Analysis of Factors Influencing Immunological Failure in Elderly People Living with HIV Undergoing Antiretroviral Therapy: A LASSO-Logistic Regression Approach.

Abstract

This study aimed to identify key predictors of immunological failure in elderly patients with HIV receiving antiretroviral therapy (ART) through machine learning approaches. We conducted a retrospective analysis of 490 elderly patients with HIV (including 120 with immunological failure) treated at Guigang People's Hospital from January 2009 to May 2024, using data extracted from the AIDS Comprehensive Prevention and Control Information System. Employing a two-stage analytical approach, we first applied least absolute shrinkage and selection operator (LASSO) regression to screen 50 potential risk factors, identifying 6 significant predictors. These were subsequently analyzed via logistic regression, revealing six protective factors: moderate disease stage [odds ratio (OR) = 0.401], AIDS stage (OR = 0.130), cotrimoxazole use (OR = 0.495), β2-microglobulin levels (OR = 0.755), platelet count (OR = 0.767), and alanine aminotransferase levels (OR = 0.760). Age was identified as an independent risk factor (OR = 1.275). Finally, the Shapley Additive explanations (SHAP) algorithm was utilized to rank feature importance, providing interpretable insights into predictor contributions. This study used machine learning (LASSO and logistic regression) to pinpoint critical predictors of immunological failure in elderly patients with HIV on ART, aiding early detection of high-risk individuals and informing prevention strategies.