Abstract
Purpose
To investigate the association between the stress hyperglycemia ratio (SHR) and short-term mortality, including 28-day and ICU mortality, in patients with traumatic brain injury (TBI).
Key methods
Data from the MIMIC-IV and eICU databases were used as the training and validation cohorts, respectively. Patients were stratified into SHR quartiles. Cox regression models (adjusted for age, sex, SOFA, APS III, heart failure, GCS, ventilator duration, neurosurgical intervention, and injury-to-ICU time) and Kaplan–Meier curves were applied to assess associations between SHR and mortality. Restricted cubic spline (RCS) models examined dose–response patterns, and subgroup analyses evaluated potential effect modification.
Results
In the MIMIC-IV cohort, SHR was positively associated with SOFA and APS III scores (both P < 0.001). Cox regression and Kaplan–Meier analyses demonstrated that higher SHR was significantly associated with increased 28-day all-cause mortality (fully adjusted HR for highest vs. lowest quartile: 1.83, 95% CI 1.18–2.87). Similar associations were observed in the external eICU cohort (adjusted HR: 2.32, 95% CI 1.01–5.36). Restricted cubic spline analysis showed a nonlinear relationship between SHR and 28-day mortality, with a mathematical inflection point around SHR ≈ 1.3.
Subgroup analyses indicated that the association between SHR and 28-day mortality was stronger in male patients (HR 2.01, 95% CI 1.23–3.29) and in those with lower APS III scores (HR 4.02, 95% CI 1.39–11.64) (both P for interaction < 0.05).In contrast, although crude Kaplan–Meier curves showed higher ICU mortality in the upper SHR quartiles, the association between SHR and ICU mortality was not statistically significant in fully adjusted Cox models in either cohort.
Conclusion
SHR was strongly associated with 28-day all-cause mortality in TBI patients. However, its association with ICU mortality was not statistically significant after adjustment and should be considered inconclusive.
Supplementary Information
The online version contains supplementary material available at 10.1186/s12883-025-04602-w.
Keywords: Traumatic brain injury, Stress hyperglycemia ratio, ICU admission, 28-day all-cause mortality rate, Cubic spline analysis
Introduction
Traumatic brain injury (TBI) is a condition caused by various factors that disrupt brain structure and function, leading to symptoms such as lethargy, coma, and impaired consciousness in most of TBI. In severe cases, TBI can result in cardiorespiratory arrest. The global incidence of TBI exceeds 50,000,000 cases, with more than 30% of patients requiring surgical treatment and more than 10% requiring ICU monitoring treatment [1–4]. This has resulted in significant social and economic burdens.
When TBI occurs, it is often accompanied by severe stress responses, which stimulate the sympathetic nervous system and simultaneously inhibit the release of insulin. This can lead to rapid fluctuations in blood sugar levels within a short period of time.Multiple studies have demonstrated that rapid fluctuations in blood glucose levels, rather than consistently high blood sugar, are more likely to increase mortality in cardiovascular patients, increase cognitive dysfunction, and lead to organ failure [5–7]. Stress-induced hyperglycemia is a key contributor to these rapid blood sugar fluctuations [8]. In cardiovascular disease, stress hyperglycemia has been shown to predict both short-term and long-term survival [9, 10]. However, blood sugar measured at admission often reflects the patient’s acute stress response and pre-existing chronic blood sugar level, which may not fully capture stress-induced changes. The stress-induced hyperglycemia ratio (SHR) is considered not only a more accurate measure of stress-induced blood sugar changes but also could reflect changes in neuroendocrine function of some diseases. It is widely used to predict outcomes in patients with cardiovascular disease and sepsis [11–16]. Unfortunately, there is a lack of studies on the role of SHR in evaluating the severity of TBI patients.
Accordingly, the purpose of the current study was to assess the relationship between SHR and 28-day and ICU all-cause mortality in TBI patients following admission to ICU from American MIMIC-IV cohort and eICU cohort.
Study methods
Data sources
Software PostgreSQL (version 13.7.2) and Navicat Premium (version 16) were used to extract information using Structured Query Language (SQL) in operation. This retrospective study utilized two databases: The MIMIC-IV database includes electronic health records from Beth Israel Deaconess Medical Center in Boston, Massachusetts, USA, containing information such as patient measurements, physician’s orders, diagnoses, treatment procedures, and deidentified free-text clinical notes from 2007 to 2020. The database included information on 1,000,000 patients. The eICU 2.0 database is an open-access electronic medical record database designed for health care professionals and researchers to access clinical data in intensive care units (ICUS). Created by the Philips Group in collaboration with the Computational Physiology Laboratory (LCP) at the Massachusetts Institute of Technology (MIT), this database covers routine data from more than 200,000 patients from 2010 to 2022, covering physiological parameters, laboratory results, medication records, and diagnostic information. The MIMIC-IV Database was used as the training cohort and the eICU Database as the validation cohort.
