Secondary Logo

Journal Logo

Admission Hyperglycemia as a Prognostic Indicator in Trauma

Yendamuri, Saikrishna MBBS; Fulda, Gerard J. MD; Tinkoff, Glen H. MD

The Journal of Trauma: Injury, Infection, and Critical Care: July 2003 - Volume 55 - Issue 1 - p 33–38
doi: 10.1097/01.TA.0000074434.39928.72

Objective: The purpose of this study was to assess the utility of two levels of hyperglycemia as predictors for mortality and infectious morbidity in traumatically injured patients.

Methods: All patients ≥ 17 years old presenting to a Level I trauma center as a “trauma alert” or a “trauma code” from January 1, 2000, through December 31, 2000, were reviewed. Hypoglycemic patients (glucose concentration < 70 mg/dL) were excluded (n = 4). Patients were considered hyperglycemic with an admission glucose concentration > 200 mg/dL (moderate hyperglycemia) or an admission glucose concentration in the upper quartile for the group (mild hyperglycemia [glucose concentration > 135 mg/dL]).

Results: Seven hundred thirty-eight patients were included in the study. Hyperglycemia was associated with increased mortality among both patients with moderate hyperglycemia (34.1% vs. 3.7%,p< 0.01) and those with mild hyperglycemia (15.5% vs. 2%,p< 0.01) compared with corresponding normoglycemic groups. Hyperglycemia proved to be an independent predictor of mortality and of hospital and intensive care unit length of stay after multiple logistic regression while controlling for age, Injury Severity Score, Revised Trauma Score, and gender. Infectious complications, including pneumonia (9.4% vs. 2%,p= 0.001), urinary tract infections (6.6% vs. 1.4%,p= 0.001), wound infections (4.9% vs. 0.6%,p= 0.039), and bacteremia (5% vs. 1.1%,p= 0.004), were significantly increased in patients with elevated glucose concentrations. Hyperglycemia is an independent predictor of increased infectious morbidity controlling for age, gender, and Injury Severity Score in multiple logistic regression models.

Conclusion: Hyperglycemia independently predicts increased intensive care unit and hospital length of stay and mortality in the trauma population. It is associated with increased infectious morbidity. These associations hold true for mild hyperglycemia (glucose concentration > 135 mg/dL) and moderate hyperglycemia (glucose concentration > 200 mg/dL).

From the Department of Surgery, Section of Trauma and Critical Care, Christiana Care Health Services (S.Y., G.J.F., G.H.T.), Newark, Delaware, and Department of Surgery, Jefferson Medical College of Thomas Jefferson University (G.J.F., G.H.T.), Philadelphia, Pennsylvania.

Submitted for publication October 8, 2002.

Accepted for publication March 10, 2003.

Presented at the 61st Annual Meeting of the American Association for the Surgery of Trauma, September 26-28, 2002, Orlando, Florida.

Address for reprints: Glen H. Tinkoff, MD, Suite 2121 MAP 2, 4735 Ogletown-Stanton Road, Newark, DE 19713; email:

Similar injuries lead to remarkably different outcomes in similar populations. Specific individuals seem to respond to the stressful situation of trauma differently; some are discharged after a relatively uneventful hospital course, whereas others develop a complicated course, with a few dying as a result of their injury. For this reason, several models have been developed to predict outcome after injury. Most of these models use physiologic and/or anatomic information to predict outcome.1,2 Laboratory tests used as markers of such adverse outcomes include base deficit, serum lactate, serum glucose, transferrin, C-reactive protein, albumin, and cholesterol.3 Several studies have demonstrated the association between glucose concentrations and outcome in both head trauma and nontrauma situations.4-6 Few studies have looked at this association within the trauma population not restricted to head trauma. Clinically significant hyperglycemia has traditionally been defined as a serum glucose concentration > 200 mg/dL.7-9 However, studies have demonstrated associations of hyperglycemia far below this level with adverse outcomes,6 encouraging a search for lower cutoffs. The association between hyperglycemia and increased infectious morbidity has an established biologic and clinical basis.9-12 In this study, we investigate the usefulness of admission serum glucose as a prognostic variable and its relationship to patient outcome and infectious morbidity in patients with traumatic injuries.

