a b s t r a c t

Emergency physicians face the challenge of rapidly identifying high-risk trauma patients. Lactate (LAC) is widely used as a surrogate of tissue hypoperfusion. However, clinically important values for LAC as a predictor of mor- tality are not well defined. Objectives: 1. To assess the value of triage LAC in predicting mortality after trauma.

2. To compute interval likelihood ratios (LR) for LAC.

Methods: Retrospective chart review of trauma patients with a significant injury mechanism that warranted labs at an urban trauma center. Outcome: In-hospital mortality. Data are presented as median and quartiles or per- centages with 95% confidence intervals. Groups (lived vs. died) were compared with Man-Whitney-U or Fisher’s-exact test. Multivariate analysis was used to measure the association of the independent variables and mortality. The interval likelihood ratios were calculated for all LAC observed values.

Results: 10,575 patients; median age: 38 [25-57]; 69% male; 76% blunt; 1.1% [n = 119] mortality. LAC was statis- tically different between groups in univariate (2.3 [1.6,3.0] vs 2.8 [1.6,4.8], p = 0.008) and multivariate analyses (odds ratio: 1.14 [1.08-1.21], p = 0.0001). Interval ratios for LR- ranged from 0.6-1.0. Increasing LAC increased LR+. However, LR+ for LAC reached 5 with LAC N 9 mmol/L and passed 10 (moderate and conclusive increase in disease probability, respectively) with LAC N 18 mmol/L.

Conclusions: In a cohort of trauma patients with a wide spectrum of characteristics triage LAC was statistically able to identify patients at high risk of mortality. However, Clinically meaningful contribution to decision- making occurred only at LAC N 9. LAC was not useful at excluding those with a low risk of mortality.

(C) 2017

1. Introduction

Trauma is the leading cause of death in the United States for ages 1-46 years and the third leading cause across all age groups [1]. A major proportion of acute deaths (<= 48 h) are attributed to exsanguina- tion with global organ failure [2]. Early detection and reversal of tissue hypoperfusion is crucial for improved outcome. Patients presenting to the emergency department (ED) after their injury may appear clinically stable despite significant ongoing occult hemorrhage. Optimal care of injured patients is dependent on early strategies to assess risk.

Uncontrolled hemorrhage results in failure of oxygen delivery, in- creasing oxygen debt, anaerobic metabolism and acid/base dysregula- tion [3]. serum lactate and Base deficit (BD), have long been used as surrogates for tissue hypoperfusion. They supplement physical exam and predictor injury severity and prognosis, more reliably than

* Corresponding author at: Department of Emergency Medicine, 450 Clarkson Avenue, Box: 1228, Brooklyn, NY 11203, USA.

E-mail address: [email protected] (S. Zehtabchi).

Abnormal vital signs [4,5]. Numerous studies have evaluated the associ- ation of LAC, BD, and LAC clearance with mortality. Most have concluded that elevated LAC, from either a venous or arterial sample, and failure to clear LAC, are highly predictive of in-hospital mortality [3,6,7,8,9,10]. However, review of the trauma literature highlights challenges in gen- eralizing existing data and limitations of available studies. Summation of data is difficult given patient variability. Existing studies are largely retrospective and registry-based. Registry sourcing imparts limitations on validity and variability of data, as inclusion criteria is significantly dif- ferent between institutions and there are no universal methods for cleaning or verification of trauma registry data. There is minimal infor- mation about hospital course and interventions included in trauma reg- istries [11]. Additionally, most prior studies have small sample sizes and are conducted in single trauma centers, potentially limiting their generalizability.

Another challenge is identifying the clinical application of specific

LAC values. If LAC is associated with mortality, what does a normal value mean for the patient? How elevated must the LAC be for the risk of death to be significant?

https://doi.org/10.1016/j.ajem.2017.10.015

0735-6757/(C) 2017

The objective of our study was twofold: 1. To assess the value of tri- age LAC in predicting mortality after trauma. 2. To determine if there is greater clinical applicability in a continuous model by computing inter- val likelihood ratios (LR) for LAC.

