Out-of-Hospital Endotracheal Intubation and Outcome After Traumatic ...

EMERGENCY MEDICAL SERVICES/ORIGINAL RESEARCH

Out-of-Hospital Endotracheal Intubation and Outcome After Traumatic Brain Injury See editorial, p. 451. Henry E. Wang, MD, MPH Andrew B. Peitzman, MD Laura D. Cassidy, PhD P. David Adelson, MD Donald M. Yealy, MD From the Department of Emergency Medicine (Wang, Yealy), Department of Surgery (Peitzman), and Department of Neurosurgery (Adelson), University of Pittsburgh School of Medicine; and the Department of Biostatistics, University of Pittsburgh (Cassidy), Pittsburgh, PA.

Study objective: Previous studies disagree about the effect of out-of-hospital endotracheal intubation on traumatic brain injury. This study compares the effects of out-of-hospital endotracheal intubation versus emergency department (ED) endotracheal intubation on mortality and neurologic and functional outcome after severe traumatic brain injury. Methods: From the 2000 to 2002 Pennsylvania Trauma Outcome Study (a registry of all patients treated at trauma centers in the Commonwealth of Pennsylvania), adult patients with head/neck Abbreviated Injury Scale score of 3 or greater and undergoing outof-hospital endotracheal intubation or ED endotracheal intubation were included. Transferred patients were excluded. The primary outcome was death (on hospital discharge). The secondary outcomes were neurologic (good versus poor, inferred from discharge to home versus long-term care facility) and functional outcome (determined from a Functional Impairment Score). The key exposure was endotracheal intubation (out-of-hospital endotracheal intubation versus ED endotracheal intubation). Using multivariate logistic regression, odds estimates for out-of-hospital endotracheal intubation were adjusted using age, sex, head/neck Abbreviated Injury Scale score, Injury Severity Score, mechanism of injury (penetrating versus blunt), admission systolic blood pressure, mode of transport (ground only versus helicopter or helicopter + ground), and the use of out-of-hospital neuromuscular blocking agents. A propensity score adjustment accounted for the potential effects of preexisting conditions, inhospital complications, and social factors (drug and alcohol use, race, and insurance coverage). Results: There were 4,098 patients with head/neck Abbreviated Injury Scale score of 3 or greater who received either out-of-hospital endotracheal intubation (n=1,797, 43.9%) or ED endotracheal intubation (n=2,301, 56.1%). Adjusted odds of death were higher for out-of-hospital endotracheal intubation than ED endotracheal intubation (odds ratio [OR] 3.99; 95% confidence interval [CI] 3.21 to 4.93). Out-of-hospital endotracheal intubation was associated with an increased adjusted odds of poor neurologic outcome (OR 1.61; 95% CI 1.15 to 2.26), moderate or severe functional impairment (Functional Impairment Score 6 to 15; OR 1.92; 95% CI 1.40 to 2.64), and severe functional impairment (Functional Impairment Score 11 to 15; OR 1.80; 95% CI 1.29 to 2.52).

0196-0644/$30.00 Copyright Ó 2004 by the American College of Emergency Physicians. doi:10.1016/ j.annemergmed.2004.04.008

NOVEMBER 2004

44:5

Conclusion: Out-of-hospital endotracheal intubation was associated with adverse outcomes after severe traumatic brain injury. The implications for current clinical care remain undefined. [Ann Emerg Med. 2004;44:439-450.]

ANNALS OF EMERGENCY MEDICINE

4 3 9

E N D O T R A C H E A L I N T U B A T I O N A N D T R A U M A T I C B R A I N I N J U R Y Wang et al

Editor’s Capsule Summary What is already known on this topic Although endotracheal intubation is clearly a valuable intervention in some cases of traumatic brain injury, the environment, experience, and skills of paramedics are very different than those of physicians within a hospital. Some previous studies have suggested that out-of-hospital intubation might be harmful, not helpful.

What question this study addressed To compare the mortality and neurologic outcome of patients with traumatic brain injury who were intubated out of hospital versus in the emergency department (ED), using a statewide trauma registry.

What this study adds to our knowledge Among 4,098 patients intubated, odds of death were about 4 times greater if they were intubated out of hospital, and various measures of neurologic functional impairment were also greater (odds ratio 1.5 to 1.9).

