AUTHOR=Liu Chunxia , Pang Ke , Tong Jianbin , Ouyang Wen , Li Liang , Tang Yongzhong 

TITLE=The association between hemoglobin A1c and all-cause mortality in the ICU: A cross-section study based on MIMIC-IV 2.0

JOURNAL=Frontiers in Endocrinology

VOLUME=Volume 14 - 2023

YEAR=2023

URL=https://www.frontiersin.org/journals/endocrinology/articles/10.3389/fendo.2023.1124342

DOI=10.3389/fendo.2023.1124342

ISSN=1664-2392

ABSTRACT=Background: Hyperglycemia has been reported to be associated with the outcomes of patients of patients in ICU. However, the correlation between the hemoglobin A1c and the long-term or short-term mortality in ICU still remains unclear. This study used MIMIC-IV database to explore the relationship between the hemoglobin A1c and the long-term or short-term mortality among patients without diabetes in ICU.
Methods:3154 critically ill patients without diabetes diagnosis before who had hemoglobin A1c measurement were extracted from the Medical Information Mart for Intensive Care (MIMIC) IV and analyzed. The major outcome included the 1-year mortality while the secondary outcomes included the 30-days mortality and 90-days mortality after discharge from ICU. Hemoglobin A1c was divided into four grades (<5.0%, <5.7%, <6.5% and ≥6.5%). COX regression was used to analyze the association between the highest hemoglobin A1c measurement and the mortality. Finally, XGBoost machine learning model and COX regression after propensity score matching (PSM) was used to verify this correlation.
Results: The study ultimately included 3154 critically patients without diabetes from the database who had hemoglobin A1c measurements. After adjusting covariates in COX regression, HbA1c < 5.0% or ≥6.5% were significantly associated with 1-year mortality (HR: 1.37; 95%CI 1.02-1.84 or HR: 1.62; 95%CI 1.20-2.18). And hemoglobin A1c ≥ 6.5% was associated with 30-day mortality (HR: 1.81; 95%CI 1.21-2.71) and 90-day mortality (HR: 1.62; 95%CI 1.14-2.29). The restricted cubic spline proved the u-shape association between hemoglobin A1c values and the 1-year mortality. In the XGBoost model, the AUC of training dataset and testing dataset were 0.928 and 0.826 respectively, while through SHAP plot, hemoglobin A1c was important for the 1-year mortality to some extent. After propensity score matching for other variables, higher level of hemoglobin A1c was still significantly associated with the 1-year mortality in COX regression.  
Conclusions: Hemoglobin A1c is significantly associated with the 1-year mortality, 30-days mortality and 90-days mortality for critically ill patients after discharge from ICU. Hemoglobin A1c < 5.0% and ≥6.5% would increase 30-day, 90-day and 1-year mortality, while medium levels (≥5.0% and < 6.5%) of hemoglobin A1c showed no significant difference in 30-day, 90-day and 1-year mortality.