If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Stress-induced Hyperglycemia Ratio as an Independent Risk Factor of In-hospital Mortality in Nonresuscitation Intensive Care Units: A Retrospective Study
Address correspondence to: Wang Rui (b. 1976), Department of Endocrinology and Metabolism, the First Hospital of Qinhuangdao, City of Qinhuangdao, Hebei, China 066000
To determine whether the stress-induced hyperglycemia ratio (SHR) is independently associated with in-hospital mortality in critically ill patients in nonresuscitation ICUs.
Methods
In this retrospective cohort study, clinical- and laboratory-related data from patients first admitted to nonresuscitation ICUs were extracted from an open-access database of >50,000 ICU admissions. Patients were assigned to one of two groups according to an SHR threshold of 1.1. The primary end point of this study was the in-hospital mortality rate. The associations between SHR and length of stay in the ICU and hospital, duration of mechanical ventilation use, and vasopressor use were secondary end points. Logistic regression models were established in the analysis of in-hospital mortality risk, and areas under the receiver operating characteristic curve (AUC) were analyzed to investigate the association between the primary end point and SHR used alone or together with the Simplified Acute Physiology Scale (SAPS) II score. The Youden index, specificity, and sensitivity of SHR and SAPS-II were also assessed.
Findings
In this study, 1859 patients were included, 187 of whom (10.06%) died during hospitalization. The group with an SHR of ≥1.1 had a greater in-hospital mortality rate (13.7% vs 7.4%; P < 0.001), longer length of stay both in the ICU and in the hospital, a longer duration of mechanical ventilation use, and a greater rate of vasopressor use. On adjustment for multivariate risk, a 0.1-point increment in SHR was significantly associated with in-hospital mortality (OR = 1.08; 95% CI, 1.00–1.16; P = 0.036). The AUC of the association between risk and the SAPS-II score was significantly greater than that with SHR (0.797 [95% CI, 0.576–0.664] vs 0.620 [95% CI, 0.764–0.830]; P < 0.001). The AUC with SAPS-II + SHR was significantly greater than that with SAPS-II used alone (0.802 [95% CI, 0.770–0.835] vs 0.797 [95% CI, 0.764-0.830]; P = 0.023). The Youden index, specificity, and sensitivity of SAPS-II + SHR were 0.473, 0.703, and 0.770, respectively.
Implications
Stress-induced hyperglycemia, as evaluated using the SHR, was associated with increased in-hospital mortality and worse clinical outcomes in these critically ill patients in nonresuscitation ICUs. SHR was an independent risk factor for in-hospital mortality, and when used together with the SAPS-II, added to the capacity to predict mortality in these patients in nonresuscitation ICUs. Prospective data are needed to validate the capacity of SHR in predicting in-hospital mortality in patients in the nonresuscitation ICU.
It has been correlated with a severe inflammatory response and has been established as a risk factor for morbidity and mortality in patients in the ICU.
In several studies, stress-induced hyperglycemia was associated with increased mortality and greater ICU length of stay (LOS) in critically ill patients; however, the threshold for defining stress-induced hyperglycemia ranges broadly, from 100 to 300 mg/dL (5.6–16.7 mmol/L).
Degree of hyperglycemia independently associates with hospital mortality and length of stay in critically ill, nondiabetic patients: results from the ANZICS CORE binational registry.
There is no effective correction for chronic hyperglycemia using the absolute value of glucose concentration in the context of a high prevalence of diabetes in the ICU. The hemoglobin (Hb)-A1c value could be considered as a measure of chronic hyperglycemia in patients with overt and latent diabetes. The glucose level on admission in critically ill patients represents the extent of acute hyperglycemia.
The stress-induced hyperglycemia ratio (SHR) is calculated based on the glucose and HbA1c concentrations on admission, and has been used to estimate stress-induced hyperglycemia while taking chronic hyperglycemia into consideration.
It is hypothesized that SHR might be a reliable predictor of stress-induced hyperglycemia in patients with or without diabetes, even in those with latent diabetes.
Medical societies recommend different targets for glucose control in patients in the resuscitation ICU and nonresuscitation ICUs.
One previously published study supported that the SHR was related to worse outcomes in ICU patients but concluded that the SHR did not predict increased mortality.
