Introduction

Coronary heart disease (CHD) remains a critical global public health concern, leading to high morbidity and mortality worldwide. It adversely affects health quality and increases healthcare costs, despite advancements in prevention and treatment methods1,2,3. In intensive care units (ICU), CHD patients often face extended stays and high mortality rates, presenting a major challenge to healthcare systems and society4. Identifying CHD patients at high risk is crucial for improving their prognosis.

Stress hyperglycemia, which commonly occurs in severe patients, signals illness severity5. This condition triggers a temporary metabolic response, elevating glucose levels during emergencies, and is linked to poorer clinical outcomes6,7. The association between hyperglycemia and poor outcomes in patients can be partly explained by the amplification of inflammatory and neurohormonal reactions in more serious cases of illness8. The Stress Hyperglycemia Ratio (SHR), by comparing Admission Blood Glucose (ABG) with baseline Glycosylated Hemoglobin A1c (HbA1c), provides a refined assessment of relative hyperglycemia9,10,11,12, offering predictive insights into potential health deterioration8.

In the ICU settings, patients with CHD are often in a high stress state, which may aggravate the condition and affect the prognosis. One study reported that 46.8% of ICU patients had CHD13. These ICU patients with CHD often experience more complications, significantly increasing mortality rates, length of hospital stay, and healthcare costs14. However, research on the prognosis of these critically ill patients with CHD remains limited15. Stress hyperglycemia, that is, an increase in blood sugar under stress conditions such as diseases or trauma, is often regarded as a response of the body to serious illness6,16,17. In ICU patients with CHD, stress hyperglycemia can lead to coronary atherosclerotic plaque instability and rupture, exacerbating myocardial ischemia, which may further result in myocardial infarction, acute heart failure, and malignant arrhythmias, severely impacting patient outcomes18,19. Therefore, the level of stress hyperglycemia can indirectly reflect the severity of diseases and the body’s ability to respond to stresses. Understanding the stress hyperglycemia level can help doctors pay more attention to blood sugar control during treatment, adjust insulin dosage or other hypoglycemic measures in a timely manner, so as to reduce the damage of hyperglycemia to the body and improve the prognosis of patients9,20. In addition, the management of stress hyperglycemia can also help reduce the risk of complications such as infection and multiple organ dysfunction in patients19. Due to the differences in stress response capabilities and severity of diseases, the level of stress hyperglycemia may also vary. By stratifying patients by stress hyperglycemia ratio, doctors can formulate personalized treatment plans for patients at different levels to achieve precision medicine.

Research has consistently shown that higher SHR values are associated with worse outcomes in various cardiac conditions, including acute myocardial infarction (AMI)9,21,22, coronary artery disease20, heart failure (HF)23,24, and ischemic stroke25,26. Nonetheless, the exploration of SHR’s influence on mortality among CHD patients receiving ICU care remains inadequately addressed. This investigation seeks to elucidate the association between SHR and all-cause mortality in this patient group.

Methods

Study population

This retrospective observational study analyzed data from the MIMIC-IV database, which encompasses records of ICU patients at Beth Israel Deaconess Medical Center spanning from 2008 to 201927. The author (Xiaofang Chen) obtained the necessary Collaborative Institutional Training Initiative license and permissions for database access.

The analysis focused on 15,402 individuals, aged 18 or older, diagnosed with CHD and admitted to the ICU on a non-consecutive basis. For those with multiple admissions, only the data on the initial stay was included. Patients lacking first-day admission data for glycosylated HbA1c and glucose were excluded, resulting in a final cohort of 2,059 patients (Fig. 1). This cohort was divided into quartiles based on their SHR on the first ICU day.

Fig. 1
figure 1

Flow chart of patient selection.

Full size image

Data collection

Data was extracted using Structured Query Language (SQL) queries in PostgreSQL (version 14.2), focusing on patient demographics, vital signs, admission severity scores (measured by Acute physiology score III (APSIII), and the Sequential Organ Failure Assessment (SOFA) score), and first 24-hour ICU laboratory results (red blood cell (RBC), white blood cells (WBC), neutrophils, lymphocyte, hemoglobin, platelets, sodium, potassium, total glyceride (TG), total cholesterol (TC), high density lipoprotein cholesterol (HDL-C), low density lipoprotein cholesterol (LDL-C), glucose, HbA1c, albumin, pH, partial pressure of CO2 (PCO2), partial oxygen pressure (PO2), lactate, aspartate aminotransferase (AST), alanine aminotransferase (ALT), blood urea nitrogen (BUN), serum creatinine (Scr), creatine kinase (CK), creatine kinase MB (CKMB), and troponin T (Tnt)).