Study population
The study population included ICU patients diagnosed with TBI. Inclusion criteria were: (1) age > 18 years, (2) ICU admission time > 96 h, (3) availability of HbA1c. Exclusion criteria included: (1) death within 6 h, (2) previous TBI with significant sequelae, such as speech impairment, motor dysfunction, (3) pre-existing cognitive disorders such as cerebrovascular dementia or Alzheimer’s disease. To evaluate the potential selection bias introduced by restricting the analysis to patients with available HbA1c values, we compared the baseline characteristics of included patients with those excluded due to missing HbA1c in the eligible cohort.
Research goals
The primary outcome was the 28-day all-cause mortality. Secondary outcomes included ICU mortality, incidence of cognitive dysfunction [Glasgow Coma Scale (GCS) < 8 [17], length of hospital stay, sequential organ failure assessment (SOFA) score and acute physiology score III (APS III).
Data extraction
Collected data included demographic information (age, sex, blood pressure, underlying conditions), GCS score, SOFA score, APS III, 28-day mortality, ICU mortality, and the length of hospital stay for TBI patients. The worsest laboratory results such as white blood count (WBC), hemoglobin (HB), neutrophil and lymphocyte counts, platelet count, prothrombin time (PT), partial thromboplastin time (PTT), international normalized ratio (INR), creatinine, BUN, glucose, C-reactive protein, triglycerides, and cholesterol were also collected within 24 h. Patients with more than 40% missing values across key laboratory and physiological variables were excluded. For variables with ≤ 40% missingness, multiple imputation by chained equations (MICE) was performed using the mice package in R version 4.2.2. Five imputations were generated. The imputation model included demographic characteristics, vital signs, laboratory tests, and major comorbidities. Outcome variables were not used to impute themselves. Continuous variables were imputed using predictive mean matching, and categorical variables using logistic or polytomous regression.
Research grouping
Based on previous reports, the blood glucose levels were aggregated from admission measurements, in order to avoid the bias, the SHR was calculated using the formula: Admission blood glucose levels (ABG)(mg/dl) / [(28.7 × HbA1c %) − 46.7] upon admission [18, 19]. Admission blood glucose was is defined as the first blood sugar measurement taken after admission.In both the training and validation cohorts, patients were divided into four SHR quartiles: T1 (0.43–1.02), T2 (1.02–1.28), T3 (1.28–1.59), and T4 (1.59–18.94).
Data analysis
Depending on the data distribution, continuous variables were reported as mean ± standard deviation (SD) or median with interquartile range (IQR). These were compared using ANOVA and Kruskal-Wallis tests as appropriate. Categorical variables are expressed as counts or percentages (%) and compared using Pearson χ2 or Fisher’s exact test.
Cox proportional hazard regression (adjusted age, gender, SOFA, APS III, heart failure, liver disease、GCS score、neurosurgical intervention、the time of ventilator、the time from injury to ICU admission) was used to calculate hazard ratios (HRs) with 95% confidence intervals (CIs). Kaplan-Meier curves were generated to estimate 28-day all-cause mortality and ICU mortality for TBI patients. In addition, a restricted cubic spine (RCS) with four nodes (at the 5th, 35th, 65th, and 95th percentiles) was applied to present the nonlinear relationship between SHR and outcomes, with nonlinearity tested using the Wald test. The subgroup analysis was conducted as a post hoc analysis, which was designed to explore potential heterogeneity in the relationship between SHR and 28-day mortality of ICU mortality across different population subgroups stratified by Gender, Age, SOFA, APS III, Liver disease, diabetes, Heart failure, and Renal disease to evaluate potential interactions in these associations. Avoiding Type I errors, the Bonferroni method was employed to adjust the significance threshold, thereby making the results of the subgroup analysis more closely aligned with the true effect.
Result
Baseline characteristic of the development cohort
As illustrated in Fig. 1, a total of 3,332 TBI patients were initially screened. After excluding 204 patients (8.1%) with more than 40% missing data, the remaining patients were evaluated according to the inclusion criteria, and 409 patients with available HbA1c were ultimately included in the final analysis. According to their SHR quartiles, TBI patients were categorized into four groups (T1, T2, T3, T4). Patients in higher SHR quartiles had higher APS III scores (P < 0.001), higher SOFA scores (P < 0.001), lower GCS scores (P < 0.001), and a higher likelihood of receiving hypertonic saline (P < 0.001), neurosurgical intervention (P < 0.01), vasopressin (P < 0.01), and invasive ventilation (P < 0.001). Comorbidities such as hypertension (P < 0.001), diabetes (P < 0.001), and liver disease (P = 0.02) were more common in higher SHR groups.Compared with the T1 group, the T3 and T4 groups had higher lactate levels (P = 0.001), higher heart rates (P = 0.012), lower mean arterial pressures (P = 0.006), higher glucose levels (P < 0.001), higher anion gaps (P = 0.004), elevated WBC counts (P < 0.001), higher BUN (P = 0.003), higher creatinine (P = 0.002), and prolonged INR (P < 0.001), PT (P < 0.001), and PTT (P = 0.01), as shown in Table 1. A comparison between included patients and those excluded due to missing HbA1c is presented in Supplementary Table S1.
Fig. 1.
Work flow of TBI patients
Table 1.