1. Methods
   1. Study design, population, and setting

This retrospective cohort study was carried out at a large, urban trauma center. The Institutional Review Board (IRB) approved the study. The hospital’s ED has an annual census of approximately 135,000 patient encounters, with over 1200 annual trauma admissions. All trauma patients (all ages) who presented to the ED during the study period (January 1st, 2011 to June 30th, 2016) with any injury mechanism of significance to warrant laboratory work up (including ve- nous LAC) were included in the study. Patients with renal or Liver failure and those with suspicion of severe infection/sepsis at time of triage

were excluded from the study.

Study protocol

We followed the recommendations proposed by Gilbert et al. to re- duce risk of bias in retrospective studies [12]. Electronic medical records (Quadramed, Quadramed Corporation, Reston, USA) were queried for all patients who had venous LAC measured upon admission to the ED. Trained research assistants reviewed electronic medical records using a pre-determined study protocol and excluded patients who presented with non-trauma related complaints. Patients were selected for the study based on a multi-tiered system of manually checking the follow- ing electronic documents: triage information, ED disposition diagnosis and hospital discharge diagnosis to confirm the traumatic nature of the ED visit. Crosschecks were performed on 10% of entries.

Information pertaining to patient demographics, the index trauma mechanism (blunt versus penetrating), vital signs, presence or absence of intracranial hemorrhage, venous LAC and BD, hospital admission and mortality were documented. If patients had additional LAC measure- ments beyond the initial, LAC clearance was also calculated.

Primary outcome

In-hospital mortality. The electronic medical record was queried for mortality data and the information was matched to the enrolled patients.

Method of measurement of LAC and outcomes

Venous LAC concentration (ABL 800 FLEX Blood gas analyzer, Radi- ometer Medical, Bronshoj, Denmark) was measured at time of initial evaluation. Triage BD was simultaneously measured with the same blood gas analyzer. LAC clearance was defined by the difference be- tween the first LAC measurement and a subsequent LAC level (at least 60 min apart, but not N 24 h from ED presentation).

Statistical analysis

Continuous variables are presented as median and quartiles and cat- egorical variables are presented as percentages with 95% confidence in- tervals. Study subjects were categorized into two groups (survived vs. died). Groups were compared with Man-Whitney-U test or Fisher’s exact test, when appropriate. Multiple logistic regression models were used to determine predictors significantly associated with mortality. We considered the following predictors: age, gender, mechanism of in- jury, intracranial hemorrhage, triage Blood pressure and heart rate, ve- nous LAC and BD. Pearson correlation coefficient was used to determine the association between triage LAC and BD.

The interval likelihood ratios for positive (LR+) and negative (LR-) tests were measured by first calculating the sensitivity and specificity

of every observed value of LAC in predicting mortality. Subsequently, likelihood ratios for positive and negative tests for all observed LAC levels were calculated using the standard likelihood ratio formula (LR+ = sensitivity/1- specificity & LR- = 1- sensitivity/specificity).

We planned a subgroup analysis a priori to repeat the same multi-

variate analysis only for admitted patients to examine whether the pre- dictive value of LAC was stronger in patients who had more severe injuries that warranted admission.

As a rule of thumb, a study is required to have at least 10 positive outcomes per number of variables included in the multivariate analysis [13]. We expect to have 10 variables in this multivariate model (Age, gender, mechanism of injury, presence or absence of intracranial hem- orrhage, hospital admission, systolic blood pressure, diastolic blood pressure, heart rate, BD and LAC). Therefore, 100 deaths are required in order for the analysis to have adequate power. The lowest reported mortality rate among the reviewed literature is 3% (range: 3% to 24%) [1,2,3,4,6,8,10]. Therefore, we would need at least 3000 subjects for the study to achieve its objective with adequate power. P value b 0.05 was considered statistically significant. Statistical analyses were per- formed with SAS version 9.4 (SAS Institute Inc., Cary, NC, USA).

1. Results

A total of 11,800 patients presented to the ED with injury during the study period. After applying the Inclusion/exclusion criteria, 10,575 pa- tients were included in the final analysis (median age: 38 [25-57]; age range 1-105 years; 69% male; 76% blunt trauma). The in-hospital mor- tality rate was 1.1% [n = 119]. The baseline characteristics of the study cohort are listed in Table 1.

In univariate analysis (Table 2) age, BD, LAC and hospital admission were statistically different between the groups. In multivariate analysis (Table 3), older age, higher LAC, larger BD and hospital admission were associated with increased mortality.

The results were not changed when only admitted patients were in- cluded in the multivariate analysis (Table 4).

