SHRs, biomarkers for dysregulated stress response, predict prognosis in sepsis patients: a retrospective cohort study from MIMIC-IV database

Abstract

Background

The dysregulated stress response is a key pathological mechanism underlying sepsis and is strongly associated with poor clinical outcomes. Stress hyperglycemia, a common manifestation of this response, may provide valuable prognostic information in sepsis patients. The stress hyperglycemia ratio (SHR) offers a more accurate reflection of the stress response and may be instrumental in assessing sepsis prognosis.

Methods

This study aimed to investigate the relationship between SHRs and clinical outcomes in sepsis patients. Data were obtained from the Medical Information Mart for Intensive Care IV database. Demographic information, intensive care unit (ICU) parameters within the first 24 h, laboratory results, insulin administration, survival time, and outcomes were extracted for analysis. Four SHR metrics (SHRfirst, SHRmin, SHRmax, and SHRmean) were calculated based on blood glucose values during the first 24 h of ICU admission (first, minimum, maximum, and mean, respectively). The predictive performance of each SHR metric was compared using the area under the receiver operating characteristic (ROC) curve. Kaplan–Meier survival analysis was performed to assess survival rates across groups defined by ROC curve-generated cut-off values. Associations between SHR and 28-day as well as 1-year mortality were further examined using both univariate and multivariate Cox regression analyses.

Results

A total of 5,025 sepsis patients were included, of whom 656 died within 28 days of ICU admission. SHR was significantly higher in the non-survivor group. Among the SHR metrics, SHRmax demonstrated the highest predictive value for both 28-day and 1-year mortality. Higher SHR values were consistently associated with increased mortality (all P < 0.001). For SHRmax, each 1-unit increase was associated with a 77% increase in mortality in univariate analysis and a 71.6% increase in multivariate analysis. Sensitivity analyses indicated that the relationship between SHR and mortality was stronger in patients without diabetes.

Conclusions

SHR serves as a robust marker of the dysregulated stress response in sepsis and holds significant prognostic value, particularly SHRmax, in predicting mortality. These findings underscore the potential clinical utility of SHR in guiding therapeutic strategies aimed at modulating the stress response and blood glucose levels in critically ill sepsis patients. Further research is warranted to explore SHR-targeted interventions in sepsis management.

Introduction

Sepsis is a life-threatening condition characterized by organ dysfunction resulting from a dysregulated host response to infection, and it remains the leading cause of mortality among critically ill patients [1]. This pathological state involves excessive and dysregulated stress responses encompassing endocrine, inflammatory, and immune pathways, which lead to systemic dysfunction, microcirculatory disturbances, and ultimately organ failure, thereby contributing to the poor prognosis frequently observed in sepsis [2]. In recent years, there has been growing research interest in factors related to the stress response that may improve outcomes in sepsis, with particular focus on biomarkers and predictive tools associated with patient mortality [3, 4].

Stress hyperglycemia, a significant manifestation of the stress response caused by infection and trauma, is commonly observed in critically ill patients [5]. This is supported by the study by Roberts et al. [6], which demonstrated a significant increase in blood glucose is provoked by the stress response to some illness or procedure, especially sepsis, ST-elevation myocardial infarction, cardiac surgery and others. Stress hyperglycemia is characterized by a transient elevation in blood glucose levels caused by impaired glucose metabolism regulation and the development of insulin resistance [5, 7]. It serves as an important marker of disease severity and has been closely associated with clinical outcomes in critically ill patients [8, 9]. Traditional assessments of stress hyperglycemia, which rely solely on blood glucose concentrations, are limited by their inability to account for chronic glycemic control [10]. To address this limitation, the stress hyperglycemia ratio (SHR), as proposed by Robert et al. [8], integrates both acute and chronic glycemic states by incorporating blood glucose and glycated hemoglobin (HbA1c), providing a more comprehensive assessment of stress response-induced alterations in glucose regulation. Specifically, the SHR calculated using the first recorded blood glucose level within 24 h of admission, referred to as SHRfirst, has been linked to sepsis outcomes. Recent studies by Yan et al. [11] have identified SHRfirst as a potential indicator of acute hyperglycemia and poor prognosis in critically ill patients.

