Abstract
OBJECTIVES: This study aimed to evaluate the association between the stress hyperglycaemia ratio (SHR) and the Haemoglobin Glycation Index (HGI), and mortality risks in critically ill patients. DESIGN: A retrospective study and machine learning (ML)-based predictive modelling. SETTING: This retrospective cohort study used data from the Medical Information Mart for Intensive Care-IV (MIMIC-IV) database. PARTICIPANTS: A total of 3106 patients were included in the study and were divided into different groups according to the value level of SHR and HGI. PRIMARY OUTCOME MEASURE: 360-day mortality. RESULTS: When treated the SHR as a continuous variable, a significant correlation exists between the SHR and 360-day mortality risks in critically ill patients (HR, 1.32; 95% CI 1.13 to 1.55). When regarded the SHR as a categorical variable, patients in the highest group were significantly associated with an increased risk of 360-day mortality compared with that of those in the lowest group (HR, 1.38; 95% CI 1.14 to 1.68). HGI, when treated as a continuous variable, was also closely associated with 360-day mortality (HR, 0.94; 95% CI 0.89 to 1.00). According to the results, the SHR index outperformed HGI at predicting all-cause 360-day mortality and adding the SHR index to the basic model for 360-day mortality improved its predictive ability (area under the curve, 0.818 for the basic model vs 0.821 for the basic model+SHR index). Furthermore, the ML-based model demonstrated the crucial contribution of SHR in predicting 360-day mortality risk of critically ill patients. Consistent with the 360-day mortality results, similar and statistically significant trends towards higher mortality at both 30 days and 90 days were observed. CONCLUSIONS: SHR and HGI showed a strong association with 360-day and short-term mortality risks. The SHR index appears to be the most promising index for prevention and risk stratification in critically ill patients.