Back to Top | Article Outline


Approval for access to patient data were obtained from the Institutional Review Board of the Christiana Care Health System. All adult patients (17 years of age and older) presenting to Christiana Hospital, a Level I trauma center, from January 1, 2000, through December 31, 2000, as a “trauma alert” or a “trauma code” were included. Definitions of trauma alert and trauma code at this hospital have been published previously.13 Data were extracted from the Trauma Database and linked to the hospital laboratory data and included admission serum glucose (in milligrams per deciliter), hospital mortality, hospital length of stay, intensive care unit (ICU) length of stay, Injury Severity Score (ISS), Revised Trauma Score (RTS), infectious complications, and demographic data. Patients with hypoglycemia, defined as a serum glucose concentration < 70 mg/dL, were excluded. The study population was divided into groups on the basis of two definitions of hyperglycemia. In the first group, the traditional glucose concentration of > 200 mg/dL was chosen (moderate hyperglycemia). In the second group, a glucose concentration of > 135 mg/dL was chosen (mild hyperglycemia) on the basis of the distribution of serum glucose in the study population. A serum glucose concentration of 135 mg/dL is the 75th percentile of this study population. All statistical analyses were performed using the Statistical Package for the Social Sciences for Windows (SPSS, Inc., Release 10.0.7, 2000). Categorical variables were analyzed with Χ2 analysis with Bonferroni correction. Univariate analysis was performed by analysis of variance. The impact of hyperglycemia on mortality was assessed using multiple logistic regression analysis, controlling for age, gender, ISS, and RTS. The impact of hyperglycemia on hospital length of stay and ICU length of stay was analyzed using multiple linear regressions, controlling for age, ISS, and RTS. Multiple linear and logistic regression modeling was performed in a backward fashion (all variables were included in the initial model, and variables that were not statistically significant were removed sequentially), using a cutoff of 0.1 for the significance of a partial t test to automatically exclude a variable from a model. A combined significance level of 0.05 was chosen for all tests.

Back to Top | Article Outline


Of the 1,921 trauma patients registered in the Trauma Database at Christiana Hospital during the study period, 817 were admitted as a trauma alert or a trauma code. Sixteen of these patients had unrecorded ages and were excluded; 59 were pediatric patients and were excluded. Of the 742 remaining patients, 4 were hypoglycemic and were therefore excluded, as this number was too small for meaningful statistical analysis. The analysis of the data from the remaining 738 patients are presented. A distribution of the cardinal variables in the study population is presented in Table 1.

Table 1

Table 1

Admission hyperglycemia was associated with an increased mortality rate for both mild and moderate hyperglycemia (Table 2). Patients with mild hyperglycemia had an increased mortality rate compared with patients with normal admission glucose concentrations (15.5% vs. 2.0%, p < 0.01); patients with moderate hyperglycemia had an even greater mortality rate compared with the corresponding normoglycemic patients (34.1% vs. 3.7%, p < 0.01).

Table 2

Table 2

The association of individual infectious morbidities with mild and moderate hyperglycemia was then explored with Χ2 analysis (Table 3). The incidence of pneumonia and bacteremia correlated with hyperglycemia irrespective of the cutoff used. However, the incidence of urinary tract infections showed an association with mild hyperglycemia. With moderate hyperglycemia, there was a trend toward an increased incidence of urinary tract infections, but the association did not approach statistical significance. The incidence of wound infections is significantly increased with moderate hyperglycemia, but not with mild hyperglycemia.

Table 3

Table 3

Age, ISS, and gender have been associated previously with an increased mortality rate in trauma patients. The distribution of these factors is shown in Table 4. Except for gender in the moderately hyperglycemic group and proportion of penetrating injury in both comparison groups, all other factors were significantly different between normoglycemic and hyperglycemic patients. To attempt to eliminate the possible confounding effects of these variables on the effect of hyperglycemia on mortality, logistic regression modeling of mortality was performed using age, ISS, RTS, gender, and hyperglycemia. The odds ratios (ORs) and their corresponding 95% confidence intervals of the variables entered into the logistic regression analyses are presented in Table 5. Even while controlling for age, gender, RTS, and ISS, hyperglycemia continued to predict mortality for both mild hyperglycemia (OR, 1.35-10) and moderate hyperglycemia (OR, 1.55-14.2). The multiple regression model of risk of death for both the cutoffs of admission serum glucose is as follows:

Table 4

Table 4

Table 5

Table 5

Ln(0.24 − 1.535 * Gluco200 + 0.056 * Age + 0.065 * ISS − 0.875 * RTS)

Ln(−0.133 − 1.282 * Gluco135 + 0.054 * Age + 0.059 * ISS − 0.889 * RTS)

Gluco200 = 1 if normoglycemic; Gluco135 = 1 if normoglycemic.

