Stress hyperglycemia ratio is a risk factor for mortality in trauma and surgical intensive care patients: a retrospective cohort study from the MIMIC-IV

Abstract

Background

Stress hyperglycemia ratio (SHR) can reduce the impact of baseline glucose on the stress hyperglycemia level. Studies have shown that SHR is associated with adverse outcomes. However, its relationship with the prognosis of trauma/surgical ICU patients has not been fully studied. The objective of this study was to explore the relationship between SHR and the short-term and long-term mortality in trauma/surgical ICU patients.

Methods

Clinical data of trauma/surgical ICU patients were extracted from MIMIC-IV. The primary outcome was 28-day all-cause mortality, and the secondary outcome was 365-day all-cause mortality. Boruta algorithm was used to screen the important features related to the 28-day mortality, and Kaplan–Meier curve, Cox proportional hazards regression, and restricted cubic spline were used to explore the relationship between SHR and clinical outcomes.

Results

A total of 1744 patients were included, of whom 786 were male and 958 were female. The 28-day and 365-day mortality rates were 14.7% and 27.2%, respectively. Multivariate Cox proportional hazards analysis showed that an increase in SHR was significantly associated with an increased risk of 28-day mortality [HR (95% CI) 1.30 (1.07, 1.58), p = 0.009] and 365-day mortality [HR (95% CI) 1.05 (1.02–1.09), p = 0.005]. Restricted cubic spline curve showed that the relationship between SHR and survival rate was "U-shaped".

Conclusions

Increase in SHR is associated with an increased risk of 28-day and 365-day all-cause mortality in trauma/surgical ICU patients.

Introduction

Stress hyperglycemia refers to the situation where the body experiences an increase in blood glucose levels in response to stressors, which is a common physiological manifestation in critically ill patients [1]. Studies have shown that hyperglycemia may contribute to adverse outcomes through mechanisms such as oxidative stress, inflammatory response and vascular endothelial dysfunction [2]. However, the elevated blood glucose level in diabetes patients admitted to the intensive care unit (ICU) may be caused by physiological responses to severe illness, chronic poor glycemic control, or both, which is difficult to distinguish clinically [3]. In 2015, Robert et al. [4] proposed the calculation of the stress hyperglycemia ratio (SHR) based on absolute levels of blood glucose and glycated hemoglobin (HbA1c) to mitigate the impact of baseline blood glucose levels on stress-induced hyperglycemia values. Since its proposal, SHR has been confirmed to be associated with adverse outcomes in various diseases [5,6,7,8,9].

Trauma and surgical critical illness patients often experience stress-induced hyperglycemia, possibly related to excessive activity of the sympathetic nervous system and insulin resistance [10]. However, it remains unclear whether SHR is associated with adverse outcomes in trauma and surgical critical patients. Therefore, this study aims to explore the relationship between SHR and short-term as well as long-term adverse outcomes in trauma and surgical critical patients.

Methods

Data source

This study used data from MIMIC-IV (version 2.2), which contains over 50,000 patients who were admitted to the ICU at Beth Israel Deaconess Medical Center (BIDMC) in Boston, Massachusetts, from 2008 to 2019 [11]. The Institutional Review Board at BIDMC approved informed consent waiver and resource sharing. Yingying Zhang obtained access to the database (certificate number: 64373576).

Patients

The subjects in this study were critically ill patients admitted to the surgical ICU or trauma ICU, excluding those who met the following criteria: (1) age < 18 years; (2) ICU stay duration ≤ 24 h; and (3) lack of glucose or HbA1c, unable to calculate SHR; in the end, 1744 patients were enrolled and divided into four groups based on the quartiles of SHR, as shown in Fig. 1.

