Perceptions of cough-audio-based tuberculosis screening and its integration into routine care: a qualitative study in South Africa and Uganda
This qualitative study in South Africa and Uganda reveals that while healthcare providers and patients view AI-driven cough-audio screening as a promising tool to enhance objectivity and linkage to testing for tuberculosis, its successful integration into routine care depends on addressing concerns about accuracy, workflow compatibility, and the necessity of maintaining trust within the provider-patient relationship.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Tuberculosis is an ancient bacterial infection that still claims more than a million lives every year, mostly in countries where resources are scarce. The disease spreads through the air when an infected person coughs, sneezes, or speaks, making early detection vital to stop the chain of transmission. For decades, health workers have relied on asking patients about their symptoms—such as a cough lasting more than two weeks, fever, or weight loss—to decide who needs further testing. However, these questions are often missed in busy clinics, or patients do not recognize them as a formal part of their visit. In recent years, scientists have developed a new digital approach: software that can listen to the sound of a cough and analyze it with artificial intelligence to spot signs of tuberculosis. While the technology shows promise in the lab, a critical question remains unanswered: how would real people, from doctors to patients, react to this tool in the middle of a crowded, under-resourced clinic?
A team of researchers set out to answer this by listening to the people who would actually use the system. They traveled to ten health facilities in two countries with high rates of tuberculosis: Uganda and South Africa. Over the course of a year and a half, they spoke with nearly two hundred people, including doctors, nurses, community health workers, and patients waiting for care. They also observed how clinics operated on a daily basis. The goal was not just to see if the software worked technically, but to understand how it would fit into the messy, human reality of a health center. They wanted to know if a phone app that listens to coughs could become a trusted part of the routine, or if it would be ignored, feared, or misunderstood.
The researchers found that while national policies in both countries say that every person entering a clinic should be screened for tuberculosis, the reality on the ground is often different. In many clinics, screening happens only when a patient looks sick or mentions a specific symptom. It is often a quiet, invisible step in a long day of work, sometimes reduced to a simple checkmark on a paper form that neither the patient nor the doctor pays much attention to. Patients frequently do not realize they are being screened; they think they are just waiting for a consultation or a medicine refill. Because the process is so hidden, it misses a chance to educate people about the disease or to encourage them to get tested if they are worried.
When the researchers introduced the idea of a cough-audio app to the staff and patients, the reaction was a mix of hope and caution. Many healthcare workers saw the technology as a way to make screening more consistent and objective. They compared the app to a stethoscope, a familiar tool that helps a doctor listen to the lungs, suggesting that this new digital tool could help them hear signs of trouble that a human ear might miss. They imagined it could help prioritize who needs a confirmatory test first, especially in crowded waiting rooms where time is short. Some patients even liked the idea that they could use the app at home to check their own cough, which might encourage them to seek help sooner rather than waiting until they feel very ill.
However, this enthusiasm was tempered by serious practical concerns. The most common worry was accuracy. Everyone agreed that the app should not replace a doctor's judgment, but rather support it. If the machine made a mistake and told a healthy person they were sick, or missed someone who was actually ill, the consequences could be severe. There were also fears about how the tool would fit into the daily rhythm of a clinic. Nurses worried that stopping to record a cough for every single patient would slow down the line, especially on busy days. They also raised questions about hygiene, noting that sharing a phone or a recording device in a room full of people with respiratory infections could spread germs if the equipment was not cleaned properly.
Trust emerged as the most important factor in whether this technology would succeed. The researchers discovered that people did not trust the technology in a vacuum; they trusted the people who used it. If a nurse or a community health worker introduced the app and explained how it worked, patients were more likely to accept it. If the app seemed to replace the human connection, or if it felt like a cold, automated process, people were hesitant. The study showed that for the tool to work, it had to be woven into the existing relationships between caregivers and patients, not stand apart from them. It needed to be part of a system that already had the resources to follow up on positive results, such as sending people for chest X-rays or treatment, which can be costly and difficult to access in some areas.
The study concluded that the value of cough-audio screening lies not just in the cleverness of the algorithm, but in how it changes the way screening is experienced. Currently, screening is often a passive, forgotten step. A well-designed app could make it a visible, structured moment where a patient and a provider engage with the risk of tuberculosis together. This could lead to better education and a clearer path to getting tested. Yet, the researchers warned that this potential is conditional. The technology will only work if it is accurate, if it is integrated smoothly into the workflow without causing delays, and if the health system has the capacity to handle the people it identifies. Without these conditions, even the most advanced tool risks becoming just another piece of equipment gathering dust in a clinic. The path forward requires careful planning and a deep understanding of the human side of healthcare, ensuring that new tools serve the people who need them most.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.