How this might change clinical practice This study demonstrated greater mortality and impairment if patients were intubated by paramedics out of hospital instead of physicians in hospital, but does not prove causation. However, it confirms the potential for worse outcomes and suggests the need for rigorous data-driven monitoring of performance (unusual in many emergency medical services systems), pending further research that pinpoints the cause of this negative result.

INTRODUCTION Background

Hypoxia is believed to be deleterious after traumatic brain injury.1-3 Current guidelines for acute traumatic brain injury therapy call for aggressive measures to prevent hypoxia, including the use of endotracheal intubation.4-6 Importance

Although endotracheal intubation is readily accomplished in the inhospital setting, multiple factors may complicate the performance of this complex procedure in the out-of-hospital setting, especially in traumatic brain injury patients.7-9 Recent efforts have introduced controversial advanced airway management techniques to the setting for this patient subset, including the use of neuromuscular blockade–assisted endotracheal intubation.10 However, results from previous studies have generated conflicting data about the potential benefit of out-of-hospital endotracheal intubation in the therapy of traumatic brain injury, with some studies suggesting positive effects and others demonstrating adverse associations.11-14

4 4 0

Goals of This Investigation

The purpose of this study was to compare the effects of out-of-hospital endotracheal intubation versus emergency department (ED) endotracheal intubation on mortality and neurologic and functional outcome after traumatic brain injury. MATERIALS AND METHODS Study Design, Setting, and Data Collection and Processing

This study was approved by the University of Pittsburgh institutional review board. We conducted a retrospective analysis using data from the Pennsylvania Trauma Outcome Study, a statewide registry of all patients presenting to accredited trauma centers in the Commonwealth of Pennsylvania. The registry includes patients presenting to a trauma center with a primary International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) injury code of 800 to 995. All trauma deaths are included in the registry, as well as patients referred from other medical facilities. Patients who are discharged from the trauma center after initial presentation or who are admitted to a regular hospital floor for less than 48 hours are excluded from the registry. The registry excludes patients sustaining drowning, poisoning, isolated asphyxiation, and isolated hip fractures (ICD-9-CM 820.0). Pennsylvania Trauma Outcome Study’s comprehensive clinical information encompasses the following broad domains: patient demographic characteristics, out-of-hospital care, process of acute care, clinical data, outcome data, final anatomic diagnoses, procedure codes, and payer class. The Pennsylvania Trauma Outcome Study is operated and supervised by the Pennsylvania Trauma Systems Foundation, an independent nonprofit agency that oversees trauma care in Pennsylvania. Registrars at each trauma center are trained to abstract data from the medical records of each eligible trauma patient. Data are recorded using a customized trauma registry software package (Collector Trauma Registry; Digital Innovation, Inc., Forest Hill, MD) and transmitted electronically on a weekly basis for centralized merging into a master data set. Quality assurance procedures are performed to ensure the accuracy and integrity of reported data; for example, extensive computer and manual verification, reabstraction of a 10% random selection of medical records, and monthly reviews of selected data entries from each site. Customized diagnostic software (Tricode; Digital Innovation, Inc.) is used to verify the consistency of all computer entries. Only data passing quality reviews are


44:5

NOVEMBER 2004


merged into the master data set. These procedures are supplemented by regular site visits and semiannual statewide trauma registrar continuing education sessions. Pennsylvania Trauma Outcome Study data have been used extensively for many previous studies.15-20 Selection of Participants

We included adult trauma patients (age 18 years) treated from January 1, 2000, to December 31, 2002. We included only patients sustaining a severe traumatic brain injury as denoted by a head/neck Abbreviated Injury Scale score of 3 or greater; by the Abbreviated Injury Scale system, this score denotes all patients sustaining a head injury considered ‘‘serious,’’ ‘‘severe,’’ ‘‘critical,’’ or ‘‘unsurvivable.’’21 We excluded patients who were transferred from another hospital facility and those not treated by advanced life support rescuers. (These patients would not have had the opportunity to be exposed to out-of-hospital endotracheal intubation.) We also excluded patients who were not intubated in either the out-of-hospital or ED phases (ie, patients intubated in the operating room, ICU, or regular hospital floor or not intubated at all).