Because the same confounding variables that influence mortality may influence septic morbidity, multiple logistic regression analysis was also performed to assess the effect of hyperglycemia on infectious morbidity independent of age, ISS, RTS, and gender. ORs with the corresponding confidence intervals are presented in Table 5. Mild hyperglycemia remained predictive of the incidence of urinary tract infections (OR, 1.21-8.8) and pneumonia (OR, 0.98-8.0) independent of ISS, RTS, age, and gender. Even though the odds ratio in the case of pneumonia spans 1.0, the partial t test for mild hyperglycemia was less than 0.1, and therefore the variable was not excluded from the model.

Hospital length of stay and ICU length of stay were significantly increased in the hyperglycemic population (Table 4). To eliminate the possible confounding effect of age, RTS, and ISS on the effect of hyperglycemia on these variables, multiple linear regression modeling was performed. This analysis proved glucose (partial t test, p < 0.01) to be an independent predictor of hospital length of stay and ICU length of stay independent of ISS (partial t test, p < 0.01), RTS (partial t test, p < 0.01), and age (partial t test, p = 0.016).

Back to Top | Article Outline


Among trauma patients, the association of outcome with hyperglycemia has been studied extensively in brain-injured patients, in whom it portends a poorer prognosis.4,14 Limited data exist in other trauma populations.15 This study looks at the prognostic value of admission serum glucose in the trauma population as a whole. In the population studied, admission serum hyperglycemia was strongly associated with increased mortality, hospital length of stay, and ICU length of stay. This remains true even if hyperglycemia is defined by a serum glucose concentration much lower than that which has been used traditionally by trauma surgeons. In the past, hyperglycemia was defined as a serum glucose concentration > 200 mg/dL. This number was based on in vitro data that were extrapolated to clinical use in diabetic patients. For trauma patients and other critically ill patients, there is no clear definition of hyperglycemia. Recent data encourage us to find new cutoffs, particularly in nondiabetic individuals. Some investigators have found better results with tighter control of serum glucose than a concentration of < 200 mg/dL.16 We chose a concentration of 135 mg/dL (75th percentile) on the basis of the distribution of serum glucose in the study population. Because hyperglycemia is a stress response and has been associated with increased ISS,17,18 this association may represent a relationship of mortality with increasing injury severity. Other variables such as age19,20 and gender may also influence such a response. The multiple logistic regression analysis controlling for ISS, RTS, age, and gender demonstrates that hyperglycemia, both mild and moderate, is indeed an independent predictor of mortality. It is interesting to note that the odds ratios of hyperglycemia in predicting mortality are much higher than those with ISS, which is itself an excellent predictor of outcome. Multiple linear regression analysis controlling for ISS, RTS, and age similarly proves that hyperglycemia is an independent predictor of hospital length of stay and ICU length of stay. The association of hyperglycemia with outcome provides another variable that can be easily obtained early in the clinical course of traumatically injured patients that may allow the ability to triage these patients appropriately and optimize resource use.

Several studies have demonstrated the increased risk of infection in diabetic patients. Given the evidence of immunosuppression induced by hyperglycemia,7,9,11,12,21-24 it is not surprising that infectious morbidity is associated with hyperglycemia, as shown in Table 3. Two associations are of particular interest. The first is the increased incidence of urinary tract infections, pneumonia, and bacteremia with the stricter definition of hyperglycemia. This may mean that tighter control of glucose than has been used in the past may be necessary to prevent these complications. The second association is the dependence of the incidence of wound infections on the level of hyperglycemia. Only moderate hyperglycemia, not mild hyperglycemia, is associated with an increased incidence of wound infection. It is to be noted that the number of patients with wound infections in this study is small, possibly impairing the statistical power required to detect differences between the comparison groups. It may also be that different areas of the body respond to hyperglycemia in different ways; this is supported by the demonstration of different effects of hyperglycemia on macrophages in different areas.22 Because this effect of hyperglycemia may be confounded by ISS, RTS, gender, and age, multiple logistic regression analyses were performed with age, ISS, RTS, and gender. The analyses proved that mild hyperglycemia is an independent predictor of pneumonia and urinary tract infection in the trauma population. In these models, moderate hyperglycemia tends to be associated with increased rates of infection, although independent association does not reach statistical significance. Although this may represent a confounding effect of the other variables inserted into the model, it may also represent lack of sufficient power to detect independent associations. It is also interesting to note that gender is an independent predictor of urinary tract infections. As may be expected, urinary tract infections are significantly more common in women than in men (5.6% vs. 1.5%, p < 0.05).