However, that study did not distinguish general ICU patients hospitalized for resuscitation purpose after surgery have extremely low mortality, as is well known and has been demonstrated in a previous study.
Impact of the time-weighted average glucose concentration and diabetes on in-hospital mortality in critically ill patients older than 75 years: a retrospective cohort study.
whether SHR is associated with mortality in critically ill patients in the nonresuscitation ICU remains unclear.
The present study used SHR as a metric for evaluating stress-induced hyperglycemia during ICU management in an attempt to determine the impact of stress-induced hyperglycemia on short-term prognosis in critically ill patients in the nonresuscitation ICU.
Materials and Methods
Study Population
In this retrospective cohort study, data were extracted from the MIMIC-III database (https://mimic.physionet.org),
an openly accessible database of >50,000 ICU admissions dated only from 2008 to 2014 at the Beth Israel Deaconess Medical Center (Boston, Massachusetts). The examination of Human Protection Study Participants (record ID 39124769) was passed to request access to this database. The study protocol was approved by the institutional review boards at Beth Israel and the Massachusetts Institute of Technology (Cambridge, Massachusetts). All patient information was processed and deidentified, and individual informed consent was waived.
Among the 46,570 first admissions, data from only patients aged >18 years were included. Exclusion criteria were death or discharge within 24 hours after admission, the lack of a record of glucose concentration or HbA1c measured on admission, and treatment in the resuscitation ICU (Figure 1). The estimated mean blood glucose level was calculated as 28.7 × HbA1c – 46.7 mg/dL. The SHR was defined as the ratio of serum blood glucose on admission to estimated mean blood glucose. The included patients were assigned to one of two groups based on an SHR threshold of 1.1 according to the study by Roberts et al.
The data were extracted using structured query language (SQL). Admission type was classified as elective, emergency, or urgent, depending on the severity of the condition. The Simplified Acute Physiology Scale (SAPS)-II score was used to evaluate disease severity during the ICU stay.
Vasopressor use was defined as continuous infusion of at least one catecholamine for at least 4 hours during the ICU stay. Patients with diabetes were identified using the ICD-9 codes 250xx and 3572.
Data on blood gas, chemistry, and point-of-care glucose measurements were collected. Glucose concentration on admission was recognized as the concentration first recorded within 24 hours after admission or last recorded within 4 hours before admission. HbA1c was the concentration first recorded during ICU treatment. Hypoglycemia was defined as a glucose concentration on admission of ≤70 mg/dL. Glucose coefficient of variation (CV) was defined as the standard variation divided by the mean of the glucose concentrations on admission.
End Points
The primary end point of this study was the in-hospital mortality rate. Other end points included the lengths of stay (LOSs) in the hospital and ICU (only the first admission to the ICU was used in the calculation of the LOS), vasopressor use, duration of mechanical ventilation, and the need for continuous renal replacement treatment.
Statistical Analysis
R shareware version 3.6.3 (R Foundation for Statistical Computing, Vienna, Austria) was used to perform the statistical analyses. Normally distributed data are expressed as mean ± SD, non-normally distributed numeric data are shown as median (interquartile range), and enumeration data are expressed as numbers and percentages. The significance of differences between groups was determined using the t test, U test, or χ2 test according to data type. Statistical significance was set at P < 0.05.
Univariate and multivariate logistic regression models were established to predict the outcomes. Three multivariate models were established. Model I was adjusted for age, sex, ethnicity, and body mass index (BMI). Model II was additionally adjusted for admission type, SAPS-II score, and Elixhauser comorbidity index. Model III was further adjusted for hypoglycemia status, glucose variation, HbA1c, and diabetes status. The association between SHR and in-hospital mortality was determined by calculating the area under the receiver operating characteristic (ROC) curve (AUC), and the DeLong test was applied for the comparison between two correlated ROC curves.
Due to a large number of patients with missing BMI values, BMI was transformed as a factor variable in the logistic regression analysis. All of the other missing values were simply excluded.
Results
Patient Characteristics
Data from a total of 1859 first admissions to nonresuscitation ICUs were included in this study, including those from 831 patients in the cardiac care unit, 404 patients in the medical ICU, 461 patients in the surgical ICU, and 163 patients in the trauma/surgical ICU (Figure 1).