ED course were considered to have received ED endotracheal intubation. In this analysis, we included only patients who received either out-of-hospital endotracheal intubation or ED endotracheal intubation. The intention of this study was to evaluate only patients whose clinical airway presentation was severe enough to warrant endotracheal intubation during the acute out-of-hospital or ED resuscitation phases. We did not use inclusion criteria according to ‘‘need’’ for endotracheal intubation because there are no validated definitions, scales, or prediction rules for this characteristic. The Glasgow Coma Scale was not used as an inclusion criterion because it is a poor predictor of the need for endotracheal intubation and was frequently missing (approximately 50%).7,25 Methods of Measurement

Covariates evaluated were applied as reported in the trauma registry. We converted continuous variables (eg, admission systolic pressure) to ordinal scales. Cases with missing measurements were excluded during multivariate model development.

Interventions

Outcome Measures

The primary intervention (exposure) of interest was the binary variable endotracheal intubation (1=out-ofhospital endotracheal intubation, 0=ED endotracheal intubation). Out-of-hospital endotracheal intubation was denoted by the presence of an endotracheal tube or other definitive airway (eg, Combitube [esophageal-tracheal twin-lumen airway device; Kendall-Sheridan Catheter Corporation, Argyle, NY], cricothyroidotomy) in the out-of-hospital phase or on arrival in the ED. Secondary airway methods performed by out-of-hospital emergency personnel (eg, Combitube, cricothyroidotomy, tracheotomy) were coded in the registry as forms of tracheal intubation; this approach was reasonable because these events are relatively rare and are performed with the same intended goal of providing an invasive artificial airway.22 ED endotracheal intubation was identified by ICD9-CM codes 96.04, 96.05 (tracheal intubation), or 31.1 (temporary tracheotomy) performed in the ED. If a patient was coded as receiving out-of-hospital endotracheal intubation and ED endotracheal intubation, the patient was considered to have received ED endotracheal intubation. This assumption is reasonable because endotracheal tubes are known to be inadvertently misplaced and to require re-placement in the ED.23,24 Patients who received endotracheal intubation at any time during the

The primary outcome was binary variable death (on hospital discharge; 1=dead, 0=alive). The secondary outcomes were neurologic and functional outcome, determined for survivors only and classified using 2 schemes. We inferred neurologic outcome using the binary variable discharge destination. We considered discharge to home, rehabilitation facility, psychiatric facility, or police custody, or signing out of the hospital ‘‘against medical advice’’ to connote ‘‘good neurologic function’’ and discharge to a nursing home, hospital, trauma center, or other long-term care facility to connote ‘‘poor neurologic outcome.’’ Although this classification scheme contains limitations, we used this system exactly as applied by previous efforts to facilitate comparison with these studies.11,14 To better assess outcome, we also examined functional abilities at discharge, as described using functional status on discharge data reported by the Pennsylvania Trauma Outcome Study registry. Functional status on discharge is reported for discharged adult survivors only and include ordinal (1 to 4) ratings of 5 functional domains: feeding, locomotion, expression, transfer mobility, and social interaction.26 (The specific definitions for each functional status on discharge subscore are listed in Appendix E1 [available online at http://www.mosby.com/AnnEmergMed].) Within each functional domain, a rating of 1 denotes complete dependence, and a rating of 4 denotes

NOVEMBER 2004

44:5


4 4 1


independence. We summed the scores for the 5 functional status on discharge domains and subtracted the total from 20 to create a ‘‘functional impairment score,’’ where functional impairment score=0 denotes no functional impairment, and functional impairment score=15 denotes maximal impairment. We considered functional impairment score ordinal ranges of 0 to 5, 6 to 10, and 11 to 15 to connote ‘‘minimal,’’ ‘‘moderate,’’ and ‘‘severe’’ impairment, respectively. In this scheme, ‘‘severe’’ impairment connotes patients with vegetative or only minimal functioning across all functional domains. For the formal analysis, we dichotomized functional impairment score by using 2 cutoffs: functional impairment score of 6 or greater and functional impairment score of 11 or greater, which resulted in 2 outcome measures: (1) moderate or severe functional impairment (functional impairment score 6 to 15), and (2) severe functional impairment (functional impairment score 11 to 15). Although the functional impairment score has not been formally validated, we believed that its use in this manner was reasonable because its components resemble those of the Disability Rating Scale, a validated scale that has been used for rating functional outcome after traumatic brain injury.27,28 Formally validated outcome scales such as the Glasgow Outcome Scale and the Disability Rating Scale were not used because they were not available in the data set or derivable from registry data.27-30

