Ateneo study finds AI TB screening cost-effective for rural Philippines

Researchers from Ateneo de Manila University have published a comprehensive study evaluating the cost-effectiveness of artificial intelligence-assisted chest radiography for tuberculosis screening in rural Philippine health units. Published in the August 2026 issue of BMC Health Services Research, the work by lead authors Harold Henrison Chiu, Bryan Christopher Lao, and Gloanne C. Adolor addresses a critical bottleneck in Philippine public healthcare: delayed diagnostic turnaround times and the scarcity of specialized radiologists in underserved regions. With the World Health Organization recording approximately 739,000 new tuberculosis cases in the Philippines in 2024, early detection remains vital to preventing severe clinical outcomes and reducing community transmission. In remote areas, patients frequently face prolonged waits for radiologist interpretation or costly teleradiology services, which often necessitate additional travel and financial strain. The research team developed a decision-analytic model simulating a cohort of 1,000 suspected tuberculosis patients undergoing chest X-rays annually over a five-year horizon. The model factored in software licensing, operational expenses, radiologist fees, and confirmatory GeneXpert testing costs. Projections indicate that AI-assisted interpretation would reduce annual operational expenditures to approximately Php 877,330, compared to Php 1.14 million for standard manual reading. This translates to a per-patient cost of roughly Php 877 under the AI model versus Php 1,142 for conventional methods. Beyond direct financial savings, the study emphasizes AI's strategic value in democratizing diagnostic access. The authors argue that for resource-constrained health systems, the primary metric for AI adoption should not be technical superiority over human experts, but rather the capacity to deliver affordable, sustainable, and equitable specialist-level triage to communities lacking medical infrastructure. Implementation feasibility hinges on integrating portable digital X-ray systems and offline-capable AI platforms that function reliably with limited bandwidth. The researchers caution that cost savings are contingent on local pricing structures; when manual reading fees are reduced or region-specific diagnostic accuracy metrics are applied, AI may remain clinically effective without always yielding direct financial savings. Furthermore, AI-generated radiology reports would still require confirmatory laboratory validation before clinical intervention. Rather than pursuing immediate nationwide deployment, the authors recommend phased pilot programs within designated rural health units. These initiatives should incorporate rigorous local validation protocols, continuous quality assurance monitoring, and transparent budgetary assessments to determine optimal integration pathways. If successfully scaled, AI-assisted diagnostic triage could significantly compress referral cycles, minimize patient drop-off rates, and establish a replicable framework for deploying computational health tools in low-resource public health ecosystems.