African infectious-disease AI review finds 43 studies, uneven coverage and

The researchers collected information about each study, including its main features, key findings and the measures used to show how well its AI models worked. The country breakdown showed that 10 studies came from South Africa, eight from Nigeria, six from Uganda, five from Ethiopia and four from Tanzania. These five countries had the largest numbers of studies included in the review.

The review shows that research is taking place in several parts of Africa. However, it does not show that AI research on infectious diseases is spread evenly across the continent or that the countries studied represent every country's research setting. Random forest was one of the most commonly used machine-learning models. Other commonly used models included support vector machines, artificial neural networks and logistic regression. Among deep-learning methods, long short-term memory was the most commonly used.

The measures of model performance depended on the models and results reported in each study. The review does not show that any one model consistently performed better than others across infectious-disease prediction, management, monitoring and outbreak forecasting. However, the authors concluded that AI is being used more often to address infectious diseases in Africa, although it is still not used as widely as it could be.

The researchers identified several barriers, including a lack of suitable policy frameworks, poor health-data quality, limited digital infrastructure, insufficient funding for health systems and research, and problems with governance. They called for investment and action to address these barriers. The review describes AI as a developing tool whose wider use depends on better data, infrastructure, funding and oversight. Overall, the researchers based their assessment of AI applications in African infectious-disease research on 43 studies selected from 5,050 records found through database searches.