Machine learning improves HIV treatment adherence predictions in Uganda

A new machine learning model developed by Claire Najjuuko aims to predict which adolescents with HIV are less likely to adhere to antiretroviral therapy in Uganda. This model incorporates socio-behavioral and economic factors, improving prediction accuracy to 80%. Previously, adherence was monitored through clinic visits and patient self-reports. The new model reduces false alarms by 14 percentage points compared to older methods, allowing healthcare providers to focus on those most at risk of nonadherence. The research highlights the importance of economic factors and personal resources, such as having a savings account, in improving treatment adherence. This interdisciplinary approach combines artificial intelligence with global health insights to enhance patient outcomes.