Association between hemoglobin glycation index and mortality in critically ill patients with intracerebral hemorrhage: a cohort study using the MIMIC-IV database

Abstract Intracerebral hemorrhage (ICH) carries high mortality in the neurocritical care setting, and metabolic risk stratification remains challenging. Conventional glycemic markers, including acute glucose and HbA1c, may be influenced by stress hyperglycemia and altered erythrocyte dynamics in critical illness. The hemoglobin glycation index (HGI), defined as the discrepancy between measured HbA1c and glucose-predicted HbA1c, may capture glycation–glucose discordance, but its relevance in critically ill ICH patients remains unclear. We performed a retrospective cohort study of 1,688 critically ill patients with ICH from the MIMIC-IV database. HGI was calculated as measured HbA1c minus predicted HbA1c derived from the first available baseline glucose measurement and categorized into quartiles. The main outcome was 365-day all-cause mortality after ICU admission. Associations between HGI and 365-day mortality were evaluated using Cox proportional hazards models. Dose–response relationships were examined with restricted cubic splines. Very low HGI was associated with higher mortality across multiple time horizons. In the fully adjusted model, compared with Q1, the adjusted hazard ratios for 7-day mortality were 0.41 (95% CI, 0.25–0.66), 0.60 (0.38–0.94), and 0.41 (0.25–0.66) for Q2, Q3, and Q4, respectively. Similar inverse associations were observed for 14-, 28-, 90-, and 365-day mortality. Restricted cubic spline analyses showed an overall inverse association between HGI and mortality, with little evidence of strong nonlinearity. In critically ill patients with ICH, very low HGI was associated with increased mortality. This finding may reflect a high-risk hematologic–metabolic phenotype involving acute stress hyperglycemia and altered HbA1c interpretation rather than a direct protective effect of higher HGI. HGI may serve as an exploratory adjunct marker for risk stratification, but further validation with detailed glucose timing, erythrocyte-related factors, and transfusion data is warranted.