Fig. 1

Inclusion flowchart

Data extraction

Structured query language (SQL) was used to obtain patient information, including gender, age; vital signs; comorbidities, including the Charlson Comorbidity Index; laboratory tests such as white blood cells, hemoglobin, platelets, creatinine, urea nitrogen, and the mean and maximum blood glucose value; medicines such as corticosteroid and insulin use; ICU events such as whether mechanical ventilation was used, whether continuous renal replacement therapy (CRRT) was used, and whether vasopressor drugs were used; Simplified Acute Physiology Score (SAPSIII); patient outcomes, including length of hospital stay, length of ICU stay, in-hospital mortality, ICU mortality, 28-day mortality, and 365-day mortality, as shown in Table 1. Follow-up began from the day of admission to ICU. All laboratory variables, medicines and disease severity scores were extracted from data collected within 24 h of the patient entering ICU. SHR was calculated using the following formula: SHR = (admission glucose (mg/dl)) / (28.7 × HbA1c (%)—46.7) [4].

Table 1 Patients’ demographics and baseline characteristics categorized by SHR

Outcomes

The primary outcome of this study is the all-cause mortality within 28 days after admission to ICU, representing the short-term prognosis, and the secondary outcome is the all-cause mortality within 365 days after admission to ICU, representing the long-term prognosis of patients.

Statistical analysis

Patients were stratified into four groups based on the quartiles of SHR. The Kolmogorov–Smimov test was employed to assess the normality of continuous variables. If the data followed a normal distribution, they were presented as mean ± standard deviation; otherwise, they were expressed as median (quartiles). Categorical variables were reported as counts (percentages). One-way analysis of variance (ANOVA) was utilized for normally distributed continuous variables, while non-normally distributed data underwent Kruskal–Wallis test. Chi-square test or Fisher's exact test was applied for categorical variables. Variables with missing rates exceeding 20% were excluded, and those with missing rates at or below 20% underwent random forest multiple imputation. Multicollinearity was assessed using variance inflation factor (VIF), and variables with VIF ≥ 5 was removed.

Prior to investigating the relationship between SHR and patient prognosis, feature selection using Boruta algorithm [12] was conducted to ascertain the predictive importance of SHR in 28-day mortality risk assessment. Kaplan–Meier survival analysis compared short-term and long-term survival rates among different SHR groups, with inter-group differences evaluated by log-rank test. Cox proportional hazards regression model assessed hazard ratios (HRs) and corresponding 95% confidence intervals (CIs) for both continuous and categorical forms of SHR in relation to 28-day and 365-day all-cause mortality outcomes. Confounding variables encompassed factors derived from clinical expertise, prior research findings, as well as important and suspicious features identified through Boruta feature selection. The trend test explores whether there is a linear correlation between outcome variables and SHR quartiles. Model 1 did not adjust for confounding variables; Model 2 adjusted for patient sex, age, height, and weight; Model 3 further adjusted for demographic characteristics including age and weight; vital signs such as heart rate, blood pressure, respiratory rate; laboratory tests including hemoglobin, platelets, oxygenation index, mean glucose value and serum sodium; ICU events like mechanical ventilation, CRTT, and vasopressor drug administration; scoring systems comprising APSIII score and Charlson comorbidity index.

Restricted cubic spline (RCS) tests employing four knots examined potential nonlinear relationships between changes in SHR levels and survival rates. Subgroup analyses based on gender, younger age (< 65 years) vs older age (≥ 65 years), and presence/absence of diabetes mellitus were performed to ensure consistency in the prognostic value of SHR indices for short-term and long-term outcomes. Two-sided p values less than 0.05 were considered statistically significant. R software (version 4.3.1) and SPSS software (version 24.0) were used for statistical analysis.

Results

Baseline data

A total of 1744 patients were included in this study (Fig. 1). According to the SHR (quartile Q1 0.22–0.87; Q2 0.87–1.04; Q3 1.04–1.30; Q4 1.30–3.32), patients were divided into 4 groups. The results of one-way ANOVA suggest that there were statistically significant differences among the four groups in gender, heart rate, systolic blood pressure, white blood cell count, hemoglobin, urea nitrogen, PH, lactate value, sodium ion concentration, potassium ion concentration, calcium ion concentration, the mean and maximum glucose values, APS III score, proportion of patients with cerebrovascular disease, diabetes, mechanical ventilation, CRRT, and use of vasoactive drugs, insulin and corticosteroid (p < 0.05) (Fig. 1).

Comparing the outcomes of all groups, it was found that there were statistically significant differences in length of hospital stay and ICU stay, in-hospital mortality, in-ICU mortality, 28-day mortality and 365-day mortality among all groups. The quartile 4 group had the longest hospital stay, ICU stay and the highest mortality at each timepoint (p < 0.05) (Table 1).