We calculated the LR (for positive and negative tests) for all the ob- served LAC levels (157 observations, LAC ranging from 0.4 to 30 mmol/L). The interval LR- for the observed LAC values ranged from

0.6 to 1. This range represents test results that do not substantially de- crease the post-test probability and therefore are not helpful in clinical decision making [14]. The interval LR+ and the corresponding LAC levels are presented in Table 5. Increasing LAC increased LR+. However, LR+ for LAC reached 5 (moderate increase in disease probability) with LAC N 9 mmol/L and LR+ passed 10 (conclusive increase in disease probability) with LAC N 18 mmol/L (Table 5) [14]. The mortality rates

Table 1

Baseline characteristics of the study population

Variable	Median and quartiles
Age	38 (25, 57)
Systolic blood pressure	136 (122, 151)
Diastolic blood pressure	83 (72, 94)
Heart rate	87 (74, 99)
Lactate	2.3 (1.6, 3.3)
Base deficit	1.3 (-0.80, 3.0)

Variables n/N

% (95% confidence interval)

Gender (male) 7266/10575

69%(66-71%)

Mechanism (blunt) 7993/10575

76% (73-78%)

Intracranial hemorrhage 237/10575 2% (1-3%)

Overall admission rate 5298/10575 50% (47-53%)

In-hospital mortality 119/10575

1.1% (0.6-2%)

Table 2

Comparison of variables between survivors and non-survivors

Table 4

Multivariate regression for identifying variables that could predict mortality among ad-

Variablea	Survived	Died Pb		mitted patients only (n = 5212)
	N = 10,456	N = 119			Variable	Odds ratio (95% confidence interval)	p
Age	38 (25-57)	47 (28, 65)	0.0005		Age	1.014 (1.00-1.04)	0.0339
Gender (male)	7191/10456	75/119	0.196		Male gender	1.62 (0.91-2.89)	0.1002
	69% (66-71%)	63% (54-71%)			Blunt mechanism of injury	1.24 (0.62-2.48)	0.549
Mechanism (blunt)	7897/10456	96/119	0.240		Intracranial hemorrhage	0.55 (0.13-2.35)	0.4166
	75% (73-78%)	81% (73-87%)			Lactate (mmol/L)	1.14 (1.08-1.21)	0.0001
Intracranial hemorrhage	235/10456	2/101	1.00		Base deficit (mEq/L)	0.89 (0.86-0.93)	0.0001
	2.3% (1.5-3%)	(2%, 0-7%)			Systolic blood pressure	0.99 (0.98-1.006)	0.2978
Overall admission rate	5212/10442	86/119	0.0001		Diastolic blood pressure	0.99 (0.97-1.01)	0.4076
	50% (47-53%)	72% (64-80%)			Heart rate	1.00 (0.99-1.01)	0.9026
Lactate (mmol/L)	2.3 (1.6-3.3)	2.8 (1.6, 4.8)	0.0089
Base deficit (mEq/L)c	1.3 (-0.8, 3.0)	0.7 (-3.7-3.2)	0.0228
Lactate clearanced	1.6 (0.2, 3.8)	0.7 (0.0, 3.7)	0.4848
Systolic blood pressure	136 (122, 152)	136 (120, 155)	0.6985	and heart rate are crude measures of perfusion and may be normal, de-
Diastolic blood pressure Heart rate	83 (73, 94) 87 (74, 99)	81 (70, 95) 89 (78, 102)	0.6223 0.1236	spite substantial oxygen debt. Vital signs have been shown to be unreli- able in the risk stratification of trauma patients [16]. The lack of

^a Continuous variables are presented as medians and 25%, 75% quartiles. Categorical variables are presented as percentages with 95% confidence intervals.

^b Continuous variables are compared using Man-Whitney-U test and categorical vari-

ables are compared using Fisher’s Exact test.

^c n = 7568.

^d n = 905.

associated with LAC cut-offs generated by interval likelihood ratios as well as the traditional “critical LAC level” is presented in Table 6. As pre- sented in the table, as LAC level crosses 9 mmol/L, the mortality rate more than triples (from 1.3% to 4.2%) and as it reaches the level of

18.5 mmol/L it further doubles (from 4.2% to 9.7%).

