A comparative analysis of the cholesterol–high-density lipoprotein–glucose index and the triglyceride–glucose index in predicting in-hospital mortality in critically ill ischemic stroke patients

Background The Cholesterol, high-density lipoprotein, and glucose (CHG) index has emerged as a potential indicator of metabolic disturbance, but its prognostic value in patients with ischemic stroke (IS) remains unclear. This study aimed to assess whether the CHG index could predict 28-day in-hospital mortality in critically ill IS patients and to compare its performance with the established triglyceride–glucose (TyG) index. Methods We conducted a cohort analysis using data from the eICU database, involving 1,670 critically ill patients diagnosed with IS between 2014 and 2015. CHG and TyG indices were computed for each patient. Their associations with 28-day in-hospital mortality were examined using multivariable Cox regression. To further investigate the associations, restricted cubic spline (RCS) analysis was conducted. Kaplan–Meier curves were used to compare outcomes across different TyG and CHG groups. Predictive accuracy was compared using receiver operating characteristic (ROC) analysis. Subgroup analyses were performed to assess consistency across different clinical characteristics. Results Among the study population, 158 (9.46%) patients died within 28 days of hospitalization. The CHG index showed a greater association with mortality (HR 1.554; 95% CI 1.198–2.018; p < 0.001) compared to the TyG index (HR 1.436; 95% CI 1.175–1.755; p < 0.001) in unadjusted models, and both remained significant after adjustment. RCS analysis demonstrated a linear relationship between both indices and 28-day in-hospital mortality. ROC curves showed similar discriminatory ability for the CHG and TyG indices. No significant interactions were observed in subgroup analyses ( p > 0.05; p for interaction >0.05). Conclusions Higher CHG index values are independently associated with increased 28-day mortality in critically ill IS patients, showing a linear relationship and predictive performance comparable to that of the TyG index.