Primary Data Analysis

We analyzed the hypothesized relationships using multivariate logistic regression. The conceptual model for each relationship was as follows: [Outcome]=fn {[endotracheal intubation (out-of-hospital endotracheal intubation vs ED endotracheal intubation)]1[covariate adjustment]1[propensity score]} We considered separate regression models to evaluate each of the 4 hypothesized relationships: d Association of death with out-of-hospital endotracheal intubation. d Association of poor neurologic outcome (discharge destination) with out-of-hospital endotracheal intubation. d Association of moderate or severe functional outcome (functional impairment score 6 to 15) with out-of-hospital endotracheal intubation. d Association of severe functional outcome (functional impairment score 11 to 15) with out-of-hospital endotracheal intubation.

4 4 2

To adjust for severity of clinical presentation, we used the covariates age, sex, head/neck Abbreviated Injury Scale score, Injury Severity Score, the presence of any other severe injury (any non–head/neck Abbreviated Injury Scale score 3), admission systolic blood pressure, mechanism of injury (penetrating versus blunt), mode of transport (ground-only versus air or air medical 1 ground), the use of out-of-hospital neuromuscular blocking agents, and a propensity score. These variables were chosen because they had plausible clinical connections to endotracheal intubation outcome and closely paralleled covariates used by previous studies of traumatic brain injury and endotracheal intubation.12,14 We included head/neck Abbreviated Injury Scale score and Injury Severity Score as covariates to account for variations in injury patterns and severity. Abbreviated Injury Scale score is a rating of the severity of injury to each of 6 body regions (head or neck, face, chest, abdomen, extremity, skin).21 As described previously, we included only patients sustaining serious, severe, critical, or unsurvivable head or neck injuries, as rated by this scheme (head/neck Abbreviated Injury Scale score 3 to 6). Injury Severity Score is based on the sum of the squares of the 3 highest Abbreviated Injury Scale scores and ranges from 0 to 75.31 Injury Severity Score was converted to ordinal ranges for this analysis (70=7). We selected Injury Severity Score and head/neck Abbreviated Injury Scale score over other methods of injury severity adjustment because Trauma–Injury Severity Score and A Severity Characterization of Trauma were missing for almost 20% of the Pennsylvania Trauma Outcome Study data, and other trauma scores were either unavailable or similarly limited by missing values.32,33 Furthermore, several previous studies have reported problems with the use of several standard trauma mortality measures for risk adjustment.33-36 We included admission systolic blood pressure as a covariate to account for variations in physiologic presentation. Admission systolic blood pressure was based on the first blood pressure measurement made on presentation to the trauma center and was characterized on an ordinal scale (60 mm Hg=1; 61 to 80 mm Hg=2; 81 to 100 mm Hg=3; 101 to 120 mm Hg=4; 121 to 140 mm Hg=5; 141 to 160 mm Hg=6; >160 mm Hg=7). Out-ofhospital systolic blood pressure was not considered because of the large number of missing values (>50%). Age was defined on an ordinal scale (18 to 30 years=1; 31 to 40 years=2; 41 to 50 years=3; 51 to 60 years=4; 61 to 70 years=5; 71 to 80 years=6; >80 years=7). We adjusted for


