South Africa’s goal is to achieve epidemic control of Tuberculosis (TB) by 2035. This would be good news indeed, as the country’s 2018 Prevalence Survey indicated an alarming 737 positive cases per 100,000 people, which was higher than expected. This situation is exacerbated by the fact that many people living with the condition haven’t been diagnosed or treated. In addition, according to the Lancet, “South Africa harbours the highest number of cases per capita of multidrug-resistant and extensively drug-resistant tuberculosis.” Against this backdrop, South Africa also has a fragile public health system due, in part, to many years of inequality.
Despite these difficulties, Dr Sipho Nyathie, Director for Health Programs for AQUITY Innovations, is upbeat that epidemic control of TB can be achieved. In the fight against TB, he is in favour of utilising innovative-technology based solutions. These include EPCON’s Artificial Intelligence (AI) model to locate high disease burden areas, as well as public-private partnerships with private sector General Practitioners (GPs).
A GP-centric model with an AI component
Following the successful public-private partnership in the Nelson Mandela Metro between local GPs working in the community and public health facilities, AQUITY Innovations, in 2018, received funding from TB Reach for improved case finding. However, the strategy of finding active TB cases, which involved healthcare workers receiving a list of patients diagnosed with TB from public health facilities, proved cumbersome, explains Dr Nyathie.
“We identified a lot of patients,” says Dr. Nyathie, “but then we needed to go to their homes and some patients stayed far away from one another. A lot of time was wasted. We needed to find clusters of TB patients so we didn’t have to move from one end of town to the other. We wanted to make use of the same GP-centric model but with an AI component.”
Higher accuracy at a lower cost
To achieve this strategy, AQUITY Innovations made use of EPCON’s state-of-the art Bayesian network model. The dashboard, or geoportal, located the all-important diseases hotspots. As more data is added to the system, specifically positive case identifications, the system improves. As it fine-tunes itself, its predictions become more accurate.
“It’s been a really exciting project,” says Dr Nyathie. “With EPCON’s easy-to-use dashboard, I can log on in Pretoria and see the performance of community healthcare workers on a day-to-day basis, allowing us to track performance.’’ Not only the yield improved compared to our previous contact investigation interventions, but the time and the costs were also so much lower.
One of the challenges of active case finding, explains Dr Nyathie, is that it’s very expensive. Studies from South Africa, China and India show each positive TB case costs in the region of $1000. However EPCON’s AI model reduces the cost to about $437 - less than half. “We absolutely want to extend our work with EPCON,” says Dr Nyathie. “We first need to demonstrate epidemic control in Nelson Mandela Bay and then show how this approach is scalable.” AQUITY Innovations is working with Nelson Mandela University to validate their results. “We need to have an academic institution looking at the different parameters to validate our results, but so far, we are very happy.”
Using AI in low and middle-income countries is a game-changer for finding TB at the granular level. Locating these clustered disease hotspots reduces costs and improves time efficiency significantly.