We also calculated the correlation of triage LAC and BD. This analysis revealed a negative correlation of modest strength between the two tests (Pearson coefficient: -0.43, p = 0.001).

1. Discussion

Our data demonstrates that increased triage LAC is “statistically” as- sociated with higher risk of mortality in trauma patients. However, de- spite statistical significance, the clinical utility of an initial LAC is less clear. Given that only a moderate increase in pre-test probability oc- curred with LAC b 9, we must consider the degree to which a single LAC upon initial Patient evaluation is clinically helpful to emergency physicians and surgeons making Treatment decisions based on pre- sumed risk.

The concept of a “Golden Hour” for optimizing trauma resuscitation and improving patient outcome is well accepted [15]. Early identifica- tion of high risk trauma patients enables physicians to target those that would benefit from aggressive resuscitation and allocation of criti- cal care resources. Initial assessment of patients is dependent on clinical examination and the limited objective information we are able to gather in the immediate stages of care. We first rely on vital signs, however, they are often misleading in the early stages of shock. Blood pressure

reliability of initial vital signs to predict trauma outcome was also dem- onstrated in our data. Heart rate and blood pressure were not found to be significantly different between groups of patients who survived or died.

injury severity scores (ISS) are commonly used in the trauma litera- ture as a predictor of mortality. However, this “anatomical” measure of injury severity is calculated in a delayed fashion once all injuries are identified. ISS is unavailable at the time of a patient’s initial ED presen- tation and cannot be used to guide resuscitation [17]. Given the inade- quacies of vital signs and unavailability of ISS early on, readily available biomarkers of occult hypoperfusion gain particular importance.

Biomarkers of anaerobic metabolism have long been used to evalu- ate acute blood loss in trauma patients. For many years, BD was the bio- marker used to guide trauma resuscitations. BD was readily available to most practitioners as part of an arterial blood gas measurement, and served as a surrogate marker for the accumulation of lactic acid in hypo- perfusion states. N 20 years ago, BD was first shown to correlate with volume requirements and outcome in trauma patients [18,19]. LAC, a di- rect byproduct of anaerobic metabolism, initially was not widely avail- able as a point of care test. However, in recent years, LAC has become readily available with quick turnaround times in the Trauma bay. Since LAC is a more precise measure of anaerobic metabolism than its surrogate, BD, we focused our study on LAC’s ability to predict mortality. We sought to determine a specific value for LAC that would guide deci- sions about clinical management.

Several prior studies deemed LAC predictive of mortality in trauma. However, most of those studies were limited by small sample size and/ or number of outcomes [9,20,21]. We sought to confirm an association between LAC and mortality in our large, diverse group of trauma pa- tients, which included all injury types and degrees of severity.

Our data showed age, LAC and BD were independent predictors of mortality. Our results, however, failed to demonstrate the clinical value of such findings. Despite finding LAC to be statistically different

Table 3

Multivariate regression for identifying variables that could predict mortality among all pa-

Table 5

Interval likelihood ratios for LAC in the study cohort

tients (n = 10,575)

Variable	Odds ratio (95% confidence inte	rval) p		0.4-3.4	1	0.9% (78/8142)
Age	1.012 (1.06-1.18)	0.0006		3.5-5.2	2	0.9% (14/1487)
Male gender	1.12 (0.70-1.78)	0.640		5.3-7	3	1.8% (8/445)
Blunt mechanism of injury	1.26 (0.74-2.16)	0.3904		7.1-9	4	2.2% (5/227)
Intracranial hemorrhage	0.52 (0.12-2.21)	0.52		9.1-12	5	4.2% (5/119)
Lactate (mmol/L)	1.11 (1.06-1.18)	0.0001		12.1-13.9	6	3.6% (2/55)
Base deficit (mEq/L)	0.93 (0.86-0.99)	0.038		14-15.5	7	9% (2/22)
Hospital admission	0.62 (0.39-0.98)	0.044		15.6-16.5	8	0% (0/17)
Systolic blood pressure	0.99 (0.98-1.004)	0.2181		16.6-18.5	9	5% (1/20)
Diastolic blood pressure	0.99 (0.98-1.01)	0.6148		N 18.5	N 10	10% (4/41)
Heart rate 1.006 (0.99-1.02) 0.3464 a The interval LR- for the observed lactate values ranged from 0.6 to 1.