44:5

NOVEMBER 2004


mode of transport as a covariate because air medical rescuers (including flight paramedics, nurses, and physicians) are often more skilled at endotracheal intubation and have access to neuromuscular blocking agents.37 We considered the effect of out-of-hospital neuromuscular blocking agents because of recent data linking out-of-hospital rapid sequence intubation to adverse traumatic brain injury outcome.14 We incorporated a propensity score as a covariate in each regression model. A propensity score is a scalar measure summarizing the effect of multiple covariates on different exposure or treatment groups.38 Propensity scores have been widely applied in the medical literature for a variety of purposes, including bias reduction, multivariable matching, and covariate adjustment. We used propensity scores to summarize the effects of selected preexisting medical conditions (cardiac disease, diabetes, gastrointestinal disease, hematologic disorders, psychiatric disorders, immunosuppression, liver disease, malignancy, musculoskeletal disorders, neurologic disorders, obesity, pulmonary disease, renal disease, substance abuse, pregnancy, endocrine disorders), inhospital occurrences (events involving pulmonary, cardiovascular, hematologic or coagulopathy, renal, hepatic, infection or sepsis, airway management, gastrointestinal, neurologic, procedural, decubiti, postoperative hemorrhage, pharmacologic, and burn complications), and social variables (drug and alcohol screening, racial categories, and insurance coverage) (Appendix E2, available online at http:// www.mosby.com/AnnEmergMed). We thought it was important to account for these variables because of previous data associating many of these covariates with trauma outcome.39-45 This strategy enabled us to account for multiple covariates with plausible effects on the exposures and outcomes. Forced incorporation of these characteristics (even through the use of stepwise selection) would place the models at risk of overcharacterization.37 Propensity scores were calculated using the Stata ‘‘pscore’’ module (Stata Corporation, College Station, TX). Logit models using out-of-hospital endotracheal intubation as the outcome variable were used to calculate propensity scores. We included only characteristics with greater than 1% incidence to avoid detecting small and potentially spurious associations. Separate propensity scores were calculated for the full cohort and the subset of survivors. We considered different combinations of preexisting conditions, inhospital occurrences, and social variables to derive propensity scores that optimally satisfied the balancing property.46 The default significance level of P value equal to .01 was used to test the balancing

NOVEMBER 2004

44:5


property. To optimize final model fit, we considered converting the propensity scores to ordinal scales (0.000 to 0.2000=1; 0.2001 to 0.4000=2; 0.4001 to 0.6000=3; 0.6001 to 0.8000=4; 0.8001 to 1.000=5). We did not evaluate interaction terms. Goodness-of-fit of the multivariate models was verified using the HosmerLemeshow test.37 All statistical analyses were conducted using Stata software (version 8.2, Stata Corporation). Sensitivity Analyses

We considered the effects of clustering by trauma center by reanalyzing the models using random-effects logistic regression. In the originally planned analysis, if a patient was dual coded as out-of-hospital endotracheal intubation and ED endotracheal intubation in the registry, the case was classified as ED endotracheal intubation; we repeated the analyses with the exclusion of these dualcoded cases. Finally, we repeated the analysis excluding patients who died in the ED. Data Presentation

We present the primary results using tables summarizing the results of each multivariate model. In each case, we describe the hypothesized associations using odds ratios (ORs) and 95% confidence intervals (CIs) for the primary exposure variable (endotracheal intubation; outof-hospital endotracheal intubation versus ED endotracheal intubation) and the risk-adjustment covariates. The effect of endotracheal intubation (out-of-hospital endotracheal intubation versus ED endotracheal intubation) is the factor of primary interest in each model.

RESULTS Characteristics of Study Subjects

Of 69,226 patients meeting Pennsylvania Trauma Outcome Study inclusion criteria, we excluded 26,104 pediatric or transferred patients. Of the remaining 43,122 patients, there were 9,720 patients with a head/neck Abbreviated Injury Scale score of 3 or greater. Of these 9,720 patients, 4,098 received endotracheal intubation during the ED phase, including 1,797 (43.9%) out-ofhospital endotracheal intubation and 2,301 (56.1%) ED endotracheal intubation. The ED endotracheal intubation group included 31 (1.0%) patients receiving a temporary surgical airway (ICD-9 31.1) and 395 (17.2%) who were originally coded as receiving both out-of-hospital endotracheal intubation and ED endotracheal intubation. Of