It is well established that the prognosis of sepsis is intimately linked to the host response [12]. As an acute phase host response, stress response has attracted much attention. However, due to inherent variability in initial blood glucose measurements, SHRfirst may not fully capture the extent of the stress response in sepsis patients [13]. Conversely, alternative SHR metrics, including SHRmax, SHRmin, and SHRmean, calculated from the respective maximum, minimum, and mean blood glucose values within the first 24 h of admission, may provide a more accurate representation of the stress response. A recent study by He et al. [13] demonstrated that SHRmax better reflects the degree of acute phase response in patients with acute coronary syndrome (ACS) compared with SHRfirst. Despite this, there is a paucity of research exploring the relationship between different SHR metrics and sepsis outcomes. Most existing studies have focused on the association between SHRfirst and short-term sepsis prognosis [14,15,16]; however, investigations into the relationship between SHR and long-term outcomes in sepsis remain limited. Although Li et al. [17] recently identified a significant link between SHRfirst and 1-year mortality in sepsis patients, other studies involving critically ill populations have suggested that SHR may not be a reliable predictor of long-term outcomes [18, 19]. Consequently, the utility of SHR in predicting long-term prognosis in sepsis remains uncertain.

Given that SHR reflects the magnitude of the stress response, this study hypothesizes that SHR serves as a valuable clinical biomarker for evaluating the prognosis of sepsis patients. The study aims to examine the association between various SHR metrics and both short-term and long-term outcomes in sepsis patients, compare the prognostic efficacy of different SHR measurements, and perform sensitivity analyses to assess the impact of diabetes mellitus (DM) on predictive accuracy.

Methods and materials

Study design and data source

This retrospective cohort study utilized data extracted from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database (version 2.2), which contains deidentified health information from over 50,000 patients admitted to the intensive care unit (ICU) at Beth Israel Deaconess Medical Center (Boston, MA, USA) between 2008 and 2019. The database access was granted to Dr. Guangjian Wang (certification number: 61644758). The study adhered to the ethical principles outlined in the Declaration of Helsinki and received approval from the institutional review boards of both the Massachusetts Institute of Technology (MIT, Cambridge, MA, USA) and Beth Israel Deaconess Medical Center (Boston, MA, USA). Given that all patient data were anonymized and deidentified, a formal ethical review and informed consent were not required for this publication.

Study population and data extraction

This study included all adult patients admitted to the ICU with sepsis, defined in accordance with the Sepsis-3 criteria (infection accompanied by organ dysfunction) [1]. Exclusion criteria were as follows: (1) patients younger than 18 years of age; (2) patients without HbA1c measurements within 24 h of ICU admission; (3) patients whose glucose levels were not obtained within 24 h of ICU admission; and (4) patients with multiple ICU admissions, from whom only data related to the first ICU admission were included. Data extraction was performed using Navicat Premium software (version 16) with structured query language. The variables extracted included demographic characteristics (age, gender, height, weight, and body mass index), comorbidities (acute myocardial infarction, heart failure, cerebrovascular disease, autoimmune disease, liver disease, DM, chronic kidney disease, mild to severe diffused kidney disease, and Charlson Comorbidity Index (CCI)), and ICU parameters within the first 24 h (temperature, heart rate, respiratory rate (RR), mean blood pressure (MBP), peripheral oxygen saturation (SpO₂), oxygenation index, and arterial blood gas values). Laboratory tests (blood cell count, renal and liver function, coagulation function, HbA1c, and blood glucose), insulin usage, survival time, and patient outcomes were also extracted. Comorbidities were identified according to the International Classification of Diseases coding system (ICD codes and item identification codes for data query and extraction were provided by the Massachusetts Institute of Technology Laboratory for Computational Physiology, downloaded from GitHub, https://github.com/ MIT-LCP/mimic-code). The maximum values for temperature, heart rate, RR, lactate, and minimum oxygenation index were collected. The average glucose (AG, mg/dl) was calculated using the formula: 28.7 × HbA1c (%) − 46.7. The SHR was calculated as blood glucose concentration (mg/dl) divided by AG [8]. SHRfirst, SHRmin, SHRmax, and SHRmean were derived from the first, minimum, maximum, and mean values of blood glucose within the first 24 h of ICU admission, respectively.

