Adaptive LASSO-Penalized Quantile Regression for Identifying Risk Factors of Left Ventricular Hypertrophy

Abstract

Left ventricular hypertrophy (LVH) is a major cardiovascular disorder associated with increased morbidity and mortality. Identifying its clinical risk factors requires statistical methods capable of handling heterogeneous medical data. This study applies the Adaptive LASSO-penalized quantile regression model, which combines robust quantile estimation with automatic variable selection.
The analysis was conducted using clinical data from 120 patients collected at the Baghdad Heart Center, Iraq, between January and December 2023. The Left Ventricular Mass Index (LVMI) was considered as the response variable, while ten demographic and clinical variables were included as predictors. The model was estimated at three quantile levels τ = 0.25,τ = 0.50,and τ = 0.75 using K-fold cross-validation for selecting the regularization parameter.
The results showed that Adaptive LASSO successfully identified the most influential predictors while producing sparse and interpretable models. Age, body mass index, systolic blood pressure, and diabetes mellitus were consistently selected across all quantiles, whereas other predictors became significant only at higher quantiles, indicating heterogeneous effects across different levels of disease severity.
These findings demonstrate that Adaptive LASSO-penalized quantile regression is an effective approach for identifying quantile-specific risk factors of left ventricular hypertrophy and provides a valuable statistical tool for analyzing heterogeneous cardiovascular data.