The clear association between hyperglycemia and adverse outcome demonstrated in this study raises the question of whether hyperglycemia is a cause versus being simply a marker of adverse outcome. Previous work in animal models demonstrated that injury leads to a hyperglycemic response and that treatment of animals with insulin leads to better outcome after hypovolemic injury.25 However, the same investigators also demonstrated a potentially beneficial effect of hyperglycemia in the decompensatory phase of hemorrhagic shock.26 Metabolic studies suggest that hyperglycemia by itself may promote proteolysis.27 This is apart from the susceptibility to infection that hyperglycemia is known to induce.

This study has several limitations. It does not analyze the effect of hypoglycemia on outcome, as glucose distribution may cause a bimodal increase in mortality. It does not separate diabetic from nondiabetic patients, which may be a confounding factor. It does not take into account the effect of administration of medications with high dextrose content or the administration of steroids, which may confound laboratory values.

Back to Top | Article Outline


We have demonstrated that admission hyperglycemia is associated with increased mortality in the trauma population. It is an independent predictor of patient outcome. Hyperglycemia is also associated with infectious morbidity in trauma patients. This may contribute to the poorer patient outcome associated with it. These relationships hold true even for more stringent definitions of hyperglycemia than those that have been conventionally used. Although a randomized, controlled trial will be needed to support cause and effect, intensive control of serum glucose concentration may improve outcome in trauma patients, as has been demonstrated in nontrauma ICU patients.

Back to Top | Article Outline


We thank Susan Mascioli, RN, for her assistance with biostatistics and Karen McFadden for her editorial assistance.