Baseline characteristic among four groups according to stress hyperglycemia ratio in the TBI patients
| Variables | Total(n = 409) | SHR_ | Statistic | P | SMD | |||
|---|---|---|---|---|---|---|---|---|
| T1(n = 101) | T2(n = 103) | T3(n = 102) | T4(n = 103) | |||||
| Gender, n (%) | 259 (63.33) | 60 (59.41) | 65 (63.11) | 66 (64.71) | 68 (66.02) | χ2 = 1.0761 | 0.783 | 0.074 |
| Age, M (Q1, Q3) | 67.00 (52.00, 79.00) | 64.00 (45.00, 82.00) | 70.00 (58.50, 79.50) | 70.00 (57.25, 79.00) | 63.00 (52.00, 75.00) | χ2 = 7.6952 | 0.053 | 0.200 |
| APSIII, M (Q1, Q3) | 39.00 (32.00, 53.00) | 34.00 (27.00, 41.00) | 38.00 (30.00, 47.00) | 40.00 (34.00, 56.00) | 46.00 (39.00, 61.00) | χ2 = 45.6002 | < 0.001 | 0.433 |
| SOFA, M (Q1, Q3) | 4.00 (2.00, 6.00) | 3.00 (2.00, 5.00) | 3.00 (2.00, 5.00) | 4.00 (2.00, 6.00) | 5.00 (3.50, 7.00) | χ2 = 32.9462 | < 0.001 | 0.397 |
| GCS, Mean ± SD | 7.69 ± 0.64 | 9.45 ± 0.56 | 8.27 ± 0.76 | 7.58 ± 0.68 | 6.74 ± 0.54 | χ2 = 34.7632 | < 0.001 | 0.423 |
| Antibiotic, n (%) | 333 (81.41) | 76 (75.24) | 85 (82.52) | 85 (83.33) | 87 (84.46) | χ2 = 3.5051 | 0.320 | 0.119 |
| Vasopressin, n (%) | 63 (15.40) | 3 (2.97) | 7 (6.79) | 20 (19.60) | 33 (32.03) | χ2 = 41.0961 | < 0.001 | 0.483 |
| Invasive_vent, n (%) | 171 (41.80) | 31 (30.69) | 37 (35.92) | 45 (44.11) | 58 (56.31) | χ2 = 15.7231 | 0.001 | 0.293 |
| Neurosurgery, n (%) | 40 (9.78) | 2 (1.98) | 9 (8.73) | 16 (15.686) | 13 (12.621) | χ2 = 12.0661 | 0.007 | 0.274 |
| Congestive heart failure, n (% | 76 (18.58) | 13 (12.87) | 22 (21.36) | 20 (19.61) | 21 (20.39) | χ2 = 2.9951 | 0.392 | 0.117 |
| Peripheral vascular, n (%) | 31 (7.58) | 4 (3.96) | 8 (7.77) | 12 (11.76) | 7 (6.80) | χ2 = 4.5341 | 0.209 | 0.154 |
| Hypertension, n (%) | 283 (69.19) | 84 (83.17) | 77 (74.76) | 61 (59.80) | 61 (59.22) | χ2 = 19.7711 | < 0.001 | 0.327 |
| Chronic pulmonary, n (%) | 67 (16.38) | 16 (15.84) | 16 (15.53) | 16 (15.69) | 19 (18.45) | χ2 = 0.4321 | 0.934 | 0.040 |
| Renal disease, n (%) | 69 (16.87) | 14 (13.86) | 16 (15.53) | 21 (20.59) | 18 (17.48) | χ2 = 1.8151 | 0.612 | 0.098 |
| Diabetes, n (%) | 133 (32.52) | 16 (15.84) | 34 (33.01) | 38 (37.25) | 45 (43.69) | χ2 = 19.7121 | < 0.001 | 0.331 |
| Liver disease, n (%) | 28 (6.85) | 9 (8.91) | 4 (3.88) | 1 (0.98) | 14 (13.59) | χ2 = 14.9461 | 0.002 | 0.294 |
| Hypertonic saline, n(%) | 319(77.98%) | 63(62.37%) | 78(75.73%) | 83(81.37%) | 95(92.23%) | χ2 = 20.9461 | < 0.001 | 0.373 |
| Lactate, M (Q1, Q3) | 2.30 (1.83, 2.81) | 2.16 (1.82, 2.48) | 2.24 (1.82, 2.63) | 2.31 (1.82, 2.81) | 2.57 (1.90, 3.95) | χ2 = 16.9962 | 0.001 | 0.358 |
| Spo2, Mean ± SD | 92.66 ± 4.46 | 93.20 ± 3.21 | 92.87 ± 3.22 | 92.74 ± 4.93 | 91.84 ± 5.85 | F = 1.7323 | 0.160 | 0.153 |
| HR, Mean ± SD | 105.38 ± 20.46 | 102.20 ± 20.47 | 103.47 ± 17.16 | 104.89 ± 20.71 | 110.91 ± 22.33 | F = 3.7193 | 0.012 | 0.222 |
| MBP, Mean ± SD | 82.15 ± 9.98 | 82.31 ± 10.84 | 84.51 ± 10.39 | 82.16 ± 9.06 | 79.64 ± 9.03 | F = 4.1973 | 0.006 | 0.252 |
| RR, Mean ± SD | 27.43 ± 6.02 | 26.60 ± 5.25 | 26.67 ± 5.21 | 28.12 ± 6.54 | 28.34 ± 6.77 | F = 2.4553 | 0.063 | 0.186 |
| T, Mean ± SD | 37.62 ± 0.74 | 37.55 ± 0.70 | 37.54 ± 0.68 | 37.64 ± 0.76 | 37.76 ± 0.81 | F = 2.0013 | 0.113 | 0.169 |