Outcomes

All patients included in this study had a minimum follow-up period of one year, as captured in the MIMIC-IV database. The primary outcomes assessed were 28-day and 1-year mortality. Mortality dates were extracted from the database to calculate survival times for each patient.

Statistical analysis

Variables with more than 60% missing data were excluded from the analysis. For variables with less than 5% missing data, mean imputation was applied. For variables with missing data rates between 5% and 60%, multiple imputation using the chained equations method (R software, mice package) was employed (see Supplementary Table S1) [20]. Following normality testing, none of the continuous variables met the assumption of normal distribution. Therefore, all continuous variables are reported as medians with interquartile ranges, while categorical variables are expressed as counts (percentages). The Mann–Whitney U Test was used to compare continuous variables, and the Fisher’s exact test was applied to categorical variables with expected counts below 10. To compare the predictive performance of different SHR levels, receiver operating characteristic (ROC) curves were generated, and areas under the curve (AUC) were calculated. Survival differences between groups, stratified by ROC-derived cut-off values, were further analyzed using Kaplan–Meier survival curves. Additionally, both univariable and multivariable Cox proportional hazards regression models were employed to examine the associations between SHR and 28-day and 1-year mortality. The multivariable Cox model was adjusted for potential confounders, including age, sex, CCI, lactate levels, and insulin administration. Sensitivity analyses were conducted using both Cox models. Statistical analysis was performed using EmpowerStats (version 6.0) and R software (version 4.4.1), with two-sided p-values of < 0.05 considered statistically significant.

Results

Baseline characteristics

A total of 5,025 patients were included in the study (Fig. 1), of whom 63.1% were male. The median age of the cohort was 68.6 years, and the median CCI, an indicator reflecting 16 comorbidities, was 5. Cardiovascular and cerebrovascular diseases were prevalent, accounting for nearly one-third to one-half of the population. A total of 656 patients died within 28 days of ICU admission. Several distinguishing characteristics were identified in the mortality group. Notably, the median age was higher in the mortality group compared with the non-mortality group (73.2 vs. 68.0). Females constituted a significantly higher proportion of the mortality group (42.2%) compared with the non-mortality group (35.9%). The CCI was also markedly higher in the mortality group, with a greater prevalence of acute myocardial infarction, heart failure, cerebrovascular disease, liver disease, and mild to severe diffuse kidney disease.

The median heart rate of the cohort was 101 beats per minute, with a MBP of 58 mmHg. Additionally, lactate levels were elevated at 2.4 mmol/L, and the oxygenation index was 193.3 mmHg. Heart rate, RR, and lactate levels were significantly higher in the mortality group whereas the oxygenation index was lower, although not significantly so (182.7 mmHg vs. 194 mmHg). Patients in the mortality group had prolonged ICU stays and mechanical ventilation times (MVt), with a median ICU stay of 5.7 days compared with 3.3 days in the non-mortality group. Although not statistically significant, a prolonged MVt was also observed in the mortality group.

Among sepsis patients, higher white blood cell counts and a slight prolongation in activated partial thromboplastin time (APTT) were noted. For patients who died within 28 days, there was evidence of worse renal and coagulation function, with a median creatinine level of 1.5 mg/dl and further prolongation of prothrombin time (15.7 s) and APTT (36.7 s). HbA1c levels were comparable between the non-mortality and mortality groups, with a median of 5.9% in all sepsis patients. However, significant differences were observed in SHR values, with higher SHRfirst, SHRmin, SHRmax, and SHRmean values in the mortality group (Table 1).