4 4 3


the 4,098 out-of-hospital endotracheal intubation or ED endotracheal intubation patients, 377 (9.6%) died in the ED. Of the 5,622 patients who did not receive out-ofhospital endotracheal intubation or ED endotracheal intubation (subjects excluded from this analysis), 490 (8.7%) were intubated at some point during the inhospital course. The mortality for this excluded group was 5.8%. Baseline univariate characteristics are described in Table 1. The groups were similar with respect to sex composition. There were higher proportions of patients aged 18 to 30 years and 31 to 40 years in the out-ofhospital endotracheal intubation group. As indicated by head/neck Abbreviated Injury Scale score, Injury Severity Score, and admission systolic blood pressure, the out-ofhospital endotracheal intubation group appeared to be more severely injured. A higher fraction of out-of-hospital endotracheal intubation patients received neuromuscular blockade and air medical involvement in treatment or care. The unadjusted overall mortality for the study group was 37.1%. There were only 9 (0.2%) burn patients in the study subset; these patients were excluded from multivariate analysis. Of the 4,098 patients, 4,096 (99.9%) sustained at least 1 other serious injury (any non–head/neck Abbreviated Injury Scale score 3). Therefore, the variable ‘‘any other severe injury’’ was excluded from multivariate analysis. Propensity scores were calculated for both the full cohort and the subset of survivors. According to the a priori decision to include only variables with a greater than 1% prevalence, 13 of 51 preexisting conditions and 26 of 46 inhospital occurrences were included in the propensity score calculation (Appendix E2, available at http://www.mosby.com/AnnEmergMed). All drug screening results were included in the propensity score, except for phencyclidine and tricycloids. All insurance coverages were included, except for government insurance (eg, Civilian Health and Medical Program of Uniformed Services, military), which encompassed only 21 (0.5%) of 4,098 patients. All racial categories were included. For the full cohort, the mean propensity scores were 0.50±0.14 for out-of-hospital endotracheal intubation and 0.39±0.16 for ED endotracheal intubation. The balancing property was satisfied for all characteristics except non– insulin-dependent diabetes mellitus and hypothermia; these characteristics were unbalanced in only 1 of 8 blocks, and thus the balancing property was considered to be satisfied. For the subset of survivors, the mean pro-

4 4 4

pensity scores were 0.44±0.15 for out-of-hospital endotracheal intubation and 0.32±0.16 for ED endotracheal intubation. The characteristics respiratory failure, drug screen positive for amphetamines, and private insurance coverage were each unbalanced for only 1 of 9 blocks, and thus the propensity score was considered to be adequately balanced for this subset. Different combinations of candidate variables (including those with 80 Missing Sex Female Male Missing Head/neck AIS score 3 4 5 6 ISS 70 Other severe injury (non–head/neck AIS score 3) Admission systolic blood pressure, mm Hg 60 61–80 81–100 101–120 121–140 141–160 >160 Missing Mechanism of injury Blunt Penetrating Burn Missing Mode of treatment and transport Air medical or [air medical + ground] Ground only Use of out-of-hospital neuromuscular blockade Status on discharge Alive Dead Neurologic outcome (survivors only, n=2,494) Good (eg, home, rehabilitation, facility, psychiatric facility, police custody, against medical advice) Poor (eg, nursing home, other hospital or trauma center) Functional impairment (survivors only, n=1,910) Mild impairment (functional impairment score 0–5)y Moderate impairment (functional impairment score 6–10)y Severe impairment (functional impairment score 11–15)y Dichotomized schemes for functional impairment Moderate or severe vs mild (functional impairment score 6–15 vs 0–5)y Severe vs moderate or mild (functional impairment score 11–15 vs 0–10)y

Out-of-Hospital Endotracheal Intubation, No. (%) (n=1,797)

ED Endotracheal Intubation, No. (%) (n=2,301)

Univariate OR (95% CI)

740 (41.2) 308 (17.1) 271 (15.1) 151 (8.4) 123 (6.8) 114 (6.3) 86 (4.8) 4 (0.2)

774 (33.6) 350 (15.2) 378 (16.4) 219 (9.5) 180 (7.8) 226 (9.8) 168 (7.3) 6 (0.3)

1.00 (N/A) 0.92 (0.77–1.11) 0.75 (0.62–0.90) 0.72 (0.57–0.91) 0.71 (0.56–0.92) 0.53 (0.41–0.68) 0.54 (0.40–0.71) N/A

473 (26.3) 1,324 (73.7) 0 (0.0)

