BMJ Open. 2025 Nov 4;15(11):e105026. doi: 10.1136/bmjopen-2025-105026.
ABSTRACT
OBJECTIVES: Current prediction models for disease progression to AIDS in people living with HIV primarily rely on traditional statistical methods. This study aimed to develop and compare four machine learning models and to create a clinically applicable nomogram for identifying risk factors associated with AIDS progression.
DESIGN: A retrospective cohort study conducted from January 2013 to December 2022.
SETTING: Yining City, Xinjiang, China.
PARTICIPANTS: Newly diagnosed HIV-infected patients (aged 18-60 years) who received antiretroviral therapy and had not progressed to AIDS at baseline.
PRIMARY OUTCOME MEASURES: Progression from HIV infection to AIDS, as defined by the Chinese Center for Disease Control and Prevention criteria.
RESULTS: Among the 2305 patients included, 652 progressed to AIDS. The cohort was predominantly male, with a mean baseline CD4 cell count of 384 cells/μL. Four machine learning models-Support Vector Machine, Random Forest, Logistic Regression and Extreme Gradient Boosting (XGBoost)-were developed. The XGBoost model demonstrated the best predictive performance (area under the curve, AUC: 0.877). Univariate and multivariate analyses identified WHO clinical stages, CD4 cell count, HIV transmission route, platelet count and haemoglobin level as significant predictors. The developed nomogram achieved an AUC of 0.840. Its calibration curve, after bias correction, showed good agreement with the ideal curve, and decision curve analysis indicated potential clinical utility.
CONCLUSIONS: In this cohort, the XGBoost model showed superior performance for predicting AIDS progression. The proposed nomogram may serve as a practical tool to facilitate rapid risk assessment in similar clinical settings. These findings suggest that enhanced monitoring and regular follow-up might be beneficial for patients with low CD4 counts for timely intervention and to improve outcomes.
PMID:41248391 | DOI:10.1136/bmjopen-2025-105026