Lactate range mmol/L

LR+ ^a Mortality rate

% (n/N)

Table 6

Mortality rates associated with various lactate ranges

Lactate range	Mortality	Mortality rate
Mmol/L	(n/N)	(%)
0.0-3.9	86/8833	0.97%
4.0-8.9	19/1448	1.3%
9.0-17.9	10/236	4.2%
N 18.5	4/41	9.7%

between survivors and non-survivors, the medians of each group were numerically very similar, thus raising the question of how useful and clinically meaningful LAC would be for clinicians making management decisions. This lack of clinical relevance is reflected in studies by Parsikia and Odom, in which the OR of LAC for mortality was 1.01 and 1.2 respec- tively [20,22]. Duane failed to find even a statistically significant differ- ence when comparing initial LAC in their cohort [23]. In a study by Pal, a more notable difference in LAC between survivors and non-survivors was observed (LAC 3.0 vs. 5.2), but investigators reported low discrim- inatory power of the biomarker due to overlap. They attributed their findings to low morality in their study population [21].

The value of triage LAC in predicting survival has been further dem-

onstrated by studies that used Receiver Operating Characteristic curves. These studies showed that LAC only had a modest Discriminatory power; area under the curve (AUC) of 0.73, 0.73, and 0.63 [10,21,22]. These findings may be explained by the multifactorial etiology of elevat- ed LAC. An elevated LAC in trauma can reflect increased production or decreased clearance, but is likely attributed to both. Considering our co- hort of patients, hypoperfusion secondary to hemorrhage is most worri- some to the emergency physician, but one must also consider the presence of a hypermetabolic state and Liver dysfunction [24]. Of note, Dezman found the AUC of initial LAC predicting 24-h mortality to be higher than analysis including serial LAC levels [10]. This suggests great- er utility of LAC when measured upon immediate presentation of the patient. Initial LAC may be relatively unaffected when compared with later LAC levels and LAC clearance due to progressive development of organ failure or other etiologies which could alter LAC. These findings indicate that LAC might not be a simple, straight forward indicator of mortality. LAC’s clinical utility rather might be influenced by a variety of factors. While exploring these factors is beyond the scope of our study, we evaluated a different approach to using LAC that might im- prove its clinical applicability in risk stratifying trauma patients.

In our analyses, we found a more convincing relationship between

LAC and mortality with a continuous approach to the study data (rather than a dichotomous normal vs abnormal approach). In our cohort, in- creasing LAC increased LR+. However, clinical significance of a test re- lies on its ability to increase the pre-test probability of the disease to an actionable (rule in or rule out the disease) post-test probability [25]. In the case of LAC, this impact was appreciated with LAC N 9 mmol/L. At this level, LR+ was 5, reflecting a moderate increase in disease post- test probability [14]. Depending on the pre-test probability of the dis- ease (determined by mechanism of injury and clinical presentation), this LR+ could generate a post-test probability that triggers aggressive resuscitation. LR+ passed 10 (conclusive increase in disease probabili- ty) with LAC N 18 mmol/L. An LR+ N 10 is associated with a significant increase in disease probability, even when the pre-test probability is not that high; signifying the value of this level of LAC in making clinical decision making [14]. In parallel to our findings, Dezman et al. founda 5- fold greater risk of death after 24 h with LAC >=7 mmol/L, using odds ra- tios [10].

These higher levels of LAC required for risk stratification vary from those we tend to associate with hypoperfusion (LAC N 4, commonly called critical lactate level) [20]. This may suggest that a single, early LAC may not adequately reflect a building oxygen debt that occurs with acute hemorrhage, unless it is above 9 mmol/L, or even 18 mmol/L.

Despite our significant findings with increasing LAC, we did not find a negative LAC to be useful, with a range of 0.6-1.0 for LR-. A negative test only decreases the disease post-test probability significantly if the LR- value is close to 0 (or b 0.1) [14,25]. Therefore, a single normal LAC should not be used as an indicator of absence of severe hemorrhage. Our finding is in contrast to the study by Dezman et al., which reported normal initial LAC to be a “reassuring” indicator of a low risk of death, with a negative predictive value of 99.72% for mortality at 24 h with a cutoff of 3.0 mmol/L. We calculated the positive and negative likelihood ratios of a LAC >= 3 mmol/L based on the raw data provided in the pub- lished article. Based on the data presented in Dezman’s article, this se- lected cut-off for LAC was associated with an LR+ of 2.2 and an LR- of

0.34 for predicting 24-h mortality; and LR+ of 1.9 and LR- of 0.5 for predicting in-hospital mortality [10]. As explained earlier in this section, LR+ and LR- at these levels have limited impact on probability of the disease and therefore are not useful in clinical decision making.