Boruta feature selection

As shown in Fig. 2, the Boruta results suggest that age, APS III score, use of vasoactive drugs, mean glucose level, SHR, and Charlson comorbidity index are important features associated with 28-day mortality, while oxygenation index, serum sodium, hemoglobin, and weight are suspicious features associated with 28-day mortality.

Fig. 2

Boruta algorithm feature selection results. The horizontal axis shows the names of each variable, and the vertical axis shows the Z-score of each variable. The boxplot shows the Z-score of each variable during the model calculation process. The green frame indicates important variables, the yellow frame indicates suspicious variables, and the red frame indicates unimportant variables. Ca calcium ion, CRRT continuous renal replacement therapy, SCr serum creatinine, APTT activated partial thromboplastin time, BUN blood urea nitrogen, PLT platelet, Lac lactate, PT partial thromboplastin time, TBL total bilirubin, HR heart rate, K potassium ion, DBP diastolic blood pressure, SBP systolic blood pressure, WBC white blood cell, Hb hemoglobin, Na sodium ion, Spo₂/fio₂ oxygenation index, SHR stress hyperglycemia ratio

The relationship between SHR and all-cause mortality

Kaplan–Meier survival curves were used to compare the 28-day and 365-day survival probability between groups. Results showed that the 28-day and 365-day survival probability of the fourth group were the lowest, and the difference was statistically significant (log-rank p < 0.001) (Fig. 3).

Fig. 3

A 28-day KM survival curve; B 365-day KM survival curve. SHR: Quartiles (0.22–0.87), Quartiles 2 (0.87–1.04), Quartiles 3 (1.04–1.30), Quartiles 4 (1.30–3.32)

The results of Cox proportional hazards regression showed that when SHR was treated as a continuous variable, the 28-day and 365-day mortality risks increased significantly as the SHR value increased by one unit (p < 0.05); when SHR was treated as a categorical variable, the 28-day and 365-day mortality risks of group 4 were significantly higher than group 1 (p < 0.05); the trend test results (p for trend) suggested that the 28-day and 365-day mortality risks increased with the increase of SHR group in models 1 and 2, and the difference was statistically significant (p < 0.05). However, there was no statistically trend in model 3 (p > 0.05) (Table 2).

Table 2 Cox proportional hazard ratios (HR) for all-cause mortality

Dose–response analysis of SHR and all-cause mortality

In the restricted cubic spline, the confounding variables from Model 3 were adjusted. Results showed that there was a nonlinear relationship between SHR and 28-day and 365-day all-cause mortality rates, presenting a U-shaped association (Fig. 4). The inflection point of the RCS curve was at SHR = 1.04.

Fig. 4

A RCS analysis of all-cause mortality at 28 days; B RCS analysis of all-cause mortality at 365 days. The curves represent the estimated adjusted risk ratios, and the shaded areas represent the 95% confidence intervals. The horizontal dashed line represents the risk ratio of 1

Subgroup analysis

Subgroup analysis was conducted based on patient's gender, age, and diabetes. Confounding variables from Model 3 were adjusted. Results showed that an increase in SHR was associated with a higher risk of 28-day and 365-day mortality in male, < 65 years, non-diabetic, and diabetic patients (p < 0.05), while an increase in SHR was only associated with a higher risk of 28-day mortality in patients aged ≥ 65 years (p < 0.05) (Fig. 5).

Fig. 5

Subgroup forest plots of all-cause mortality at 28 days and 365 days. HR risk ratio, CI confidence interval

Discussion

Main findings

The results of this study suggest that elevated SHR is associated with increased short-term and long-term mortality risk in trauma and surgical ICU patients, and the association remains significant after adjusting for multiple variables. The subgroups analysis indicates that the relationship between SHR and prognosis is not affected by the presence of diabetes.

Relation with previous evidence

Traumatized patients with high blood glucose upon admission have a significantly increased risk of death [13, 14]. In a retrospective evaluated 1799 patients in the surgical care unit found that the average blood glucose of patients who died in hospital was significantly higher than that of patients who survived, suggesting that hyperglycemia was associated with an increased risk of death of patients in the surgical care unit [15]. However, there is no study explores the relationship between SHR and prognosis in trauma and surgical critical care patients.