Comorbidities were delineated using the International Classification of Diseases, 10th Revision (ICD-10) and ICD-9 codes, including CHD, HF, hypertension, atrial fibrillation (AF), dyslipidemia, respiratory failure, stroke, type 2 diabetes, acute renal failure (ARF), chronic kidney disease (CKD), AMI, percutaneous coronary intervention (PCI), coronary artery bypass grafting (CABG) and sepsis. The follow-up commenced at ICU admission and concluded at the patient’s death.

Due to the prevalence of missing data in MIMIC-IV, the study employed multivariate imputation by chained equations (MICE) to mitigate bias in data analysis, adhering to Fully Conditional Specification for imputing each incomplete variable28. Among the variables included in the study, the missing value was 50.7% for LDL-C, 49.0% for HDL-C, 48.1% for TC, 46.3% for BMI, 46.3% for TG, 42.2% for lymphocytes, 42.2% for neutrophils, 38.5% for lactate, 37.8% for PCO2, 37.8% for PO2, 37.0% for Tnt, 36.6% for albumin, 36.2% for PH, 34.5% for CK, 29.2% for CKMB, 23.2% for ALT, 22.9% for AST, and 0.5% for platelets. All the above independent variables were included in the imputation model for imputation modeling. Due to the substantial missing data, we performed a sensitivity analysis comparing imputed and non-imputed datasets. The comparison demonstrated consistent trends between both models (Supplementary Table 1).

Primary outcomes and clinical definitions

The primary outcome of this research was all-cause mortality, encompassing both in-hospital and ICU mortality rates. SHR was determined using the formula: SHR = (admission glucose) (mmol/L) / (1.59 * HbA1c [%] – 2.59)8. CHD was defined as stable angina, unstable angina, myocardial infarction (MI), and ischemic heart disease3. CKD was defined by an estimated glomerular filtration rate of 60 ml/min/1.73 m2or less, as delineated by the CKD-Epidemiology Collaboration (CKD-EPI) formula29. Acute Kidney Injury (AKI) was diagnosed following the Kidney Disease: Improving Global Outcomes (KDIGO) criteria, which stipulate an increase in SCr to ≥ 1.5 times baseline within the previous seven days, an SCr rise of ≥ 0.3 mg/dl within 48 h, or urine output less than 0.5 ml/kg/h for six hours or more30.

Statistical analysis

Continuous data were summarized as means ± standard deviation or medians with interquartile ranges, and were analyzed via the ANOVA analysis. Categorical data, expressed as frequencies and percentages, were compared using Fisher’s exact test or Pearson’s chi-square test as appropriate. For variables with significant differences, the Benjamini-Hochberg method was used to perform post-hoc multiple comparison to correct P values to observe the differences between groups. Survival differences across varying levels of SHR were examined through Kaplan-Meier analysis, while disparities were tested using log-rank methods.

The association between SHR and the primary outcome was quantified using Cox proportional hazard models to estimate hazard ratios (HRs) and 95% confidence intervals (CIs). Several models were adjusted for variables associated with prognosis: Model 1 applied no adjustments; Model 2 was adjusted for demographic and baseline health characteristics, specifically age, sex, and body mass index (BMI); Model 3 was additionally adjusted for variables related to patient health and disease status, encompassing age, sex, BMI, HF, type 2 diabetes, AKI, CKD, AMI, PCI, AF, sepsis, TG, TC, High-Density Lipoprotein (HDL), sodium, potassium, heart rate, and Mean Blood Pressure (MBP). SHR values were analyzed as both continuous and categorical variables, the latter with the lowest quartile serving as a reference. A restricted cubic splines model was leveraged to explore the dose-response relationship between SHR and outcomes.

To assess the prognostic consistency of SHR across various demographics and medical histories, subgroup analyses were executed according to gender, age (above versus equal to or below 65 years), BMI (30 kg/m2 or higher versus lower), and the existence of specific medical conditions, including diabetes, hypertension, AMI, CKD, HF, AF and sepsis. Likelihood ratio tests were implemented to ascertain the link between SHR and these stratification variables.

Data analysis was performed with R software (version 4.2.2), and a two-tailed p-value of less than 0.05 was indicative of statistical significance.

Results

This study included 2,059 patients, with a mean age of 69.92 (SD = 12.88) years; 65.95% (1,358) were male. The average SHR across all patients was 1.20 ± 0.61. In-hospital and ICU mortality rates were 8.50% and 5.25%, respectively.

Baseline characteristics

Participants were stratified into quartiles based on admission SHR levels [Q1: 0.19–0.88; Q2: 0.88–1.07; Q3: 1.07–1.37; Q4: 1.37–15.09). The average SHR values for these groups were 0.74 ± 0.13, 0.97 ± 0.55, 1.20 ± 0.08, and 1.88 ± 0.85, respectively. Patients in the Q4 group had more severe illness indicators, including faster heart rates at admission, and higher incidences of HF, AF, respiratory failure, AKI, CKD, and AMI. This group also showed higher potassium, BUN, and Scr levels, but lower albumin, pH, and PO2 levels compared to those with lower SHR values ((all P < 0.05).

