facilitated public–private mix intervention on tuberculosis notifications in Viet Nam

Article Title: Evaluation of the tuberculosis notification impact from an intermediary-facilitated public–private mix intervention in Viet Nam: A population-level quasi-experimental study

Article References: Vo, L. N. Q., Dong, T. T. T., Huynh, H. B., Luong, B. A., Mo, H. T. L., Codlin, A. J., Forse, R., Mai, T. D. T., Nguyen, L. P., Creswell, J., Dinh, L. V., Nguyen, H. B., Vu, P. X., Nguyen, T. D., Nguyen, L. H., Dang, T. M. H., Van Luu, S., Do, N. H., Truong, H. T., … Lönnroth, K. (2026). Evaluation of the tuberculosis notification impact from an intermediary-facilitated public–private mix intervention in Viet Nam: A population-level quasi-experimental study. PLOS Medicine, 23(9), e1005144. https://doi.org/10.1371/journal.pmed.1005144

Image Credits: AI Generated

DOI: 10.1371/journal.pmed.1005144

Keywords: tuberculosis, public–private mix, Viet Nam, TB notification, quasi-experimental study, intermediary agency, active case finding, health systems, PLOS Medicine, treatment coverage, global health financing, epidemiology