| Urine output, M (Q1, Q3) | 1605.00 (1035.00, 2390.00) | 1390.00 (950.00, 2300.00) | 1650.00 (1039.50, 2320.00) | 1650.00 (1192.00, 2345.00) | 1585.00 (956.00, 2602.50) | χ2 = 3.4972 | 0.321 | 0.132 |
| Glucose, M (Q1, Q3) | 162.58 (129.00, 221.00) | 114.00 (101.00, 127.00) | 143.00 (129.00, 161.00) | 180.50 (162.00, 217.50) | 249.00 (201.00, 312.50) | χ2 = 252.1232 | < 0.001 | 1.000 |
| AG, M (Q1, Q3) | 16.00 (14.00, 19.00) | 15.00 (14.00, 18.00) | 16.00 (14.50, 19.00) | 16.00 (14.00, 19.00) | 16.00 (15.00, 20.00) | χ2 = 13.1792 | 0.004 | 0.280 |
| WBC, M (Q1, Q3) | 12.60 (9.40, 16.60) | 11.10 (8.40, 13.80) | 12.10 (9.65, 15.55) | 12.65 (9.03, 16.78) | 15.00 (11.85, 20.40) | χ2 = 38.9242 | < 0.001 | 0.455 |
| BUN, M (Q1, Q3) | 18.00 (14.00, 26.00) | 16.00 (12.00, 23.00) | 17.00 (13.00, 23.00) | 20.00 (15.00, 27.00) | 21.00 (15.00, 30.00) | χ2 = 13.9522 | 0.003 | 0.167 |
| Creatinine, M (Q1, Q3) | 1.00 (0.80, 1.30) | 0.90 (0.70, 1.20) | 1.00 (0.70, 1.20) | 1.05 (0.90, 1.40) | 1.10 (0.80, 1.55) | χ2 = 15.3572 | 0.002 | 0.143 |
| INR, M (Q1, Q3) | 1.20 (1.10, 1.40) | 1.20 (1.10, 1.30) | 1.10 (1.10, 1.30) | 1.20 (1.10, 1.30) | 1.30 (1.10, 1.70) | χ2 = 27.9522 | < 0.001 | 0.293 |
| PT, M (Q1, Q3) | 13.10 (12.00, 15.10) | 13.00 (12.00, 14.40) | 12.60 (11.50, 13.90) | 13.15 (12.12, 14.59) | 14.40 (12.60, 18.55) | χ2 = 25.6722 | < 0.001 | 0.290 |
| PTT, M (Q1, Q3) | 29.40 (26.40, 32.80) | 29.50 (27.40, 32.40) | 28.10 (26.45, 31.85) | 28.40 (25.80, 32.27) | 31.10 (26.85, 39.55) | χ2 = 11.2842 | 0.010 | 0.217 |
¹ Pearson χ² test, ² Welch’s t-test, ³ Mann–Whitney U test, ⁴ Student’s t-test
Association between SHR and 28-day all-cause mortality of TBI patients
In this study, TBI patients were divided into four groups based on their SHR. In the training cohort, 28-day all-cause mortality increased across SHR quartiles (T1: 15.84%, T2: 11.65%, T3: 21.57%, T4: 33.98%), as shown in Fig. 2A.
Fig. 2.
A The incidence of 28-day and ICU all-cause mortality among four groups according to stress hyperglycemia ratio in the TBI patients. B The incidence of ICU all-cause mortality among four groups according to stress hyperglycemia ratio in the TBI patients
In the univariate Cox regression analysis (Model 1), the T4 group had a significantly higher hazard of 28-day mortality compared with the T1 group (HR = 2.41, 95% CI 1.33–4.36, P = 0.004).This association remained significant after adjusting for age and sex (Model 2: HR = 2.26, 95% CI 1.25–4.09, P = 0.007) and persisted in the fully adjusted model (Model 3: HR = 1.83, 95% CI 1.18–2.87, P = 0.018), as shown in Fig. 3. The Kaplan–Meier survival curves demonstrated significantly higher 28-day mortality in the T3 and T4 groups compared with the T1 group (Fig. 4A).
Fig. 3.
Cox hazard regression for the association of stress hyperglycemia ratio with 28-day and ICU all-cause mortality. Model 1: unadjusted. Model 2 (core): adjusted for age, gender, APS III, congestive heart failure, chronic liver disease. Model 3 (fully adjusted): model 2 plus Glasgow Coma Scale on admission, neurosurgical intervention, invasive mechanical ventilation during the first ICU day, early antibiotic therapy and vasopressor use. SHR quartile 1 serves as reference. HR = hazard ratio; CI = confidence interval
Fig. 4.