An increase in SHR was associated with higher WBC, neutrophils, glucose, lactate, CK, and Tnt levels, but lower lymphocyte, RBC, and sodium levels (all with p-values less than 0.05). An elevated SHR was associated with extended ICU stays (3.09 vs. 3.29 vs. 4.06 vs. 5.07 days, P = 0.003) and higher ICU (1.94% vs. 3.69% vs. 4.47% vs. 10.87%, P<0.001) and in-hospital mortality (4.47% vs. 6.21% vs. 7.59% vs. 15.73%, P<0.001). Patients in Q4 experienced longer hospital stays compared to those in lower quartiles. More detailed baseline characteristics are provided in Table 1.

Table 1 Baseline characteristics of critical ill patients with CHD grouped according to SHR quartilesa.
Full size table

Primary outcomes

Kaplan–Meier survival analysis demonstrated significant disparities in the in-hospital mortality across SHR quartiles during both 1-month and 3-month follow-up periods (log-rank p <0.001 for both, Fig. 2a and b). Similarly, significant differences in ICU mortality were noted at 1-month and 3-month follow-ups (log-rank p less than 0.001 for both, Fig. 2c and d).

Fig. 2
figure 2

Kaplan–Meier survival analysis curves for all-cause mortality. Footnote SHR quartiles: Q1: 0.19–0.88; Q2: 0.88–1.07; Q3: 1.07–1.37; Q4: 1.37–15.09. Kaplan–Meier curves showing probability of hospital mortality according to groups at 1 month (A), and 3 months (B), and ICU mortality according to groups at 1 month (C), and 3 months (D).

Full size image

The Cox proportional hazards model affirmed a significant correlation between SHR and in-hospital mortality in both Model 1 (HR, 1.17 [95%CI 1.07–1.28] P<0.001) and Model 3 (HR, 1.16 [95%CI 1.02–1.32] P = 0.022) when the SHR was considered a continuous variable. For nominal SHR, significant associations with in-hospital mortality were observed in both Model 1 (Q1 vs. Q2: HR, 1.72 [95% CI 1.01–2.94] P = 0.048; Q3: HR, 1.96 [95% CI 1.17–3.28] P = 0.01; Q4: HR, 3.02 [95% CI 1.90–4.80] P<0.001) and Model 3 (Q1 vs. Q2: HR, 1.62 [95% CI 0.94–2.79] P = 0.080; Q3: HR, 1.71 [95% CI 1.01–2.89] P = 0.045; Q4: HR, 2.67 [95% CI 1.64–4.34] P < 0.001). The risk of in-hospital mortality escalates with increasing SHR (Table 2). A similar trend was noted for ICU mortality. The restricted cubic splines model suggested a J-shaped relationship between SHR and the risks of both in-hospital and ICU mortality (P for non-linearity = 0.002, Fig. 3). This study used the graphical method to prove the premise of the Cox model. It can be observed that the curves on the KM graph did not intersect. Hence, the premise of the Cox model is proved to be valid. Meanwhile, we calculated the AIC values of Model 1-Model 3 when SHR was used as a continuous variable and a nominal variable (Table 3) to avoid possible overfitting. As shown in Table 3, the AIC value of Model3 was the lowest. According to the AIC values, the final adjusted Model 3 was considered the best model.

Table 2 Cox proportional hazard ratios (HR) for all-cause mortality.
Full size table
Fig. 3
figure 3

Restricted cubic spline curve for the SHR hazard ratio. A Restricted cubic spline for hospital mortality. B Restricted cubic spline for ICU mortality. HR, hazard ratio; CI, confdence interval; ICU, intensive care unit; SHR, stress hyperglycemia ratio.

Full size image
Table 3 Akaike Information Criterion (AIC) for Cox proportional hazard ratios (HR).
Full size table

Subgroup analyses

To delve deeper into the impact of SHR on primary outcomes, subgroup analyses were executed based on gender, age, BMI, diabetes, hypertension, AMI, CKD, HF, AF and sepsis (Fig. 4). Notable interactions indicated that SHR was particularly efficient for predicting the risk of mortality in females, individuals older than 65 years, and patients with HF but without AMI, CKD or sepsis.

Fig. 4
figure 4

Forest plots of hazard ratios for the primary endpoint in different subgroups. HR, hazard ratio; CI, confidence interval; BMI, body mass index; HBP, hypertension; AMI, acute myocardial infarction; CKD, chronic kidney disease; HF heart failure; AF atrial fibrillation.