574 (24.9) 1,725 (75.0) 2 (0.9)

1.00 (N/A) 0.93 (0.81–1.08) N/A

333 451 964 49

(18.5) (25.1) (53.6) (2.7)

20 (1.1) 96 (5.3) 418 (23.3) 663 (36.9) 455 (25.3) 78 (4.3) 67 (3.7) 1,797 (100.0)

644 715 914 28

(28.0) (31.1) (39.7) (1.2)

1.00 (N/A) 1.22 (1.02–1.46) 2.04 (1.73–2.40) 3.38 (2.07–5.52)

89 (3.9) 228 (9.9) 807 (35.1) 788 (34.2) 309 (13.4) 42 (1.8) 38 (1.7) 2,299 (99.9)

1.00 (N/A) 1.87 (1.09–3.23) 2.30 (1.40–3.81) 3.74 (2.27–6.18) 6.55 (3.87–11.09) 8.26 (4.12–16.58) 7.85 (3.86–15.95) N/A

120 (5.2) 80 (3.5) 183 (8.0) 312 (13.6) 551 (23.9) 536 (23.3) 482 (20.9) 37 (1.6)

1.00 (N/A) 0.48 (0.33–0.70) 0.41 (0.30–0.56) 0.38 (0.29–0.50) 0.27 (0.21–0.35) 0.21 (0.16–0.28) 0.21 (0.16–0.28) N/A

1,573 (87.5) 219 (12.2) 4 (0.2) 1 (0.1)

2,033 (88.4) 263 (11.4) 5 (0.2) 0 (0.0)

1.00 (N/A) 1.08 (0.89–1.30) 1.03 (0.28–3.86) N/A

1,218 (67.8) 579 (32.2) 858 (47.7)

370 (16.1) 1,931 (83.9) 135 (5.9)

1.00 (N/A) 0.09 (0.08–0.11) 14.66 (11.98–17.98)

926 (51.5) 871 (48.5)

1,652 (71.8) 649 (28.2)

1.00 (N/A) 2.39 (2.10–2.73)

739 (81.8)

1,344 (85.5)

1.00 (N/A)

164 (18.2)

247 (15.5)

1.21 (0.97–1.51)

288 (43.8) 125 (19.0) 244 (37.1)

761 (60.7) 205 (16.4) 287 (22.9)

1.00 (N/A) 1.61 (1.24–2.09) 2.25 (1.80–2.80)

— —

— —

1.98 (1.63–2.41) 1.99 (1.61–2.46)

296 95 186 292 362 284 251 31

(16.5) (5.3) (10.4) (16.2) (20.1) (15.8) (14.0) (1.7)

AIS, Abbreviated Injury Scale; ISS, Injury Severity Score, N/A, not applicable. *ORs reflect differences in distributions between exposure groups (endotracheal intubation; out-of-hospital endotracheal intubation vs ED endotracheal intubation). The dichotomized schemes for functional impairment score (eg, moderate or severe vs mild, severe vs moderate or mild) were used in multivariate modeling. y Functional impairment score.

NOVEMBER 2004

44:5


4 4 5


differences in the parameter estimates or the resulting inferences. LIMITATIONS

The most evident limitation of this effort is the use of a preexisting, unvalidated registry. We believe that this approach was justified for several reasons. The use of preexisting data is efficient; we had immediate access to a large-scale body of data with numerous characteristics for outcome identification and risk adjustment. The Pennsylvania Trauma Outcome Study registry collects data on more than 10,000 patients per year using rigorous quality control measures, including detailed audits and the use of automated error-checking software. Although quality of abstraction does not denote data accuracy, our goal was to identify associations, not to develop precise prediction models; the degree of data accuracy was likely adequate for this application. A prospective data collection effort would have been expensive, difficult, laborious, inefficient, and subject to the same concerns about data reliability, accuracy, and validity. Although we applied an unvalidated measure, the functional impairment score, we used this scale only to differentiate broad ranges of (not minute differences in) functional outcome. We believe that this measure was more informative than the surrogate marker discharge

