Developing a Logistic Regression Machine Learning Model that Predicts Viral Load Outcomes for Children Living with HIV in Gutu District, Zimbabwe

HIV remains a major public health issue globally, particularly in poor resource settings such as the Gutu district of Zimbabwe. The study aimed to develop a predictive viral load outcome model for HIV children based on the CRISP-DM research process. Secondary clinical data for children aged 0–17 years in Gutu were retrieved from the Demographic Health Information System (DHIS2). The study identified age, adherence status, gender, and geographical location as correlated with viral load outcomes. A supervised machine learning logistic regression model was trained with data balance and proper management of complexities. Grid search-based hyperparameter tuning was performed to improve model performance further. The evaluation metrics were accuracy, sensitivity, F1 Score, and area under the receiver operating characteristic curve (AUC-ROC). The model’s performance resulted in 89% accuracy, with all the metrics showing a strong performance. A confusion matrix was used to visualize the results. The findings add to the knowledge on viral load outcome prediction and HIV care in Zimbabwe. The findings suggest that early diagnosis and targeted interventions can improve viral load outcomes in children in Gutu. For future research, the development of the model will be based on more representative data sets and applied to other settings to determine differences in other regions and understand the dynamics of HIV care in children.