Table 1 Baseline characteristics of the study population

Fig. 1

Flow chart of the study process. Note: ICU: intensive care unit; MIMIC-IV: Medical Information Mart for Intensive Care IV; HbA1c: glycated hemoglobin

Comparison between various SHRs

Given that SHR has been established as a reliable biomarker for stress response in critically ill patients, we further investigated the association between different SHRs and mortality in sepsis patients. In Fig. 2, the left panel illustrates the predictive efficiency of various SHRs on 28-day mortality, while the right panel depicts their performance for 1-year mortality. Among the SHRs examined, SHRmax exhibited the highest predictive accuracy for both 28-day and 1-year mortality, despite lower AUC values. Specifically, the AUC for SHRmax was 0.632 for 28-day mortality and 0.607 for 1-year mortality. AUC values for other SHRs are also presented in Fig. 2. Cut-off values, derived from ROC curves, were identified as 1.1556 for SHRfirst, 1.1526 for SHRmin, 1.3038 for SHRmax, and 1.3342 for SHRmean 28-day mortality. For 1-year mortality, the cut-offs were 1.1361, 1.1682, 1.3075, and 1.2619, respectively. In addition, the detailed other results provided in the supplementary material (Additional file: Table S2).

Fig. 2

ROC curves for 28-day and 1-year mortality. Note: ROC: receiver operating characteristic, SHR: stress hyperglycemia ratio, AUC: areas under the curve

In addition, a Kaplan–Meier survival analysis was performed, yielding notable results. Statistically significant differences (P < 0.001) in survival rates were observed between higher and lower SHR groups based on the cut-off values, regardless of the method used to calculate SHR. Patients in the higher SHR groups consistently demonstrated lower survival rates, while those in the lower SHR groups had higher survival rates. Figures 3 and 4 present the survival curves for 28-day and 1-year mortality, respectively. These findings underscore the predictive value of both SHRfirst and SHRmax in mortality outcomes, with higher SHRs being associated with increased mortality risk.

Fig. 3

Kaplan–Meier survival curves for different SHRs and 28-day mortality, with cut-off values generated from the ROC curves in Fig. 2. Note: ROC: receiver operating characteristic, SHR: stress hyperglycemia ratio. All P < 0.001

Fig. 4

Kaplan–Meier survival curves for different SHRs and 1-year mortality, with cut-off values generated from the ROC curves in Fig. 2. Note: ROC: receiver operating characteristic, SHR: stress hyperglycemia ratio. All P < 0.001

The correlation between different SHRs and mortality

To quantify the correlation between SHRs and mortality, further analysis was conducted (Table 2) using the Cox proportional hazards regression model. Statistically significant associations with mortality were observed in both univariate and multivariate models. Variables such as age, gender, CCI, the Sequential Organ Failure Assessment (SOFA) score, lactate levels, and insulin usage were identified as potential confounders based on univariate analysis and their clinical significance, and were subsequently adjusted for in the multivariate model. All SHRs, irrespective of the calculation method, demonstrated a significant correlation with increased 28-day mortality and 1-year mortality, though to varying degrees. For instance, each unit increase in SHRmax was associated with a 77% increase in 28-day mortality in the univariate model and a 71.6% increase in the multivariate model. The hazard ratios (HRs) and 95% confidence intervals (CIs) were 1.770 (1.580, 1.982) and 1.716 (1.480, 1.989), respectively.

Furthermore, the correlation between SHRs and mortality was assessed in subgroups stratified by cut-off values, with detailed results provided in the supplementary material (Additional file: Table S3).

Table 2 HR and 95% CIs of the correlations between SHRs and mortality

Sensitivity analysis

Given the influence of various factors on blood glucose levels, a sensitivity analysis was performed, focusing on patients with and without a history of DM. Age, gender, CCI, SOFA score, lactate levels, and insulin usage were adjusted for in the multivariate model (Table 3). The analysis revealed that all SHRs were significantly associated with both 28-day and 1-year mortality, regardless of DM status. However, stronger correlations were observed in non-DM patients compared to DM patients. For instance, in the multivariate model, SHRfirst was associated with an HR of 2.446 for 28-day mortality in non-DM patients, compared to 1.702 in DM patients. Conversely, SHRmin showed higher HRs in DM patients. For 28-day mortality in the multivariate model, the HRs and 95% CIs were 2.911 (1.880, 4.507) in DM patients and 2.867 (1.764, 4.661) in non-DM patients with similar trends observed for 1-year mortality.

