Synergistic predictive value of dynamic glycemic trajectories and variability metrics for 28-day mortality in critically ill heart failure

Glucose dynamics is one of the unique mechanisms in patients with critically ill heart failure (HF). The aim of this study is to evaluate the impact of dynamic blood glucose trajectories on 28-day mortality in critically ill HF patients. Latent Category Growth Model (LCGM) was used to classify patients’ blood glucose trajectories during the first 4 days of intensive care unit (ICU) admission. Kaplan-Meier survival analysis and Cox regression assessed the association between admission blood glucose levels, glucose trajectories, and 28-day mortality in critically ill HF patients. Subgroup analyses evaluated the robustness of the findings. A total of 6062 patients with critically ill HF were included in this retrospective cohort study, with 28-day mortality occurring in 1306 (21.54%) patients. The Kaplan Meier survival curve shows that the survival probabilities of different blood glucose trajectories from high to low are: class 1 > class 3 > class 2 > class 4, and there are significant inter class differences. COX regression confirms that the predictive ability of blood glucose trajectory classification for mortality in patients with critically ill HF is superior to the blood glucose coefficient of variation. Subgroup analysis further evaluated the consistency of the association between blood glucose latent trajectory classification and 28-day mortality in different patient characteristics. Dynamic blood glucose trajectories and variability indicators provide complementary information for predicting 28-day mortality in critically ill HF patients.