In this cohort, 187 patients (10.06%) died in the hospital. The distributions of age, sex, ethnicity, BMI, admission type, diabetes and hypoglycemia rates, admission glucose concentration, glucose CV, and mean insulin use per day were similar between the surviving and nonsurviving groups. The SHR was greater, HbA1c level was lower, steroid use was more likely, the prevalence of sepsis was greater, disease conditions (as assessed using SAPS-II score) were more severe, and the rates of comorbidities were greater in the nonsurviving group (Table I).
Table IClinical characteristics of surviving and nonsurviving critically ill patients in nonresuscitation ICUs.
Characteristic
Survivors
Nonsurvivors
P
(n = 1672)
(n = 187)
Age, no. (%)
0.096
<45 y
165 (33.1)
24 (25.5)
45–<65 y
163 (32.7)
27 (28.7)
≥65 y
171 (34.3)
43 (45.7)
Female, no. (%)
1025 (61.3)
107 (57.2)
0.314
White ethnicity, no. (%)
1137 (68.0)
122 (65.2)
0.494
Body mass index, mean ± SD, kg/m2
29.58 ± 11.48
29.39 ± 6.68
0.730
Admission type, no. (%)
0.175
Elective
46 (2.8)
1 (0.5)
Emergency
1584 (94.7)
182 (97.3)
Urgent
42 (2.5)
4 (2.1)
Diabetes, no. (%)
712 (42.6)
69 (36.9)
0.157
Admission glucose, mean ± SD, mg/dL
163.21 ± 100.58
176.92 ± 79.22
0.072
HbA1c, mean ± SD, %
6.83 ± 2.12
6.47 ± 1.43
0.024
SHR, mean ± SD
1.11 ± 0.52
1.30 ± 0.52
<0.001
Hypoglycemia, no. (%)
86 (8.6)
15 (11.2)
0.409
Glucose CV, mean ± SD
0.21 ± 0.16
0.20 ± 0.14
0.576
Insulin amount, mean ± SD, IU/d
52.46 ± 118.19
59.41 ± 104.14
0.441
Steroid use, no. (%)
38 (2.3)
11 (5.9)
0.007
SAPS-II score, mean ± SD
30.96 ± 11.76
46.26 ± 15.25
<0.001
Elixhauser van Walraven comorbidity score, median [interquartile range]
Association Between Stress-induced Hyperglycemia and Outcomes
Using an SHR threshold of 1.1, the group with the greater SHR (≥1.1) had a greater rate of in-hospital mortality (13.7% vs 7.4%; P < 0.001), longer LOSs in both the ICU and hospital, a longer duration of mechanical ventilation, as well as a greater prevalence of vasopressor use (Table II).
Table IIEffects of stress-induced hyperglycemia on clinical outcomes.
Effect
SHR <1.1
SHR ≥1.1
P
Vasopressor use, no. (%)
190 (17.8)
187 (23.7)
0.002
Mechanical ventilation, median [interquartile range], h
0.00 [0.00, 14.17]
0.00 [0.00, 35.38]
<0.001
CRRT, no. (%)
12 (1.1)
16 (2.0)
0.165
Hospital LOS, median [interquartile range], d
6.09 [3.73, 10.68]
6.91 [4.15, 12.22]
0.001
ICU LOS, median [interquartile range], d
2.17 [1.48, 4.03]
2.88 [1.81, 5.68]
<0.001
In-hospital mortality, no. (%)
79 (7.4)
108 (13.7)
<0.001
CRRT = continuous renal replacement treatment; LOS = length of stay; SHR = stress-induced hyperglycemia ratio.
In the crude analysis, a 0.1-point increment in SHR was significantly associated with a greater risk for in-hospital mortality (OR = 1.09; 95% CI, 1.04–1.14; P < 0.001). After adjustment for the risk for multivariates, a 0.1-point increment in SHR remained significantly associated with in-hospital mortality in models I, II, and III (Table III).
Table IIICrude and after-adjustment analyses of relationships between mortality and 0.1 stress-induced hyperglycemia ratio increment in critically ill patients in nonresuscitation ICUs.