Table 3 Sensitivity analyses on DM and non-DM populations

Discussion

Infection is often a common cause of the host stress response. Given that they are the representative population of the dysregulated stress response, this study focused specifically on sepsis patients. While it is well established that SHR serves as a reliable indicator of the stress response, there is a notable lack of research investigating the association between various SHRs and sepsis patient outcomes. Our study yielded the following key findings: (1) Significant differences were detected in SHRfirst, SHRmin, SHRmean, and SHRmax between patients in the mortality and non-mortality groups, with all four SHR metrics being markedly elevated in the mortality group. (2) Among the four SHR metrics, SHRmax demonstrated the strongest predictive performance for both 28-day and 1-year mortality. (3) ROC curve analysis identified distinct cut-off values for each SHR metric, all of which showed robust predictive capability, with Kaplan–Meier survival analysis further confirming that patients with elevated SHR values experienced poorer prognoses. (4) In addition to univariate analysis, multivariate analysis—adjusted for confounding factors—revealed a significant association between various SHRs and patient prognosis. (5) Sensitivity analysis demonstrated that SHR had significant prognostic value in both diabetic and non-diabetic patients, though the association was more pronounced in those with diabetes.

The study cohort consisted of 5,025 patients observed over a 28-day period following ICU admission, during which 656 patients died, while 4,369 survived. The mortality group exhibited distinct characteristics: (1) Patients in this group were older, with an increased proportion of females. (2) CCI scores were significantly higher in the mortality group, particularly for cardiovascular and cerebrovascular diseases, as well as liver and kidney diseases. (3) Patients in the mortality group had significantly worse vital signs, including higher heart rate, increased RR, and lower MBP. Interestingly, these patients had longer ICU stays but slightly shorter overall length of stay (LOS). One hypothesis is that these patients are more critically ill, so the need for ICU treatment is longer, but the high in-hospital mortality leads to shorter LOS, possibly due to higher acuity and in-hospital mortality. (4) Organ dysfunction, particularly of the liver, kidneys, and coagulation system, was prevalent in the mortality group. (5) No significant differences were observed in the maximum temperature, MVt, oxygenation index, hemoglobin, HbA1c, or other admission indicators between the two groups, raising attention. These parameters, measured at ICU admission, may not fully capture the disease’s progression or severity, and pre-admission interventions could have led to their normalization.

All SHRs are critical for evaluating the prognosis of sepsis patients, with SHRmax emerging as the most reliable predictor. The reasoning behind SHRmax’s superiority in representing the stress response in sepsis, compared with SHRfirst, SHRmin, and SHRmean is straightforward. First, the host stress response in sepsis is intimately linked to blood glucose regulation. Sepsis can provoke a dysregulated stress response, resulting in a sharp increase in blood glucose levels [6]. This stress hyperglycemia arises from hormonal and neural alterations that influence carbohydrate metabolism, and its pathophysiology differs from that of chronic hyperglycemia seen in type 2 DM [7]. In response to stress, activation of the hypothalamic-pituitary axis and the sympathetic nervous system triggers a marked elevation in counter-regulatory hormones (catecholamines, corticosteroids, glucagon, and growth hormone). These hormones enhance gluconeogenesis and glycogenolysis, leading to a significant rise in blood glucose [21, 22]. As blood glucose increases, pancreatic β cells secrete insulin to reduce hepatic glucose production and facilitate glucose uptake and storage in the liver, muscles, and adipose tissue. While insulin is the only hormone that lowers blood glucose, its regulatory role is frequently compromised in critically ill patients. In such cases, pancreatic β cell dysfunction or impaired insulin secretion may occur, often accompanied by insulin resistance due to the dysregulated stress response, making it difficult to control blood glucose levels effectively, ultimately resulting in stress hyperglycemia [23, 24]. Second, SHRmax is a better indicator of the stress response severity because SHR is adjusted for chronic blood glucose levels via HbA1c measurement, thereby directly reflecting acute glycemic fluctuations and their association with the stress response. Additionally, SHRmax minimizes the confounding factors that may obscure the relationship between SHR and the stress response. In critically ill patients, unclear pre-ICU interventions, disease progression, or other factors can trigger an elevated stress response after ICU admission. Initial blood glucose measurements in the ICU may not reliably capture the degree of the stress response due to variability in timing and other factors. However, this limitation can be mitigated by using the maximum blood glucose value within the first 24 h post-admission, which more accurately reflects the severity of the dysregulated stress response. It is important to acknowledge that post-ICU blood glucose levels may not always correlate precisely with the severity of the stress response because of insulin or other therapeutic interventions. Nevertheless, elevated blood glucose levels typically signify a heightened stress response. Our analysis was adjusted to account for insulin use, and we specifically chose a 24-hour window to minimize the instability introduced by therapeutic interventions. Given that SHRmax better represents the degree of the stress response, our findings further suggest that an intensified stress response is often associated with a worse prognosis in sepsis patients.