Back to Top | Article Outline


1. Senkowski CK, McKenney MG. Trauma scoring systems: a review. J Am Coll Surg. 1999;189:491-503.
2. Van Camp LA, Delooz HH. Current trauma scoring systems and their applications. Eur J Emerg Med. 1998;5:341-353.
3. Dunham CM, Frankenfield D, Belzberg H, Wiles CE III, Cushing B, Grant Z. Inflammatory markers: superior predictors of adverse outcome in blunt trauma patients? Crit Care Med. 1994;22:667-672.
4. Lannoo E, Van Rietvelde F, Colardyn F, et al. Early predictors of mortality and morbidity after severe closed head injury. J Neurotrauma. 2000;17:403-414.
5. Graf WD, Cummings P, Quan L, Brutocao D. Predicting outcome in pediatric submersion victims. Ann Emerg Med. 1995;26:312-319.
6. Gore DC, Chinkes D, Heggers J, Herndon DN, Wolf SE, Desai M. Association of hyperglycemia with increased mortality after severe burn injury. J Trauma. 2001;51:540-544.
7. Pomposelli JJ, Baxter JJ III, Babineau TJ, et al. Early postoperative glucose control predicts nosocomial infection rate in diabetic patients. JPEN J Parenter Enteral Nutr. 1998;22:77-81.
8. Zerr KJ, Furnary AP, Grunkemeier GL, Bookin S, Kahere V, Starr A. Glucose control lowers the risk of wound infection in diabetics after open heart operations. Ann Thorac Surg. 1997;63:356-361.
9. Golden SH, Peart-Vigilance C, Kao WH, Brancati FL. Perioperative glycemic control and the risk of infectious complications in a cohort of adults with diabetes. Diabetes Care. 1999;22:1408-1414.
10. Demerdash TM, Seyrek N, Smogorzewski M, Marcinkowski W, Nasser-Moadelli S, Massry SG. Pathways through which glucose induces a rise in [Ca2+]i of polymorphonuclear leukocytes of rats. Kidney Int. 1996;50:2032-2040.
11. Hostetter MK. Handicaps to host defense: effects of hyperglycemia on C3 and Candida albicans. Diabetes. 1990;39:271-275.
12. Kjersem H, Hilsted J, Madsbad S, Wandall JH, Johansen KS, Borregaard N. Polymorphonuclear leucocyte dysfunction during short term metabolic changes from normo- to hyperglycemia in type 1 (insulin dependent) diabetic patients. Infection. 1988;16:215-221.
13. Tinkoff GH, O'Connor RE, Fulda GJ. Impact of a two-tiered trauma response in the emergency department: promoting efficient resource utilization. J Trauma. 1996;41:735-740.
14. Chiaretti A, De Benedictis R, Langer A, et al. Prognostic implications of hyperglycaemia in paediatric head injury. Childs Nerv Syst. 1998;14:455-459.
15. Kassum DA, Thomas EJ, Wong CJ. Early determinants of outcome in blunt injury. Can J Surg. 1984;27:64-69.
16. Van den Berghe GW, Wouters P, Weekers F, et al. Intensive insulin therapy in critically ill patients. N Engl J Med. 2001;345:1359-1367.
17. Kenney PR, Allen-Rowlands CF, Gann DS. Glucose and osmolality as predictors of injury severity. J Trauma. 1983;23:712-719.
18. Pomerantz WJ, Hashkes PJ, Succop PA, Dowd MD. Relationship between serum glucose and injury severity score in childhood trauma. J Pediatr Surg. 1999;34:1494-1498.
19. Desai D, March R, Watters JM. Hyperglycemia after trauma increases with age. J Trauma. 1989;29:719-723.
20. Ernst AA, Casaletto J, Nick TG, Weiss SJ. Serum glucose levels in elder trauma victims. Acad Emerg Med. 1999;6:1177-1779.
21. Khaodhiar L, McCowen K, Bistrian B. Perioperative hyperglycemia, infection or risk? Curr Opin Clin Nutr Metab Care. 1999;2:79-82.
22. Kwoun MO, Ling PR, Lydon E, et al. Immunologic effects of acute hyperglycemia in nondiabetic rats. JPEN J Parenter Enteral Nutr. 1997;21:91-95.
23. Oldenborg PA, Sehlin J. Hyperglycemia in vitro attenuates insulinstimulated chemokinesis in normal human neutrophils: role of protein kinase C activation. J Leukoc Biol. 1999;65:635-640.
24. von Kanel R, Mills PJ, Dimsdale JE. Short-term hyperglycemia induces lymphopenia and lymphocyte subset redistribution. Life Sci. 2001;69:255-262.
25. Swerlick RA, Drucker NA, McCoy S, Pearce FJ, Drucker WR. Insulin effectiveness in hypovolemic dogs. J Trauma. 1981;21:1013-1021.
26. Pearce FJ, Drucker WR. Glucose infusion arrests the decompensatory phase of hemorrhagic shock. J Trauma. 1987; 27:1213-1220.
27. Flakoll PJ, Hill JO, Abumrad NN. Acute hyperglycemia enhances proteolysis in normal man. Am J Physiol. 1993;265(5 pt 1):E715-E721.
Back to Top | Article Outline


Dr. Philip S. Barie (New York, New York): Colleagues from Christiana Health Center have reported that adult patients, who are hyperglycemic in the trauma bay, have a risk of mortality that is increased 8- to 10-fold. The risk exists for serum glucose concentrations as low as 135 mg/dL and likely lower. This mortality is related to an increased risk of nosocomial infection and, had it been quantified, to the organ dysfunction syndrome that remained a major cause of late death after injury.

Given the known, profound adverse effects of hyperglycemia and immune function, host immunosuppressant effects leading to organ dysfunction is a very plausible hypothesis. In other related literature, it is established that failure to control hyperglycemia during cardiac surgery triples the risk of sternal wound infection, and that a serum glucose concentration greater than 200 mg/dL on postoperative day 1, even transiently, quadruples the risk of surgical site infection after major noncardiac surgery. Recently, a randomized multicenter trial for critically ill surgical patients demonstrated that mortality was halved when serum glucose was kept under 110 mg/dL with exogenous insulin.

This study suggests that trauma patients are no different. The authors are to be commended. The audience is advised that a prospective study from Baltimore, to be presented later today as a poster, is corroborative, albeit with a smaller adverse impact on mortality when serum glucose greater than 200 mg/dL was used as a cutpoint.

We do need a prospective study of insulin administration in the trauma bay. I would like the authors to comment on how that study might be designed.

From your logistic regression model, can you determine how tightly glucose must be controlled? How soon? For how long? Should insulin be given as a continuous infusion with or without dextrose in a clamp technique or in response to frequent periodic determinations?

How will the imperative of early enteral feeding impact design and conduct? Should total parenteral nutrition be abandoned as an immunologic and hepatic toxin?