Full size image

Discussion

This study found a strong link between the SHR and in-hospital and ICU mortality rates in patients with CHD. An elevated SHR was associated with extended ICU stays and higher ICU and in-hospital mortality. Notably, patients in the Q4 had significantly longer hospital stays than those in other groups. Our subgroup analysis revealed that SHR consistently predicted mortality in patients, including those with HF, highlighting its stable prognostic value in CHD.

This research is pioneering in analyzing the relationship of SHR with all-cause mortality among CHD patients using a U.S. public critical care database. The findings underscore SHR as a reliable independent predictor of higher mortality in these patients, even when considering various confounding factors. This study introduces a direct, effective method for assessing the risk of stress hyperglycemia among CHD patients admitted to the ICU.

Hospitalized patients with hyperglycemia, particularly those admitted for HF, MI, ischemic stroke, or severe illness, often face increased mortality risks31,32,33,34,35,36,37,38. Hyperglycemia can signal a temporary physiological response to acute illness, defined by a relative glucose level rise due to inflammatory and neurohormonal disruptions from severe conditions. This stress response involves the hypothalamic-pituitary-adrenal axis and sympathoadrenal system, increasing proinflammatory cytokine release and inducing stress hyperglycemia5.

Previous research has debated the reliability of admission glucose levels as an indicator of stress hyperglycemia, especially in diabetic patients with poor glycemic control39. Studies by Gregory W. et al.8 and Marenzi et al.40 have shown that SHR, unlike admission glucose levels, is significantly associated with critical illness and provides a better prediction of in-hospital mortality in AMI cases. SHR offers a new perspective on the link between hyperglycemia and clinical outcomes by adjusting glucose levels to HbA1c. Given the widespread availability of glucose and HbA1c testing, this simple calculation can provide valuable prognostic information for hospitalized patients.

Numerous studies have shown a strong link between the SHR and poor outcomes. Stress hyperglycemia is a proven marker for elevated mortality risks in AMI patients31,37,41,42. Additionally, evidence indicates that in non-obstructive MI, stress hyperglycemia correlates with more extensive myocardial cell death and worse prognosis, both short-term and long-term43,44. Research by Li et al.45has identified a U-shaped relationship between SHR and mortality in critical conditions, suggesting both high and low SHR levels worsen the prognosis of acute decompensated HF and diabetes, as noted by Zhou et al. Sia et al. highlighted SHR as a key independent factor for one-year mortality in ST-segment elevation myocardial infarction (STEMI) cases, irrespective of diabetes status46. Xu et al. have demonstrated the link between SHR and 30-day mortality in STEMI patients, noting that adding SHR to the Thrombolysis in Myocardial Infarction (TIMI) risk score improves its predictive power47. Zhang et al. have identified increased SHR as an independent risk factor for mortality in ICU patients48.

The underlying mechanism of the connection between SHR and mortality may involve diminished endothelium-dependent vasodilation, reduced platelet anti-aggregation, and increased sympathetic nerve activity due to inflammation49,50,51. Stress hyperglycemia contributes to oxidative stress and inflammation, worsening endothelial dysfunction6, thrombosis, and ischemia-reperfusion injury, thereby increasing myocardial damage21. It also promotes a pro-thrombotic state and activates the neuroendocrine system, increasing catecholamines and cytokines, which impair vascular endothelial function44,52. These mechanisms help explain the connection between hyperglycemia and adverse outcomes, especially in severe illnesses, by exacerbating inflammatory and neurohormonal responses [8].

Our analysis substantiates SHR as an independent indicator of mortality in ICU patients with CHD. Our findings can facilitate the early identification of high-risk CHD patients within ICU. Strict glycemic control is required to improve outcomes in this high-morbidity and high-mortality population.

Strengths and limitations

The principal strength of this investigation lies in confirming the SHR as a significant independent predictor of mortality in critically ill CHD patients upon ICU admission in a U.S. cohort. However, given the retrospective and observational design of the study and the inherent limitations of the MIMIC-IV database, it is impossible to examine causal inferences and account for variables such as disease severity, baseline characteristics, admission diagnoses, and participants’ socioeconomic status, potentially leading to biased results. Moreover, as a monocentric study with a relatively constrained sample size, despite comprehensive multivariate adjustments and subgroup analyses, this study may not fully eliminate bias from unconsidered confounding factors. The analysis emphasizes the baseline prognostic value of SHR in CHD without examining SHR fluctuations during hospital stays. Hence, further investigation is warranted to ascertain whether SHR changes could be used to predict mortality. Prospective cohort studies are needed to confirm our findings.

Conclusions

This study underscores the utility of SHR as a prognostic tool for forecasting in-hospital and ICU mortality among critically ill CHD patients. Further research, particularly randomized studies, is required to examine how glycemic control based on SHR might enhance outcomes in those with severe CHD.