Patients exhibiting elevated SHR values generally demonstrate poor prognoses, with increased SHR strongly associated with an increased risk of mortality. The observed poor clinical outcomes are likely attributable to an excessively dysregulated stress response. Stress hyperglycemia is highly prevalent in sepsis and has been recognized as an important marker of disease severity in critically ill patients [8]. Previous studies have established an association between SHRfirst and sepsis prognosis. For instance, Yan et al. [11] found that higher SHRfirst values were significantly linked to an increased risk of 28-day and in-hospital mortality in sepsis patients, compared with lower SHRfirst values. However, there is a paucity of studies investigating the correlation between SHRmax, which may better capture the severity of the stress response, and sepsis prognosis. Additionally, limited research has explored the potential utility of SHRmean and SHRmin in this context. Our findings contribute valuable insights into this area, confirming that higher SHR values correspond with more severe stress responses. It is well established that a dysregulated stress response plays a crucial role in sepsis progression, with the degree of this response closely tied to patient prognosis [12]. SHRs may more directly reflect the severity of the stress response. Leveraging the unique feature, this study examines the relationship between SHRs and patient outcomes in sepsis from the perspective of stress response severity. SHRmax emerged as the most accurate predictor of prognosis, suggesting that elevated blood glucose levels are a better indicator of stress response, and that an excessively dysregulated stress response may be a key risk factor for increased mortality in sepsis. Although SHRmean reflects the average state of stress response and was therefore included in this analysis, our study confirmed that the maximum degree of the stress response, rather than the average state, more accurately predicts patient prognosis. Therefore, further studies are warranted to investigate the clinical relevance and significance of SHRmean.

SHRs have been found to be more effective in prognosticating outcomes in sepsis patients without DM, although they remain significant in those with DM. To further evaluate the differential efficacy of SHRs across populations, we conducted a sensitivity analysis. This approach was necessitated by the potential variation in blood glucose dynamics between patients with and without DM, even when exposed to a similar degree of stress response. Notably, our findings revealed that increases in SHRmax, SHRmean, and SHRfirst were more strongly associated with a higher risk of mortality in sepsis patients without DM compared with those with DM. This effect was particularly pronounced when acute blood glucose changes were proportionally similar to chronic levels. Patients with DM exhibited a lower risk of mortality than their non-DM counterparts, a finding that corroborates the work of Li et al. [17], who similarly identified SHRfirst as a predictor of all-cause mortality in sepsis patients without DM. A plausible explanation for this observation is that chronic DM is often characterized by persistent inflammation and oxidative stress, which may activate the endogenous antioxidant response system, thereby mitigating some of the adverse pathophysiological effects of acute inflammation and stress [25, 26]. Of particular note, an inverse association was observed with SHRmin, where patients without DM had a lower risk of mortality compared with those with DM. This may be attributable to the occurrence of hypoglycemia, often resulting from overly aggressive glucose-lowering therapy, a phenomenon more common in patients with DM.