A The Kaplan-Meier survival curves of 28-day all-cause mortality among four groups according to stress hyperglycemia ratio in the TBI patients. B The Kaplan-Meier survival curves of ICU all-cause mortality among four groups according to stress hyperglycemia ratio in the TBI patients
Restricted cubic spline (RCS) analysis assessed the relationship between SHR and 28-day all-cause mortality. The RCS results demonstrated a strong correlation between the SHR and 28-day all-cause mortality, close to 0.001, with a statistically significant nonlinear relationship between the hazard ratio (HR) and mortality (p = 0.002), as shown in Fig. 5. A change in the slope of the RCS curve was observed around SHR ≈ 1.3, representing a mathematical inflection point rather than a biological threshold. To further explore this pattern, we conducted a sensitivity analysis comparing SHR ≥ 1.3 vs. < 1.3, which showed results consistent with the overall trend. The results are consistent with the main analysis trend, as shown in supplementary Table 1.
Fig. 5.
Restricted three-spline analysis among four groups according to stress hyperglycemia ratio in the TBI patients
Association between SHR and ICU mortality of TBI patients
In this study, TBI patients were divided into four groups based on their SHR. In the training cohort, ICU mortality differed across SHR quartiles (T1: 8.91%, T2: 1.95%, T3: 10.78%, T4: 23.30%), as shown in Fig. 2B.
However, Cox regression analyses did not identify an independent association between SHR and ICU mortality.In Model 1, none of the SHR quartiles showed significantly different hazards of ICU mortality compared with T1 (T4 vs. T1: HR = 1.39, 95% CI 0.64–3.03, P = 0.411).Similarly, no significant association was observed after adjustment for age and sex (Model 2: HR = 1.63, 95% CI 0.71–3.74, P = 0.249) or in the fully adjusted model (Model 3: HR = 1.04, 95% CI 0.44–2.44, P = 0.993), as shown in Fig. 3. In contrast, Kaplan–Meier curves suggested higher ICU mortality in the T3 and T4 groups (Fig. 4B). This discrepancy between crude (KM) and adjusted (Cox) results, together with the small number of ICU deaths, indicates that SHR was not independently associated with ICU mortality.
Subgroup analysis
Subgroup analyses were conducted to evaluate whether the association between SHR and 28-day mortality differed across demographic and clinical strata, including sex, age, APS III category, and comorbidities (congestive heart failure, peripheral vascular disease, hypertension, chronic pulmonary disease, renal disease, diabetes, and liver disease). The overall hazard ratio per unit increase in SHR was modest but statistically significant (HR 1.09, 95% CI 1.00–1.19, P = 0.04), as shown in Fig. 6.The association between higher SHR and increased mortality appeared stronger in male patients (HR 2.01, 95% CI 1.23–3.29, P = 0.005) than in female patients (HR 1.08, 95% CI 0.98–1.19, P = 0.123), with a significant interaction (P for interaction = 0.016).
Fig. 6.
Subgroup analyses
Similarly, the mortality risk associated with SHR was more pronounced among patients with lower illness severity (APS III < 38) (HR 4.02, 95% CI 1.39–11.64, P = 0.01) compared with those with higher APS III scores (HR 1.04, 95% CI 0.94–1.15, P = 0.502), with a significant interaction (P for interaction = 0.012).For other subgroups—including age strata, congestive heart failure, peripheral vascular disease, hypertension, chronic pulmonary disease, renal disease, diabetes, and liver disease—the associations between SHR and mortality were directionally similar but did not reach statistical significance (all P for interaction > 0.05). The detailed effect sizes are presented in Supplementary Table 2.
Validation analysis
There were significant differences in APS III (P < 0.001), SOFA scores (P = 0.023), congestive heart failure (P = 0.041), peripheral vascular disease (P = 0.001), renal disease (P = 0.0102), heart rate (HR) (P = 0.001), AG (P < 0.001), WBC (P < 0.001), INR (P < 0.001), PT (P < 0.001), and PTT (P < 0.001) between the different SHR groups of TBI patients (see Supplementary Table 3). SHR in these patients was associated with varying risks of 28-day all-cause mortality in the external validation cohort (P < 0.001). Although crude comparisons showed numerical differences in ICU all-cause mortality (P < 0.001), this association did not remain statistically significant after multivariable adjustment, indicating that SHR was not independently associated with ICU mortality. Cox proportional hazard regression was applied to analyze the risk of 28-day and ICU all-cause mortality.In the univariate retrospective analysis of 28-day all-cause mortality, T4 had a significantly higher mortality than T1, as shown in Supplementary Table 3(HR = 4.82, 95% CI 2.21–10.49, P < 0.001). The multivariate regression analysis confirmed that the 28-day all-cause mortality remained elevated across SHR groups, with T4 showing the highest risk, as shown in Supplementary Table 3.There was no statistically significant difference in in- ICU mortality, as shown in Supplementary Table 4(HR = 0.99, 95% CI 0.44–3.85, P = 0.988).The K-M curve was used to estimate in-hospital and ICU mortality in patients with TBI. Statistically significant differences in mortality were found between the different SHR groups. 28-day and ICU mortality was higher in the T3 and T4 groups compared to the T1 group, as depicted in supplement Fig. 1. Given the small number of ICU deaths and the inconsistency between Kaplan–Meier curves and the adjusted Cox models, the relationship between SHR and ICU mortality should be considered inconclusive and exploratory.
