Using Machine Learning to Predict Poor Adherence to Medication Among Adolescents Living with HIV

Nearly 85% of the 1.7 million adolescents living with HIV reside in Sub-Saharan Africa. Even with the expansion and access to free antiretroviral therapy (ART), adherence to ART among adolescents is low, increasing the potential for the virus to further spread.

In collaboration with the International Center for Child Health and Development (ICHAD) and the AI for Health Institute (AIHealth), Claire Najjuuko, a doctoral student at WashU’s McKelvey School of Engineering in the Division of Computational and Data Science, set out to develop a machine learning model to predict poor adherence, specifically, which adolescents living with HIV would be less likely to adhere to ART medication. With such knowledge, healthcare practitioners could implement interventions for those identified as less likely to adhere to the treatment plan.

“The current way of practice is that adolescents go to the clinic every month or two for medication refills, and a healthcare practitioner checks how many pills they have left compared to what is expected, and asking the adolescent questions regarding missed doses to establish if they are adhering to the therapy. This project to predict future non adherence can have real impact if implemented in the right way,” noted Najjuuko.

Using data from ICHAD’s Suubi+Adherence study (2012-2025), Claire developed a machine learning model by incorporating socio-behavioral and economic factors, including adherence history. Ultimately, the model analyzed data from 647 adolescents with complete data at 48 months follow-up. Findings recently published in AIDS, indicate that the model accurately identifies 80% of adolescents at risk of nonadherence while lowering the false alarm rate to 52% —14 percentage points lower than a model based solely on adherence history. By reducing false alarms, this model can help healthcare providers focus interventions on those who need them the most, improving patient outcomes while reducing unnecessary follow-ups and provider fatigue. The predictive tool can help identify adolescents at the highest risk of treatment failure, and enable early targeted interventions. However, the tool is still preliminary and its accuracy could be improved by incorporating HIV phenotypic and clinical data.

Claire worked under the mentorship of Dr. Fred Ssewamala, Director of ICHAD at the Brown School and Dr. Chenyang Lu, Director of AIHealth at McKelvey School of Engineering. Prior to joining the PhD program at WashU, Claire served as the Senior Data Manager for ICHAD’s field office in Uganda. The goal of her work is to use artificial intelligence and data science to help improve adolescents’ compliance with treatment in low-resource settings. Read the full story on the McKelvey School of Engineering’s news website.