The triglyceride high-density lipoprotein cholesterol-glucose body mass index (TyHGB) is a novel composite indicator integrating triglyceride, high-density lipoprotein cholesterol, fasting glucose and body mass index and may provide a more comprehensive assessment of systemic metabolic status than current indices. This multicenter landmark cohort study aimed to determine the association between TyHGB measured at the landmark time and 28-day mortality in patients present in the intensive care units (ICUs) at that time and to evaluate its potential utility as a prognostic marker.

This multicenter observational cohort study included 1649 patients from 113 ICUs in China. The association between TyHGB and 28-day mortality was evaluated using three sequentially adjusted logistic regression models with restricted cubic splines: Model 1 adjusted for demographics; Model 2 further adjusted for disease severity and organ dysfunction; Model 3 additionally adjusted for therapeutic interventions. Subgroup analyses with interaction tests were conducted. Model performance was assessed by discrimination (AUC), calibration (Brier score and calibration slope) and internal validation via bootstrap resamples.

Restricted cubic spline analysis demonstrated a positive association between TyHGB and 28-day mortality across fully adjusted models (P < 0.05). Subgroup analyses identified a statistically significant interaction with diabetes (P for interaction = 0.011). In the overall cohort, adding TyHGB to the base model improved AUC from 0.731 to 0.745 (ΔAUC = +0.014, P = 0.043), with significant improvements in reclassification (NRI = 0.328, P < 0.001; IDI = 0.012, P = 0.009) and model fit (LRT P < 0.001). Among diabetic patients (N = 401), TyHGB significantly improved AUC from 0.673 to 0.723 (ΔAUC = +0.050, P = 0.049; NRI = 0.515, P < 0.001; IDI = 0.041, P = 0.022). In contrast, no significant improvement was observed in non-diabetic patients (ΔAUC = +0.009, P = 0.152).

Among patients present in participating ICUs at the landmark time, higher TyHGB was associated with subsequent 28-day mortality. However, adding TyHGB to the baseline risk prediction model led to only small improvements in overall predictive performance. Findings in diabetic patients are exploratory and require further validation.