Discussion
With advancements in economic and social development, the incidence of TBI has increased annually, exceeding 300,000 new cases each year, resulting in significant economic and social burdens. TBI often induces stress response, and stress-induced hyperglycemia is related to the intensity of the stress state and the baseline blood glucose levels. The SHR measures changes in stress-induced blood glucose levels. This study aimed to explore the relationship between SHR, 28-day all-cause mortality, and ICU all-cause mortality in patients with TBI, using MIMIC-IV and eICU as the training and validation cohorts, respectively. We found that SHR was closely associated with 28-day all-cause mortality, whereas its association with ICU all-cause mortality was inconsistent and did not remain significant after adjustment. In the validation cohort, SHR showed crude differences in ICU mortality across quartiles, but this association did not persist in multivariable models. This discrepancy between crude and adjusted results, combined with the small number of ICU deaths and the inconsistency between Kaplan–Meier curves and Cox regression, indicates that the relationship between SHR and ICU mortality is not robust. Therefore, the ICU mortality findings should be regarded as inconclusive and exploratory, and SHR cannot be considered a reliable predictor of ICU mortality based on the present data.
Stress-induced hyperglycemia is a common and persistent condition in TBI patients, often leading to high variability in blood glucose levels. The SHR measurement had some limitations, such as dependency of HbA1c and single-point glucose. The research had excluded the TBI patients lacking the HbA1c, and averaged the glucose counts within 24 h after admission. This variability can increase the risk of cardiovascular events and mortality in cardiovascular disease patients [20–24]. Stress-induced hyperglycemia is influenced by both the intensity of the stress response and baseline blood glucose levels. Rapid fluctuations in stress-induced blood glucose levels can damage endothelial cells, peripheral nerve dysfunction, and immune system damage, leading to microvascular thrombosis, inflammatory cytokine storms, and organ dysfunction [25–27]. Additionally, patients with TBI often undergo surgical treatment and ICU admission, which can cause hormonal imbalances and may necessitate intensive insulin-based glucose-lowering therapy. Nutritional support in the ICU can also elevate SHR.
Previous studies have reported that a high SHR is associated with baseline blood glucose levels and stress states. To further assess the SHR’s role in risk stratification values in outcome measurement, subgroup analysis was conducted in different subgroups within the training cohort, including sex, age, APS III, congestive heart failure, peripheral vascular disease, hypertension, chronic lung disease, renal disease, diabetes, and liver disease. In the training cohort, an elevated SHR was significantly correlated with increased mortality risk across various subgroups, including sex, age, APS III score, congestive heart failure, peripheral vascular disease, hypertension, chronic lung disease, renal disease, diabetes, and liver disease. However, no significant correlations were found in the assessment of mutual relationships. The risk was higher in males (95% CI: 2.01 [1.23–3.09], P = 0.016) and in patients with an APS III score (95% CI: 4.02 [1.39–11.64], P = 0.012). Although SHR showed a strong association with 28-day all-cause mortality, prospective studies are still needed for further validation.
Similarly, there were significant differences in baseline characteristics such as APS III, SOFA score, WBC, INR, PT, PTT between the different SHR groups of TBI patients in the validation cohort, though SHR was associated with the severity of TBI patients. In the assessment of mortality, SHR in TBI patients was associated with varying risks of 28-day all-cause mortality in the validation cohort. However, SHR did not show an independent association with ICU mortality in TBI patients.
This study has several limitations. First, the analysis was restricted to patients with available HbA1c measurements because HbA1c is required to calculate the SHR. This led to the exclusion of a large proportion of TBI patients who did not undergo HbA1c testing. Patients who have HbA1c recorded are typically those with known or suspected diabetes, more chronic comorbidities, or closer metabolic surveillance, which means that the included cohort may represent a systematically different subset of the overall TBI population. This selection mechanism introduces the potential for selection bias and may limit the generalizability of our findings, as the association between SHR and outcomes may differ in patients without chronic metabolic disease or those who do not routinely receive HbA1c testing. In addition, SHR was calculated using the first recorded glucose measurement after ICU admission. A single glucose value may not accurately reflect the dynamic glycemic fluctuations or the sustained hyperglycemic stress that occurs throughout the ICU stay. Glycemic variability, repeated episodes of hyperglycemia, and treatment-related changes in glucose levels may all contribute to prognosis but are not captured by a single time-point measurement. This limitation may lead to misclassification of metabolic stress and could attenuate the observed associations between SHR and clinical outcomes.
