Machine learning predicts HIV treatment nonadherence in Uganda

Nearly 85% of adolescents with HIV live in sub-Saharan Africa. This region accounts for half of the world’s almost 40 million HIV cases. In Uganda, the government provides free antiretroviral treatment (ART), but many adolescents between the ages of 10 and 16 do not stick to their medication. This lack of adherence raises concerns about further spreading the virus. Claire Najjuuko is a doctoral student at Washington University in St. Louis. She worked as a data manager in Uganda and aims to use artificial intelligence to help adolescents adhere to their HIV treatment. Her research was published online on February 25, 2025, in the journal AIDS. With guidance from her advisors at the university, Najjuuko developed a machine learning model to predict which adolescents are likely to stop taking their medication. Knowing this can help healthcare providers give extra support to those at risk before they fall behind on treatment. Currently, clinics check adolescents' medication refills and ask about missed doses to assess adherence. Najjuuko's model, however, could change this approach. It could significantly improve healthcare by identifying 80% of adolescents at risk of nonadherence, while reducing false alarms by 14%. To create this model, Najjuuko analyzed data from a study involving 647 adolescents in southern Uganda. She considered various factors affecting adherence, including economic and social variables. Factors such as household income, self-confidence, and family dynamics were key indicators of whether young people would stick to their treatment. Adolescents often struggle to adhere to treatment as they seek independence and face stigma associated with HIV. Some research indicates that having a savings account can positively influence adherence. When young people have resources, they are more likely to take care of themselves. Taking ART can also be challenging due to side effects like nausea. Difficulty accessing food or transportation to clinics can further complicate adherence. Najjuuko’s model has the potential to inform personalized strategies for supporting adolescents, showcasing the importance of combining AI with health research. This research illustrates the collaboration between various fields at WashU, bringing together artificial intelligence and global health to improve health outcomes for vulnerable populations.