Robert et al. [4] found that an elevated SHR at admission was associated with an increased risk of adverse events such as transfer to the ICU and in-hospital mortality, and proposed that SHR could be used as a predictor of adverse outcomes in hospitalized patients. Subsequently, the relationship between SHR and various diseases, especially cardiovascular diseases, was extensively studied. Liu et al. [16] found that SHR was associated with the long-term all-cause mortality rate of patients with acute myocardial infarction in the critical condition. When SHR was greater than 1.3, the patients' 1-year all-cause mortality and long-term all-cause mortality risk significantly increased, and the conclusion remained consistent across different ethnic groups. He et al. [17] found that SHR was associated with increased in-hospital mortality and 1-year mortality risk in patients with severe coronary artery disease. Subgroup analysis suggested that the conclusion was consistent in diabetic patients and non-diabetic patients. Yang et al. [5] also found that SHR was associated with increased 2-year major adverse cardiovascular and cerebrovascular events and all-cause mortality risk in patients undergoing drug-eluting stent implantation for acute coronary syndrome. Yan et al. [18] studied the correlation between SHR and 28-day all-cause mortality in patients with severe sepsis, and the results showed that SHR was an independent predictor of 28-day all-cause mortality in severe sepsis, and the results were consistent in diabetic and non-diabetic patients, further expanding the applicable population of SHR. Our study found that high SHR was associated with increased short-term and long-term all-cause mortality risk in surgical and trauma ICU patients, and the results were not affected by whether the patients had diabetes, consistent with previous research findings.

Possible explanations for findings

When the body is in a state of stress, such as major surgery or trauma, the hypothalamic–pituitary–adrenal cortex axis (HPA) and sympathetic adrenal medulla axis are activated, and amount of cortisol, glucagon and catecholamine are released. Meanwhile, cytokine release and insulin resistance can lead to stress hyperglycemia [19,20,21]. By introducing HbA1c to correct past blood glucose levels, the SHR can be better optimized and evaluated in critically ill patients during stress states, thus making it more reasonable to study the relationship between SHR and disease prognosis [22]. The specific mechanism by which SHR increases the risk of death is unclear, and may be attributed to the inflammatory response caused by high blood sugar. First, higher SHR is the result of complex interactions between various hormones and cytokines, including interleukin-6 and tumor necrosis factor-α, which may lead to increased mitochondrial reactive oxygen species production in endothelial cells, thereby leading to endothelial dysfunction [23, 24]. Second, stress-induced hyperglycemia can cause non-enzymatic glycation of platelet glycoproteins, promoting platelet activation and increasing the risk of thrombosis [25]. Third, high blood sugar can activate the immune system and increase the risk of infection [26], and prospective cohort studies have shown that an increase in SHR is positively associated with the occurrence of pulmonary infections during the hospital stay of patients with acute myocardial infarction [27]. The above mechanisms may lead to an increased risk of death in surgical and trauma patients.

Implications for clinical practice

The positive correlation between SHR and poor prognosis in trauma and surgical patients was found for the first time, emphasizing the clinical significance of assessing SHR in surgical and trauma patients admitted to ICU, but whether refined management of SHR is associated with improved outcomes needs further research.

Study limitations

Our study has some limitations. First, limited by retrospective study design, the results may be influenced by some unmeasured confounders. Although we try to collect data comprehensively, there are still some confounders cannot be obtained or analyzed. For example, we cannot distinguish whether glucose injection is used as a drug solvent or for other therapeutic purposes (hypoglycemia, high potassium, high sodium, etc.), so we did not include this factor in the Cox proportional hazards regression model. Second, we cannot determine whether there is a causal relationship between SHR and increased mortality, further studies should be performed to explore the cause–effect relationship between SHR and the prognosis of trauma and surgical ICU patients.

Conclusions

After adjusting for multiple confounding variables, a higher SHR was associated with an increased short-term and long-term mortality risk in surgical and trauma ICU patients, which can be used to predict poor prognosis in surgical and trauma ICU patients.