Identifying important risk factors for survival of HIV-infected patients using censored quantile regression

Abstract

Introduction:This study aimed to estimate the effect of potential risk factors on survival of human immunodeficiency virus/acquired immunodeficiency syndrome (AIDS) patients using censored quintile regression model. Material and methods:We used a dataset from a (registry-based) retrospective cohort study conducted in Tehran (from April, 2004 to March, 2018. Demographic information, such as age, sex, marital status, and educational level as well as behavioral information, including being-in-prison, drug/alcohol abuse and smoking, antiretroviral therapy, co-infection with tuberculosis (TB), and CD4+ cell count, were investigated as potential risk factors for AIDS progression. Censored quintile regression was used to estimate and investigate these factors for AIDS progression.Results: Mean age of patients was 33.93 years. Time to progression ranged from 0.01 to 223.17 months, and mean of time to progression was 40.55 months. A total of 1,249 (50.5%) patients experienced an event by end of the study. Impact of age, gender, prison, being addicted, being infected with tuberculosis, and using highly active antiretroviral therapy (HAART) were significant for most of quintiles (p < 0.05). Conclusions:It was shown that age, being prisoned, TB infection, and HAART were significantly associated with a lower time in AIDS progression. Censored quintile regression could be an appropriate choice for considering time-varying effects and easy interpretation of regression coefficients in analyzing AIDS progression data.

Determining the Level of Importance of Variables in Predicting Kidney Transplant Survival Based on a Novel Ranking Method

“…They presented a Bayesian risk model by combining the variables identified from 3 different approaches (medical literature, data mining methods, and elastic net regression). Hamidi et al, 22 using the RF technique, identified the important risk factors in kidney transplant patients. They identified cold ischemic time, recipient age, creatinine level at discharge, and donors' age as the most important factors.…”

Joint prediction scheme of pulmonary infection risk after renal transplantation

“…In addition to traditional data mining methods, in recent years, some intelligent data mining methods often utilized in machine learning have also been applied to this prediction field [8]. For example, some data-driven random survival forest mining approaches are proposed in [9], [11], [12]. Moreover, some decision-tree-based schemes [16], proposed to determine the risk factors in lung infection, could also been found in [17][18][19].…”