Prediction of lack of sustained viral suppression in cisgender women living with HIV: a decision tree model.

Sustained viral suppression is a central indicator of HIV care and reflects the interaction between individual, social, and programmatic vulnerabilities affecting engagement. This study aimed to develop and evaluate a decision tree model to predict lack of sustained viral suppression among cisgender women living with HIV. This retrospective observational study used data from 236 cisgender women at a specialized HIV care service in Southeast Brazil between 2014 and 2024. The outcome was viral load detectable as a proxy of treatment nonadherence. Independent variables represent dimensions of vulnerability. A Classification and Regression Tree model was fitted. The dataset was divided into training and test sets, and 10-fold cross-validation was used to guide pruning. Mean age at diagnosis was 39.3 years, and mean CD4 nadir was 489.6 cells/µL. The model showed moderate performance (AUC = 0.82). As immunological status improved, age at diagnosis, ART-related adverse events, and mental health problems contributed to progressively more complex vulnerability profiles associated with the outcome. The findings suggest that interpretable models may contribute to the exploratory identification of vulnerability profiles related to continuity of HIV care.