Abstract
INTRODUCTION: The objective of the current investigation was to develop a clinical predictive model for virological treatment failure in HIV patients with low level viremia. METHODS: The study included 786 patients with HIV-associated low-level viremia (LLV). Using Lasso and multivariable logistic regression, we developed a predictive model from clinical and laboratory variables to identify significant predictors. This predictive model was presented as a nomogram and subsequently transformed into a scoring system. Following model construction, internal validation was performed to evaluate the model's calibration capability and clinical utility. RESULTS: The final model incorporated five predictors (HLLV, NVP/3TC/AZT, WHO stage 1, ART delay, triglyceride) into a point-based scoring system. Using the Youden index, a threshold of 6 points was determined. The model demonstrated good performance, with training and internal validation AUCs of 0.762 and 0.759, respectively, and satisfactory calibration and diagnostic accuracy. CONCLUSION: New scoring system predicts virological failure in low-level viremia, supporting early clinical intervention.