Study Reveals How AI can Identify Areas with High Tuberculosis Prevalence
A new study published in the Journal of Tropical Medicine and Infectious Disease has revealed that artificial intelligence could be used to identify and map areas with a high prevalence of tuberculosis.
The research, conducted in Lagos, Ogun, Oyo and Osun States, used the EPCON AI-powered Epi-control platform’s hot spot mapping model and demonstrated a 75 per cent accuracy in predicting TB hotspots in Nigeria, than traditional approaches.
EPCON is a healthcare impact organisation specialising in the use of AI to estimate disease burden, predict its evolution and the effect of interventions.
In a statement by Candice Burgess-Look, the intervention of AI in finding people with TB would hasten the progress of fighting the treatable TB disease.
According to the World Health Organisation, TB is an infectious disease caused by a bacterium that affects the lungs of infected persons.
It spreads through the air when infected persons sneeze, cough or spit.
The WHO further states that TB is the second leading infectious killer after COVID-19.
According to the National Tuberculosis, Leprosy and Buruli Ulcer Control Programme, Nigeria ranks sixth among 30 high-burden TB countries globally.
It further noted that the TB burden in Nigeria was 21,000 annually with only 2,384 cases diagnosed.
Continuing, the statement emphasised that about 590,000 new TB cases were reported, of this figure, 140,000 are persons positive for the Human immunodeficiency virus.
Commenting on the study, the Senior Program Manager, Community Programs of the Society for Family Health, Dr. Abiola Alege, said, “One of the biggest benefits of the platform is that we can implement programs more efficiently and we’ve been able to diagnose at least 12,000 more persons with TB that would have been missed completely.
“We have also been able to put family members on CB preventive treatment. We have been able to halt the spread by quickly picking up these people who would have remained in the community, spreading the disease. The model has helped us to divert our resources, our limited resources, to where it would be most impactful.”
Read Also:
The Chief Executive Officer at EPCON, Caroline Van Cauwelaert, stated that TB was difficult to detect in low and middle-income countries due to the overstretched healthcare systems.
She said, “With COVID reversing two decades of progress, this ability to detect TB is transformative for African countries. TB is particularly challenging to detect in Low and Middle-income Countries with over-stretched healthcare systems and it carries a high economic burden.
“Without proper treatment, two-thirds of those infected with TB will die. When TB can be quickly and effectively detected, it can be cured – the antibiotics have been available since the 1950s.”
The study used Bayesian machine learning, the AI engine from EPCON, to process large-scale epidemiological and clinical data to predict disease spread and impact.
“The findings from the Nigeria case study found that the yield in population clusters that were predicted by the AI model to have a high positivity for TB rates were at least 1.75 times higher than the yield from locations chosen by standard approaches.
“It concluded that the AI model has the potential to help active case finding (ACF) implementers find areas with high TB positivity and potentially discover undiagnosed TB within these communities.
“Using our Epi-control platform which integrates routine programme data with local sociodemographic and contextual information, we identified TB hotspots for targeted active case finding with impressive results.
“The TB positivity yields at these model-predicted hotspots were significantly higher than those at conventionally selected sites, 73 per cent higher in Lagos, 95 per cent in Ogun, 103 per cent in Osun and 75 per cent in Oyo,” the statement further read.
EPCON, in partnership with AQUITY Innovations, also developed a TB risk predictive model for the Nelson Mandela Bay area in South Africa that increased detection rates from 0.2 per cent in random screening to 0.9 per cent using the model-based approach.
The statement further noted that the Epi-control platform has drastically reduced the cost of finding undiagnosed TB cases in South Africa from $1748 to $437 when compared to conventional approaches.
It added that the SFH in Nigeria, with the use of the AI, now focused on areas where people with TB were likely to be found.
SOURCE: HealthWise