Is there an alternative hypothesis, the testing of which should be the focus of this study? What if the problem fundamentally is not hyperglycemia but catabolism?

Administration of insulin as an anabolic hormone may be the crucial aspect of this line of investigation. If catabolism can be mitigated or reversed sooner, outcomes might indeed be better. Is that the problem at its essence?

Perhaps beta-blockade will prove beneficial for more than wound healing in burned children. With apologies to those who believe that outcomes are worse in male patients with sepsis—and I am not among them—should we be giving anabolic hormones sooner to more patients?

Fundamentally, are we changing the resuscitation paradigm? Rather than debating hemodynamic oxygen transport and acid based endpoints, are we reinvigorating debate about resuscitating the immune system? Again, I commend the authors and thank the Association for the privilege of the floor.

Dr. Grant Bochicchio (Baltimore, Maryland): I also wanted to thank you for your excellent presentation, in which your data were really complementary to our data, as Dr. Barie mentioned, which is a prospective study that we will be presenting later today. I have a few following questions that I was unable to answer from your study.

First, did you stratify outcome by procedures (i.e., by abdominal surgeries, orthopedic surgeries, neurosurgical procedures) regarding the two groups. Second, did you compare (even though the ISSs are the same) or stratify them by Abbreviated Injury Scale score? That is, did you compare the traumatic brain-injured patients in both groups so that you were sure that you are comparing “apples to apples?”

Finally, did you separate out community-acquired infections, that is, infections that were already present or incubating in the field and diagnosed within 48 to 72 hours? This would definitely skew your data in terms of infection rate, prognosis, and nosocomial infections and should be included in your data.

Dr. Basil A. Pruitt, Jr. (San Antonio, Texas): Did you perform a multiple regression analysis of all the variables you recorded and find that hyperglycemia significantly reduced the residual variance beyond that accounted for by all those other variables? Second, because hyperglycemia is a natural response to injury and is induced by the increased secretion of catecholamines and glucocorticoids, did you measure circulating levels of those hormones? Finally, did you have any information about when the patients last ate?

Dr. Saikrishna Yendamuri (closing): The first question was how should insulin be administered if the decision is made to perform a randomized trial. Before we do that, though, what we should do is determine the appropriate cutoff. The cutoff used for this study doesn't have a strict biologic basis.

One way of determining that is to analyze glucose as a continuous variable using a receiver operating characteristic analysis with a larger population and trying to find that discriminating level. Next, that has to be validated first before we go ahead and perform a trial.

As for the catabolism being the primary problem, there's enough evidence to state that hyperglycemia itself leads to catabolism. There's an excellent study by Jeeranandam et al. that proved that hyperglycemia, independent of insulin status, by itself can lead to a catabolic effect.

Although insulin administration might be part of the solution, it may not be all of it. It would have to be tied to glucose control.

I agree with Dr. Barie and I think this has an impact on the kind of resuscitation that we perform and the kind of fluids that we use for resuscitation. I would just like to reiterate the fact that we should avoid hyperglycemia with any therapeutic intervention that we may use in the process.

Dr. Bochicchio, one of the questions that was raised was whether we had information on possible infections before the patient arrived in the trauma bay. Hyperglycemia may have an influence not just on incidence of infections but also on the clinical progress and impact of those infections. Therefore, even if you have data that showed that somebody had prior infections in the field, it may not invalidate the effect of hyperglycemia on that process. In this study, we did not have these data.

As for Abbreviated Injury Scale score, we really did not look at that, but it's something that we could look at. Also, we did not look at the kind of procedures and the specific kind of trauma in this population. Patients with head trauma might be a confounding factor that we did not look at.

Dr. Pruitt, no, we did not have an idea of when the patient last ate. We did not measure levels of cortisol or epinephrine. The addition of glucose to the logistic regression model did reduce the residual variance. However, if we had a model that had ISS, RTS, and age and added glucose to that, there was a minor improvement.

Substitution of ISS with glucose in this model also led to a reduction of variance in the regression. The reason we performed that analysis was that ISS is not a figure that we often have in the trauma bay. Thus, if you have a model with just RTS, age, and glucose, we think it predicts mortality as well as a model that has age, ISS, and RTS.


Prognosis; Trauma; Glucose; Sepsis; Mortality; Hyperglycemia; Outcome.

© 2003 Lippincott Williams & Wilkins, Inc.