Second, the study population was derived primarily from Western white individuals in publicly available critical care databases, which limits the assessment of racial or ethnic variability in the relationship between SHR and clinical outcomes.Third, although the study was externally validated using the eICU database, it was not cross-validated with additional independent datasets, and the applicability of our findings to other healthcare settings remains uncertain.Finally, as a retrospective cohort study, causal inference cannot be established, and prospective, multicenter studies are needed to further validate the prognostic value of SHR in TBI. Moreover, as with all observational studies, the possibility of residual confounding remains. Certain clinically relevant factors—such as the timing of interventions, specific ICU management protocols, nutritional status, and other unmeasured or unavailable variables in the databases—could not be fully accounted for, and may have influenced both SHR values and patient outcomes.
Conclusion
In patients with TBI, SHR was associated with indicators of disease severity and demonstrated a strong and consistent association with 28-day all-cause mortality. However, its association with ICU all-cause mortality was not statistically significant after adjustment and should be considered inconclusive. Further prospective studies are required to clarify the role of SHR in ICU-specific outcomes.
Supplementary Information
Acknowledgements
Not applicable.
Authors’ contributions
All work was approved by the co-authors. Da-cheng Wang, Yan Zhuang and Lin Li made significant contributions to conception and study design. Cuijuan Zheng, Xinyuan Zhang and Da-cheng Wang completed data acquisition. Xue He and Hai-cheng Yang performed data analysis and interpretation; Yan Zhuang and Xue He have written the draft of the article and critically revised it. No conflicts of interest exist in the submission of this manuscript. We would like to declare on behalf of all co-authors that the work described was original research that has not been published previously and is not under consideration for publication elsewhere, in whole or in part. All authors read and approved of the final manuscript.
Funding
This work was supported by the National Natural Science Fund of China (Grant NO.82074379, 82274433).
Data availability
The datasets generated and/or analysed during the current study are not publicly available but are available from the corresponding author on reasonable request.
Declarations
Ethics approval and consent to participate
All experiments in this study were carried out in accordance with the Declaration of Helsinki. The data used in this study is a public database and will not cause adverse effects on patients. Therefore, this ethics and the consent to participate of the study was waived by the ethics committee of the Affiliated Huaian First people’s Hospital of Nanjing Medical University.
Consent for publication
Not applicable.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Xue He, Cuijuan Zheng and Lin Li contributed equally to this work and should be considered co-first authors.
References
- 1.Tatyana M, Shirin M, Angela C. Traumatic brain injury: sex, gender and intersecting vulnerabilities. Nat Rev Neurol. 2018;14(12). 10.1038/s41582-018-0091-y. [DOI] [PubMed]
- 2.Krassimir D, Jonathan G, Pinxia C, et al. Traumatic brain injury: intraoperative management and intensive care unit multimodality monitoring. Anesthesiol Clin. 2023;41(1). 10.1016/j.anclin.2022.11.003. [DOI] [PubMed]
- 3.Global. regional, and National burden of neurological disorders, 1990–2016: a systematic analysis for the global burden of disease study 2016. Lancet Neurol. 2019;18(5). 10.1016/s1474-4422(18)30499-x. [DOI] [PMC free article] [PubMed]
- 4.Ruoran W, Jing Z, Jing X, et al. Incidence and burden of acute kidney injury among traumatic Brain-Injury patients. Risk Manag Healthc Policy. 2021;14(0). 10.2147/rmhp.S335150. [DOI] [PMC free article] [PubMed]
- 5.Hironori T, Noriaki I, Jin K, et al. Glycemic variability determined with a continuous glucose monitoring system can predict prognosis after acute coronary syndrome. Cardiovasc Diabetol. 2018;17(1). 10.1186/s12933-018-0761-5. [DOI] [PMC free article] [PubMed]
- 6.Julien B, Hassane A, Vincent B, et al. Individualised versus conventional glucose control in critically-ill patients: the controling study-a randomized clinical trial. Intensive Care Med. 47(11). [DOI] [PMC free article] [PubMed]
- 7.Yohei H, Noriaki I, Jin K, et al. Prognostic importance of glycemic variability on left ventricular reverse remodeling after the first episode of ST-segment elevation myocardial infarction. Cardiovasc Diabetol. 2023;22(1). 10.1186/s12933-023-01931-3. [DOI] [PMC free article] [PubMed]
- 8.Paul M, Musa K, Moses G, et al. Hyperglycemia in severe traumatic brain injury patients and its association with thirty-day mortality: a prospective observational cohort study in Uganda. PeerJ. 2021;9(0). 10.7717/peerj.10589. [DOI] [PMC free article] [PubMed]