LAC clearance has been proposed as a better predictor of outcome than a single admission LAC. Several studies have demonstrated that a failure to clear LAC was strongly associated with mortality in varioUS settings [9,20,26,27,28]. In contrast to these earlier studies, our data did not support LAC clearance as a predictor of outcome. We found no significant difference in LAC clearance between survivors and non- survivors. However, our data was retrospective and intervals between initial and subsequent LAC measurements were not consistent. In addi- tion, a large proportion of our patients did not have a repeat LAC level.

We found a less than optimal correlation between LAC and BD. BD is a calculated value that reflects the amount of additional base that would have to be added to the blood to neutralize the PH. Therefore, BD repre- sents the overall acid-base status, while LAC is an indicator of lactic ac- idosis due to tissue hypoperfusion. The cause of an abnormal BD acidosis could be any unmeasured anions: ketones, alcohol, renal failure, and even hyperchloremia of saline resuscitation, as well as elevated LAC. Therefore, it is not surprising that LAC and BD are not in 100% concor- dance [29,30,31].

1. Limitations

Our study had several limitations. Data was retrospective, thus lim- iting our ability to control for confounding variables. However, we employed the steps recommended by Gilbert et al. to improve the valid- ity of our results [12]. We did not have a complete record of patient co- morbidities. Certainly, LAC might have been altered in patients with sig- nificant baseline liver or renal disease. Our database was from a single hospital center in an urban location. This may raise questions about gen- eralizability to other settings, such as a rural location, where transport time to the hospital may be significantly delayed.

1. Conclusion

In a large cohort of trauma patients with a wide spectrum of charac- teristics (age, disease severity, injury mechanism), triage LAC had poor discriminatory power in identifying high risk patients. Although a statis- tically significant difference in LAC existed between survivors and non- survivors, clinical application was limited. A value was only meaningful to decision-making (moderate increase in pre-test probability) with LAC N 9 mmol/L.

Funding source

None.

Conflict of interest

None.

References

1. Injury Prevention & Control [Internet]. Centers for disease control and prevention; 2017Available from: https://www.cdc.gov/injury/wisqars/.
2. Sauaia A, Moore FA, Moore EE, et al. Epidemiology of Trauma deaths. J Trauma Inj In- fect Crit Care 1995;38(2):185-93.
3. Regnier M-A, Raux M, Manach YL, et al. Prognostic significance of blood lactate and lactate clearance in trauma patients. Anesthesiology 2012;117(6):1276-88.
4. Callaway DW, Shapiro NI, Donnino MW, Baker C, Rosen CL. Serum lactate and base deficit as predictors of mortality in normotensive elderly blunt trauma patients. J Trauma Inj Infect Crit Care 2009;66(4):1040-4.
5. Heffernan DS, Thakkar RK, Monaghan SF, et al. Normal presenting vital signs are un- reliable in geriatric blunt Trauma victims. J Trauma Inj Infect Crit Care 2010;69(4): 813-20.
6. Sammour T, Kahokehr A, Caldwell S, Hill AG. Venous glucose and arterial lactate as biochemical predictors of mortality in clinically severely injured trauma patients-a comparison with ISS and TRISS. Injury 2009;40(1):104-8.
7. Lavery RF, Livingston DH, Tortella BJ, Sambol JT, Slomovitz BM, Siegel JH. The utility of venous lactate to triage injured patients in the trauma center. J Am Coll Surg 2000; 190(6):656-64.
8. Schmelzer TM, Perron AD, Thomason MH, Sing RFA. Comparison of central venous and arterial base deficit as a predictor of survival in acute trauma. Am J Emerg Med 2008;26(2):119-23.
9. Manikis P, Jankowski S, Zhang H, Kahn RJ, Vincent J-L. Correlation of serial blood lac- tate levels to organ failure and mortality after trauma. Am J Emerg Med 1995;13(6): 619-22.
10. Dezman Z, Comer A, Smith G, Narayan M, Hirshon J. 266 elevated initial lactate and failure to clear lactate predicts 24-hour mortality in trauma patients. Ann Emerg Med 2014;64(4).
11. Zehtabchi S, Nishijima DK, McKay MP, Mann NC. Trauma registries: history, logistics, limitations, and contributions to emergency medicine research. Acad Emerg Med 2011 Jun;18(6):637-43.
12. Gilbert EH, Lowenstein SR, Koziol-McLain J, Barta DC, Steiner J. Chart reviews in emergency medicine research: where are the methods? Ann Emerg Med 1996; 27(3):305-8.
13. Peduzzi P, Concato J, Kemper E, Holford TR, Feinstein AR. A simulation study of the number of events per variable in logistic regression analysis. J Clin Epidemiol 1996;49(12):1373-9.
14. Grimes DA, Schulz KF. Refining clinical diagnosis with likelihood ratios. Lancet 2005;

