Machine learning algorithms to predict the risk of hyperlipidemia in people with HIV after starting HAART for 6 months.

OBJECTIVE

The purpose of this study was to use machine learning models to predict the risk of hyperlipidemia in people with HIV (PWH) for 6 months after starting HAART, to improve early intervention efforts and prevent further progression to cardiovascular and cerebrovascular diseases.

METHODS

This study enrolled HAART-naive individuals who visited the clinic at Beijing Ditan Hospital between January 2015 and January 2023. All clinical features were extracted from the electronic medical records. A classification prediction model was established based on various machine learning algorithms, including Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Light Gradient Boosting Machine (LightGBM), to predict the risk of hyperlipidemia based on accuracy, positive-predictive value, negative-predictive value, sensitivity, and specificity. Receiver operating characteristic (ROC) curve, precision-recall curve, and decision curve analyses were used to visually evaluate the model.

RESULTS

A total of 2479 participants (median age, 33 years) were included, of which 2380 (96.01%) were male and 99 (3.99%) were female. The LightGBM model performed the best among all the models in both the training and testing sets. This model performed well in the decision curve analysis (DCA), and baseline high-density lipoprotein cholesterol (HDL-C), baseline triglycerides, baseline viral load, age, albumin, monocyte count, baseline CD4 + cell count, uric acid level, lymphocyte count, and sex were the top 10 predictive risk factors for hyperlipidemia in PWH who started HAART treatment for 6 months, based on SHAP analysis.

CONCLUSION

This study demonstrated that the LightGBM model can effectively predict the risk of hyperlipidemia in PWH after starting HAART treatment for 6 months and reminded physicians closely to monitor serum lipid levels or the timely addition of lipid-lowering drugs, which helped prevent the occurrence of cardiovascular diseases among PWH.