Parameter
OR
95% CI
P
Crude
1.09
1.04–1.14
<0.001
Model I
1.10
1.05–1.15
<0.001
Model II
1.06
1.00–1.11
0.034
Model III
1.08
1.00–1.16
0.036
Model I was adjusted for age, sex, and ethnicity, and body mass index. Model II was adjusted additionally for the admission type, SAPS-II score, and Elixhauser van Walraven comorbidity index. Model III was further adjusted for hypoglycemia, glucose variation, HbA1c, and diabetes status.
The AUC with SAPS-II–defined risk was significantly greater than that with SHR (0.797 [95% CI, 0.576–0.664] vs 0.620 [95% CI, 0.764–0.830]; P < 0.001). The AUC with SAPS-II + SHR was significantly greater than that with SAPS-II used alone (0.802 [95% CI, 0.770-0.835] vs 0.797 [95% CI, 0.764–0.830]; P = 0.023) (data not shown). The Youden index, specificity, and sensitivity with SAPS-II + SHR were 0.473, 0.703, and 0.770, respectively (Figure 2).
Figure 2Receiver operator characteristic curves for stress-induced hyperglycemia ratio (SHR), Simplified Acute Physiological Scale (SAPS) II score, and combined SHR + SAPS-II. The numbers shown beside cutoff points were Youden index (specificity, sensitivity).
In this retrospective study of data from critically ill patients in nonresuscitation ICUs, stress-induced hyperglycemia, as measured using the SHR, was independently associated with in-hospital mortality after adjustment for the risk for disease severity, hypoglycemia, diabetes, and glucose variation. SHR used in conjunction with SAPS-II score added to the capacity of predicting in-hospital mortality in patients in nonresuscitation ICUs.
According to previously published studies, stress-induced hyperglycemia may cause more severe adverse outcomes than chronic hyperglycemia.
The interaction of acute and chronic glycemia on the relationship of hyperglycemia, hypoglycemia, and glucose variability to mortality in the critically ill.
; on the other hand, the stress response in patients with acute disease is usually due to consequential increased secretion of stress hormones, such as cytokines, cortisol, and catecholamines.
All of these factors lead to poor prognosis in critically ill patients with acute hyperglycemia, also known as stress-induced hyperglycemia. The SHR indicator considers chronic hyperglycemia in the evaluation of glycemic control, making it a good indicator of stress-induced hyperglycemia. Previously published studies have mostly been focused on the impact of stress-induced hyperglycemia on general ICU patients or on a specific type of disease.
Impact of the time-weighted average glucose concentration and diabetes on in-hospital mortality in critically ill patients older than 75 years: a retrospective cohort study.
Intraoperative infusion of dexmedetomidine for prevention of postoperative delirium and cognitive dysfunction in elderly patients undergoing major elective noncardiac surgery: a randomized clinical trial.
Is the Enhanced Recovery After Surgery (ERAS) program effective and safe in laparoscopic colorectal cancer surgery? A meta-analysis of randomized controlled trials.
Therefore, it is necessary to evaluate the overall effect of SHR on the prognosis of patients in the resuscitation ICU. In the present study, each 0.1-point increment in SHR was associated with a 8% increased risk for in-hospital mortality after adjustment for multivariates, and increased SHR was related to worse clinical outcomes, including greater ICU LOS and mechanical ventilation, similar to the findings from other populations.
In a prior study, the use of SHR-assessed stress-induced hyperglycemia together with the acute physiology, age, chronic health evaluation (APACHE) II score added to the capacity to predict prognosis in ICU patients.
The findings from the present study in a nonresuscitation ICU also supported this conclusion that SHR added the SAPS-II score to a significantly increased capacity to predict the risk for in-hospital mortality. It was not surprising that the increase in AUC was small when SHR was used with the SAPS-II score. Given the large number of variables already included in the process of calculating the SAPS-II score, the predictive power was small but not trivial.
However, the relationship between absolute glucose control and mortality in critically ill patients remains controversial. Some studies have reported a significant association between glucose control and adverse short-term outcomes,
A prospective randomised multi-centre controlled trial on tight glucose control by intensive insulin therapy in adult intensive care units: the Glucontrol study.
