The Independent Uganda: You get the Truth we Pay the Price
For years, Uganda’s TB response has relied heavily on symptom-based screening. It is a low-cost approach, but one that experts increasingly describe as blunt and inadequate
Kampala, Uganda | PATRICIA AKANKWATSA | Uganda’s response to tuberculosis is entering a new and more technologically driven phase. But as health experts caution, the shift is proving far from simple.
At a recent Tuberculosis AI Virtual Symposium held ahead of World TB Day, health officials, researchers and policymakers gathered to examine how artificial intelligence could reshape the country’s fight against the disease. The meeting, held under the theme “Yes! We can end TB: The role of Artificial Intelligence in ending the TB epidemic in Uganda,” focused on a central question: can technology help close Uganda’s long-standing gaps in detecting tuberculosis?
The answer, many agreed, is yes, but only if deeper structural challenges are addressed. Uganda continues to face a heavy tuberculosis burden. According to the Ministry of Health, an estimated 96,000 people fall ill with TB each year. But nearly 30,000 of those cases go undiagnosed or unreported annually. In 2024 alone, about 88,000 cases were officially recorded, against an estimated 99,000 infections, leaving a significant gap.
These are not just numbers. Each missed case represents a person who may not receive treatment in time, and who may unknowingly pass the disease on to others. In a country where tuberculosis remains the leading cause of death among people living with HIV –accounting for around 60% of deaths in that group—the consequences are particularly severe.
For years, Uganda’s TB response has relied heavily on symptom-based screening. It is a low-cost approach, but one that experts increasingly describe as blunt and inadequate. Evidence presented at the symposium suggests that this method can miss up to half of all active TB cases, especially among people who do not show clear symptoms.
The result has been slow progress, experts said. Since 2015, TB incidence in Uganda has declined by only about 2.3%, far below the pace needed to meet global elimination targets.
Dr. Cissy Kityo, Executive Director of the Joint Clinical Research Centre, acknowledged both progress and persistent shortcomings.
“We still have a problem. We have made commendable progress, but gaps remain: delayed diagnosis, missed cases, and treatment gaps that are still costing lives,” she said.
She stressed that behind every statistic is a human story. “Every gap we mention costs lives; not one life, but several. Every misdiagnosis is not just a statistic. It is a person, a family, a community losing someone who could have been saved.”
Uganda is classified globally as a high TB and TB/HIV burden country, reflecting both the scale of the disease and the limitations of current detection systems. While access to diagnostic tools has expanded over the past decade, reliance on traditional screening methods remains widespread.
That is where artificial intelligence is beginning to enter the picture. “AI is not a distant, futuristic concept,” Dr. Kityo said. “It is already here in Uganda, helping detect TB earlier, track outbreaks faster, and support health workers to make better decisions.”
Three main applications are driving this shift: computer-aided chest X-ray analysis, predictive mapping of TB hotspots, and AI-supported patient follow-up systems.
Digital chest X-rays, combined with computer-aided diagnosis (CAD), are among the most visible tools. Uganda currently operates 79 portable digital X-ray units and 16 mobile clinics, with dozens more expected to be deployed.
Between 2021 and 2025, these systems screened more than 190,000 people. About 30% of the scans showed abnormalities, leading to over 6,000 confirmed TB diagnoses after further testing.
For Dr. Henry Luzze of the National TB and Leprosy Programme, the significance is clear.
“Traditional screening methods are inefficient,” he said, noting that earlier studies found symptom-based approaches can miss up to 50% of TB cases. “This means many people are only diagnosed when the disease has already advanced or spread.”
Artificial intelligence, he argued, offers a way to change that by identifying cases earlier and more accurately. Beyond diagnosis, AI is also being used to predict where TB is most likely to occur.
Philip Tumwesigye of the Infectious Disease Institute presented findings from the EPCON platform, an AI-powered surveillance tool designed to detect and map disease patterns in low-resource settings.
“Incidence is still estimated at about 99,000, but we reported only about 88,000 patients in 2024,” he said. “That shows we still have a persistent gap.”
Traditional methods of identifying TB hotspots often rely on broad assumptions. AI, by contrast, analyses large datasets, including demographic, geographic and epidemiological information, to pinpoint high-risk areas more precisely.
“What EPCON does is generate actionable recommendations,” Tumwesigye explained. “It helps us move from reacting to cases to anticipating where they will occur.”
Early results suggest that focusing on the top 20% of high-risk communities identified by AI could increase case detection by up to 75%.
Such gains could transform Uganda’s approach; from a reactive system to a predictive one. “This allows us to find TB where it is most likely to be missed—and to find it faster, smarter and more efficiently,” he said.
But alongside the optimism, experts repeatedly returned to a critical issue: data.
Artificial intelligence systems are only as reliable as the information they process. In Uganda, data challenges remain significant—ranging from incomplete reporting to inconsistent formats and inaccurate geographic tagging.
“Garbage in, garbage out,” Tumwesigye warned.
Poor-quality data can lead to flawed predictions, misdirected interventions, and wasted resources. The problem is compounded by fragmented health information systems that do not always communicate effectively with one another.
Although Uganda has developed standards for data exchange, implementation is uneven. Legacy systems, incompatible platforms and inconsistent naming conventions continue to hinder progress.
Ethical concerns
Governance and ethical concerns also featured prominently in discussions. Dr. Paul Mbaka, Assistant Commissioner and head of the Health Information Division, emphasised the need for strong regulatory frameworks—particularly when AI is used in clinical decision-making.
“In clinical care, there is a risk of harm and potential litigation if proper safeguards are not in place,” he said.
He noted that many AI tools are still being piloted without full regulatory approval, especially for diagnostic use.
“There is some exposure, because we may not yet have fully validated these systems to replace expert judgement,” he added.
Other concerns include data privacy, system transparency and the “black box” nature of some AI models, which can make it difficult for clinicians to understand how decisions are reached.
Global guidelines exist, from organisations such as the World Health Organization and the Africa Centres for Disease Control and Prevention, but translating them into practice remains a challenge.
Benefits outweigh risks
Despite these obstacles, there was broad agreement among participants that the potential benefits of AI outweigh the risks, provided implementation is carefully managed.
A key theme throughout the symposium was the need to move beyond small-scale pilot projects and expand successful innovations nationwide.
“This symposium is not just another meeting,” Dr. Kityo said. “It is about actionable insights, not theory. If we have successes, we scale them up.”
At the Joint Clinical Research Centre, she added, artificial intelligence is already being integrated into a new five-year strategic plan, covering service delivery, research and public health programmes.
“We see AI as a catalyst,” she said. “It is not replacing health workers—but empowering them.”
Uganda’s ambitions align with global targets under the End TB Strategy, which aims to reduce TB deaths by 95% and incidence by 90% by 2035. But current trends suggest the country is not on track to meet those goals.
With incidence declining by just over 2% in nearly a decade, incremental progress is unlikely to be enough.
Artificial intelligence offers a potential breakthrough, but not a guaranteed one. Its success will depend on stronger data systems, clearer regulations, sustained investment and effective collaboration between government, researchers and development partners.
As Dr. Kityo put it: “Let us envision a TB programme where no patient is left undiagnosed, where data guides every decision, and where innovation becomes routine; not exceptional.”
