Predicting depression severity among people living with HIV in Oman using random forest

Depression is the most common psychiatric comorbidity among people living with HIV and is associated with an increased risk of disease progression. However, existing screening tools generally do not account for sociodemographic and psychosocial factors. Machine learning models offer a promising approach by capturing complex interactions in variables that traditional methods overlook. This cross-sectional, multicentre study included people living with HIV attending infectious disease clinics in four hospitals in Oman. Data were collected between January 2023 and February 2025. Adults aged 18 years and above with a diagnosis of HIV infection were included. Of the 289 recruited participants, 256 with complete data on all model variables were included in the final analysis. Depression severity was assessed using the PHQ-9 scale and sociodemographic and psychosocial variables were collected using validated questionnaires. A Random Forest regression model was implemented using the JASP Machine Learning module. The depression prevalence among people living with HIV was 26.3%. The model demonstrated moderate predictive accuracy, explaining approximately 35% of the variance in PHQ-9 scores. Self-efficacy and perceived social support were the strongest predictors of depression severity. Findings underscores the potential of machine learning- particularly Random Forest- in HIV and mental health care using routinely collected data.