365(9469):1500-5.

1. Cowley RA. A total emergency medical system for the state of Maryland. Md State Med J 1975;45:37-45.
2. Paladino L, Sinert R, Wallace D, Anderson T, Yadav K, Zehtabchi S. The utility of base deficit and arterial lactate in differentiating major from Minor injury in trauma pa- tients with normal vital signs. Resuscitation 2008;77(3):363-8.
3. Baker SP, O’neill B, Haddon W, Long WB. The injury severity score. J Trauma Inj Infect

Crit Care 1974;14(3):187-96.

1. Davis JW, Shackford SR, Mackersie RC, Hoyt DB. Base deficit as a guide to volume re- suscitation. J Trauma Inj Infect Crit Care 1988;28(10):1464-7.
2. Rutherford EJ, Morris JA, Reed GW, Hall KS. Base deficit stratifies mortality and de- termines therapy. J Trauma Inj Infect Crit Care 1992;33(3):417-23.
3. Odom SR, Howell MD, Silva GS, et al. Lactate clearance as a predictor of mortality in trauma patients. J Trauma Acute Care Surg 2013;74(4):999-1004.
4. Pal JD, Victorino GP, Twomey P, Liu TH, Harken AH, Betts JM. Admission serum lac- tate levels do not predict mortality in the acutely injured patient. J Trauma Inj Infect Crit Care 2004;57(2):452.
5. Parsikia A, Bones K, Kaplan M, et al. The predictive value of initial serum lactate in trauma patients. Shock 2014;42(3):199-204.
6. Duane TCAM, Ivatury RR, Dechert T, et al. Blood glucose levels at 24 hours after trau- ma fails to predict outcomes. J Trauma Inj Infect Crit Care 2008;64(5):1184-7.
7. Andersen LW, Mackenhauer J, Roberts JC, Berg KM, Cocchi MN, Donnino MW. Etiol- ogy and therapeutic approach to elevated lactate levels. Mayo Clin Proc 2013; 88(10):1127-40.
8. Zehtabchi S, Kline JA. The art and science of probabilistic decision-making in emer- gency medicine. Acad Emerg Med 2010 May;17(5):521-3.
9. Abramson D, Scalea TM, Hitchcock R, Trooskin SZ, Henry SM, Greenspan J. Lactate clearance and survival following injury. J Trauma Inj Infect Crit Care 1993;35(4): 584-9.
10. Regnier M-A, Raux M, Manach YL, et al. Prognostic significance of blood lactate and lactate clearance in trauma patients. Anesthesiology 2012;117(6):1276-88.
11. Mcnelis J, Marini CP, Jurkiewicz A, et al. Prolonged lactate clearance is associated with increased mortality in the surgical intensive care unit. Am J Surg 2001; 182(5):481-5.
12. Martin MJ, Fitzsullivan E, Salim A, Brown CV, Demetriades D, Long W. Discordance between lactate and base deficit in the surgical intensive care unit: which one do you trust? Am J Surg 2006;191(5):625-30.
13. Wilson RF, Sibbald WJ. Approach to acid-base problems in the critically III and in- jured. J Am Coll Emerg Physic 1976;5(7):515-22.
14. Aduen J, Bernstein WK, Miller J, et al. Relationship between blood lactate concentra- tions and ionized calcium, glucose, and acid-base status in critically ill and noncritically ill patients. Crit Care Med 1995;23(2):246-52.