In the present study, admission glucose and glucose CV were not significantly different between the surviving and nonsurviving groups in the nonresuscitation ICU. Therefore, the absolute glucose level and glucose variation did not seem to be reliable predictors of prognosis in critically ill patients, at least not in those in the resuscitation ICU. However, the SHR seems to be an independent predictor of in-hospital mortality across the glycemic spectrum in critically ill patients.
Presently, there are still many controversies. Regarding glucose targets in ICU patients, a target of 140 to 180 mg/dL is recommended for most critically ill patients,
As stress-induced hyperglycemia plays an important role in predicting mortality in nonresuscitation ICU patients, the translation of its clinical significance into specific values applicable to clinical practice is another issue of concern. Future studies, such as prospective, randomized, controlled studies, should shed light on this topic.
The present study had several strengths. First, data from critically ill patients from nonresuscitation ICUs were selected, excluding the confounding factors of low mortality in patients in resuscitation ICUs. Second, SHR—which was more reliable than was the absolute value of glucose concentration when considering HbA1c and admission glucose together—was employed to evaluate the extent of stress-induced hyperglycemia in the present study. Third, the sample size in this study was relatively large, and the publicly available data had credibility in critical care research. Fourth, three models were built and univariate and multivariate regression analyses were used to investigate the association between stress-induced hyperglycemia and mortality, ensuring the robustness of the results of this study. Fifth, ROC curve analysis was employed in this study, which indicated that the SHR when added to the SAPS-II score increased the capacity to predict in-hospital mortality in the nonresuscitation ICU. To the best of our knowledge, this is the first study to determine whether the SHR is an independent predictor of the short-term prognosis of patients in the nonresuscitation ICU.
This study had some limitations. First, this was a retrospective single-center study in critically ill patients, and the results need to be further validated in prospective, multicenter studies. Second, the data in this study were extracted from a public database, and there were some missing values for variables, such as HbA1c. This study selected only those without missing HbA1c values to calculate SHR, which might have caused a latent selection bias. Third, the observational nature of this study precludes proof of causality, and the results can be considered only as hypothesis generating.
Conclusions
In the present study, stress-induced hyperglycemia, as evaluated using SHR, was associated with increased in-hospital mortality and worse clinical outcomes in critically ill patients in the nonresuscitation ICU. SHR was an independent risk factor for in-hospital mortality and, when used with the SAPSII score, increased the capacity to predict mortality in patients in nonresuscitation ICUs.
Acknowledgments
Thanks to the team who established and published the MIMIC-III database for the public; this was quite a selfless gift to those eager to elucidate hypotheses but who are challenged by a lack of available data. Thanks to Dr. Li Haibo for his sharing of some brilliant R packages in statistical analysis. Thanks to Dr. Yang Qilin for his help with the basic training of clinical study design. Thanks to Dr. Ma Chunming for raising important suggestions in the revision of the manuscript.
Author Contributions
Tian Yiming was responsible for data cleaning, statistical analysis, and article writing. Wang Rui (b. 1983) and Zhang Mengmeng participated in data cleaning. Li Tao and He Yang participated in the development of the research protocol. Wang Rui (b. 1976) was responsible for the design of the research protocol.
Declaration of Interest
The authors have indicated that they have no conflicts of interest with regard to the content of this article.
Degree of hyperglycemia independently associates with hospital mortality and length of stay in critically ill, nondiabetic patients: results from the ANZICS CORE binational registry.
Impact of the time-weighted average glucose concentration and diabetes on in-hospital mortality in critically ill patients older than 75 years: a retrospective cohort study.
The interaction of acute and chronic glycemia on the relationship of hyperglycemia, hypoglycemia, and glucose variability to mortality in the critically ill.
Intraoperative infusion of dexmedetomidine for prevention of postoperative delirium and cognitive dysfunction in elderly patients undergoing major elective noncardiac surgery: a randomized clinical trial.
Is the Enhanced Recovery After Surgery (ERAS) program effective and safe in laparoscopic colorectal cancer surgery? A meta-analysis of randomized controlled trials.
A prospective randomised multi-centre controlled trial on tight glucose control by intensive insulin therapy in adult intensive care units: the Glucontrol study.
00413-1&id=ga1.jpg){kind=link}
00413-1&id=gr1.jpg){kind=link}
00413-1&id=gr2.jpg){kind=link}