- 9.Michał T, Maryla K-T, Christopher P, et al. The role of stress hyperglycemia and hyperlactatemia in non-diabetic patients with myocardial infarction treated with percutaneous coronary intervention. Cardiol J. 2023;31(4). 10.5603/CJ.a2023.0041. [DOI] [PMC free article] [PubMed]
- 10.Yan L, Xiaowen L, Yinhua Z, et al. Impact of glycemic control status on patients with ST-segment elevation myocardial infarction undergoing percutaneous coronary intervention. BMC Cardiovasc Disord. 2020;20(1). 10.1186/s12872-020-01339-x. [DOI] [PMC free article] [PubMed]
- 11.Jie Y, Yitian Z, Chen L, et al. The impact of the stress hyperglycemia ratio on Short-term and Long-term poor prognosis in patients with acute coronary syndrome: insight from a large cohort study in Asia. Diabetes Care. 2022;45(4). 10.2337/dc21-1526. [DOI] [PubMed]
- 12.Fengjuan Y, Xiehui C, Xiaoqing Q, et al. Association between the stress hyperglycemia ratio and 28-day all-cause mortality in critically ill patients with sepsis: a retrospective cohort study and predictive model establishment based on machine learning. Cardiovasc Diabetol. 2024;23(1). 10.1186/s12933-024-02265-4. [DOI] [PMC free article] [PubMed]
- 13.Qing Z, Jie Y, Wenyao W, et al. The impact of the stress hyperglycemia ratio on mortality and rehospitalization rate in patients with acute decompensated heart failure and diabetes. Cardiovasc Diabetol. 2023;22(1). 10.1186/s12933-023-01908-2. [DOI] [PMC free article] [PubMed]
- 14.Justin BE-T, Songzhu Z, Guy B, et al. Visit-to-Visit glycemic variability and risks of cardiovascular events and All-Cause mortality: the ALLHAT study. Diabetes Care. 2019;42(3). 10.2337/dc18-1430. [DOI] [PMC free article] [PubMed]
- 15.John Peter M, Elif Ilhan E, Gregory Peter M, et al. Relationship between glycated hemoglobin and stroke risk: a systematic review and meta-analysis. J Am Heart Assoc. 2018;7(11). 10.1161/jaha.117.007858. [DOI] [PMC free article] [PubMed]
- 16.Elahe P, Maryam J, Sima H, et al. Improved glycemic status, insulin resistance and inflammation after receiving oral oleoylethanolamide supplement in people with prediabetes: a randomized controlled trial. Diabetol Metab Syndr. 2022;14(1). 10.1186/s13098-022-00848-3. [DOI] [PMC free article] [PubMed]
- 17.Yelena GB, Alice B, Nancy RT, et al. Diagnosing level of consciousness: the limits of the Glasgow coma scale total score. 2021, 38(23). 10.1089/neu.2021.0199 [DOI] [PMC free article] [PubMed]
- 18.Yuhao S, Weiguo F, Yang L, et al. Glycemic variability and in-hospital death of critically ill patients and the role of ventricular arrhythmias. Cardiovasc Diabetol. 2023;22(1). 10.1186/s12933-023-01861-0. [DOI] [PMC free article] [PubMed]
- 19.Olivier L, Danielle P, Jean-Charles P. Update on glucose in critical care. Nutrition. 2018;59(0). 10.1016/j.nut.2018.06.027.
- 20.Jing H, Shichen L, Fang L, et al. Glycemic control is related to cognitive dysfunction in Chinese children with type 1 diabetes mellitus. J Diabetes. 2018;10(12). 10.1111/1753-0407.12775. [DOI] [PubMed]
- 21.Yifei M, Chunfang W, Jingyi L, et al. Impact of short-term glycemic variability on risk of all-cause mortality in type 2 diabetes patients with well-controlled glucose profile by continuous glucose monitoring: a prospective cohort study. Diabetes Res Clin Pract. 2022;189(0). 10.1016/j.diabres.2022.109940. [DOI] [PubMed]
- 22.Jinghao C, Qing Y, Jingyi L, et al. Impact of the complexity of glucose time series on All-Cause mortality in patients with type 2 diabetes. J Clin Endocrinol Metab. 2022;108(5). 10.1210/clinem/dgac692. [DOI] [PMC free article] [PubMed]
- 23.Peter JG, Francesco C. The 2019 ESC guidelines on diabetes, pre-diabetes, and cardiovascular diseases developed in collaboration with the EASD: new features and the ‘Ten commandments’ of the 2019 guidelines are discussed by professor Peter J. Grant and professor Francesco Cosentino, the task force chairmen. Eur Heart J. 2019;40(39). 10.1093/eurheartj/ehz687. [DOI] [PubMed]
- 24.Jia Z, Qian Z, Hongqiu G, et al. Comparison of stress hyperglycemia ratio and glycemic gap on acute ICH in-hospital outcomes. Ann Clin Transl Neurol. 2024;11(6). 10.1002/acn3.52063. [DOI] [PMC free article] [PubMed]
- 25.Mohammed A, Luo Y, Wang K, et al. Stress hyperglycemia ratio as a prognostic indicator for long-term adverse outcomes in heart failure with preserved ejection fraction. Cardiovasc Diabetol. 2024;23(1):67. 10.1186/s12933-024-02157-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Vincent J, Sakr Y, Singer M, et al. Prevalence and outcomes of infection among patients in intensive care units in 2017. JAMA. 2020;323(15):1478–87. 10.1001/jama.2020.2717. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Sharif A, Fayed M. Evaluation of multiple organ dysfunction score (MODS) and the sequential organ failure assessment (SOFA) score as in-hospital outcome predictors among cases of hydrogen cyanamide exposure: a cross-sectional study. Environ Sci Pollut Res Int. 2021;28(31):42161–76. 10.1007/s11356-021-13655-6. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The datasets generated and/or analysed during the current study are not publicly available but are available from the corresponding author on reasonable request.