Several additional findings from our study warrant brief discussion: (1) Our results demonstrate a strong correlation between SHRs and sepsis prognosis, indicating that physicians may only need to obtain a single SHR measurement to assess patient outcomes. This substantially enhances the clinical utility and feasibility of SHR as a prognostic tool. (2) A more detailed analysis found that, unlike SHRfirst, SHRmin, and SHRmean, SHRmax was more effective in evaluating prognosis at a lower value. Different clinical interventions, intensity, and frequency most likely cause this result. For example, interventions may be more aggressive with a higher SHRmax, and glucose-raising therapy may even be possible with a lower SHRmin. (3) The AUC values for the different SHRs, while not exceptionally high, varied in sensitivity and specificity, and the results were consistent with expectations. This is because the stress response to sepsis involves multiple physiological mechanisms, including inflammation, immunity, and coagulation, in addition to hyperglycemia. The elevation of blood glucose levels in sepsis patients represents just one facet of the broader dysregulated stress response. He et al. [13] recently demonstrated that combining SHRmax with the SOFA score significantly improves the prognostic evaluation of ICU patients with ACS, underscoring the utility of SHR as a significant prognostic marker in sepsis. Our results similarly suggest that combining SHR with other clinical indicators may enhance prognostic efficacy in sepsis. (4) The study’s findings suggest that various SHRs are consistently associated with sepsis prognosis, regardless of whether patients have DM. This could be attributed to SHR’s ability to account for chronic glycemic status, making it more reflective of acute blood glucose fluctuations caused by physiological stress. The lack of a need to differentiate between patients with or without DM further highlights the clinical applicability of SHR. (5) Our study strongly suggests that SHR plays a critical role in sepsis prognosis, which raises the possibility that blood glucose control may benefit sepsis patients. Historically, blood glucose management strategies have been explored, with Van den Berghe et al. [27] reporting in 2001 that maintaining blood glucose ≤ 110 mg/dl through insulin therapy improved outcomes in critically ill patients. However, subsequent studies have raised concerns about this approach, as strategies focused solely on glycemic control did not consistently improve prognosis and were associated with increased hypoglycemia and mortality [28, 29]. Despite these concerns, SHR offers an advantage over simple blood glucose measurements by more accurately reflecting the severity of the stress response. Future research should investigate whether blood glucose control strategies aimed at reducing SHR levels could improve sepsis outcomes. (6) Furthermore, this study highlights the importance of modulating the stress response in sepsis management. The prior research partially supported this approach: Ma et al. [30] demonstrated that dexmedetomidine can activate the cholinergic anti-inflammatory pathway, ameliorating the dysregulated response following renal ischemia-reperfusion injury. Therefore, interventions aimed at controlling the stress response, mitigating stress hyperglycemia, and subsequently reducing SHRs may hold potential for improving clinical outcomes in sepsis patients.

Some limitations of this study should be acknowledged. As a retrospective analysis utilizing the MIMIC-IV database, certain inherent constraints arise because of the nature of the data extracted. Specifically, the study lacks detailed information on blood glucose management strategies prior to ICU admission and post-discharge treatment, which may introduce potential gaps in treatment-related variables. Although we sought to ensure the robustness of our findings by adjusting for insulin use and other covariates, the possibility of residual unmeasured confounders cannot be excluded. Consequently, the generalizability of our results is limited, necessitating cautious interpretation. Future multicenter prospective studies are warranted to further validate these findings.

Conclusion

SHRs, particularly SHRmax, serve as important indicators of the severity of dysregulated stress response and are significantly associated with both short- and long-term prognosis in sepsis patients. Specifically, elevated SHRs are linked to poor prognostic outcomes. Additionally, SHRs can be effectively used to assess prognosis in sepsis patients with and without DM, with greater utility observed in those without DM. These findings underscore the clinical relevance of SHRs in the management of sepsis. Furthermore, it is plausible to hypothesize that strategies aimed at modulating the stress response and implementing SHR-targeted glycemic control may hold substantial clinical significance, potentially becoming integral components of critical care management. Large-scale prospective studies are essential to further substantiate these conclusions.