Table 2. Adjusted odds of death.*

Variable Endotracheal intubation (out-of-hospital endotracheal intubation vs ED endotracheal intubation) Age (ordinal) Sex (male vs female) Head/neck AIS score ISS (ordinal) Mechanism of injury (penetrating vs blunt) Systolic blood pressure on admission (ordinal) Mode of treatment/transport (ground only vs air or [air + ground]) Out-of-hospital neuromuscular blockade use Propensity score (ordinal)

Table 3. Adjusted odds of poor neurologic outcome (discharge to long-term care facility, hospital, or other trauma center destination).* Death on Hospital Discharge, OR (95% CI) 3.99 (3.21–4.93) 1.49 0.89 2.30 1.19 4.38 0.67 2.28

(1.42–1.56) (0.74–1.07) (2.02–2.60) (1.08–1.31) (3.36–5.71) (0.64–0.71) (1.83–2.85)

0.48 (0.38–0.60) 1.18 (1.06–1.31)

*Includes only patients from full cohort with complete data (n=4,008). HosmerLemeshow goodness of fit P=.83. The calculated propensity score was based on selected preexisting conditions, inhospital occurrences, results of drug screening, and insurance coverage (Appendix E2) and was parameterized on an ordinal scale (0.000– 0.2000=1; 0.2001–0.4000=2; 0.4001–0.6000=3; 0.6001–0.8000=4; 0.8001–1.000=5).

4 4 6

destination, which has been used by other authors.11,14 Discharge destination has similarly not been validated. We used the combination of Injury Severity Score, head/neck Abbreviated Injury Scale score, and admission systolic blood pressure as the primary risk-adjustment variables. We also incorporated a propensity score based on preexisting conditions, inhospital occurrences, and social variables to account for potential selection bias in the exposure groups. As discussed previously, this riskadjustment strategy represented the best of the available analytic options. The registry did not contain details about the course of out-of-hospital airway management. For example, we could not identify failed out-of-hospital endotracheal intubation efforts. We also could not adjust for the presence of factors known to affect out-of-hospital endotracheal intubation efforts (eg, trismus and the presence of gag reflex).7-9 We also did not have information about the course of ED airway care; for example, did a patient who did not appear to require endotracheal intubation in the out-of-hospital setting subsequently deteriorate in the ED and require intubation? We accepted these limitations because the broader goal of this analysis was to draw a connection between successful out-of-hospital endotracheal intubation and inhospital outcomes. It is conceivable that exposure to a failed out-of-hospital endotracheal intubation may have hindered a patient’s outcome. As described by Dunford et al,47 potentially deleterious

Poor Neurologic Outcome, OR (95% CI)

Variable Endotracheal intubation (out-of-hospital endotracheal intubation vs ED endotracheal intubation) Age (ordinal) Sex (male vs female) Head/neck AIS score ISS (ordinal) Mechanism of injury (penetrating vs blunt) Systolic blood pressure on admission (ordinal) Mode of treatment/transport (ground only vs air or [air + ground]) Out-of-hospital neuromuscular blockade use Propensity score (survivors only, ordinal)

1.61 (1.15–2.26) 1.61 0.77 1.37 1.14 0.88 0.83 1.39

(1.51–1.72) (0.60–1.00) (1.13–1.64) (0.99–1.32) (0.51–1.51) (0.76–0.90) (1.00–1.91)

0.92 (0.65–1.29) 1.04 (0.90–1.20)

*Includes only survivors with complete data (n=2,452). Hosmer-Lemeshow goodness of fit P=.91. The propensity score used in this model was calculated according to survivors only.


44:5

NOVEMBER 2004


events such as bradycardia and hypoxia can occur during out-of-hospital endotracheal intubation efforts, and there is speculation that these events may be linked to traumatic brain injury outcome.3 Murray et al12 found that unsuccessful intubation was associated with a 2.58-times increased odds of death compared with nonintubation. Given these previous observations, we speculate that were we able to identify the failed out-of-hospital endotracheal intubation subset, we would probably have observed even greater adjusted odds of death and functional impairment than those presented in the current analysis. We adjusted for transport mode to control for exposure to air medical paramedics, who are generally more skilled at endotracheal intubation and carry neuromuscular blocking agents. However, it is conceivable that this characteristic simply reflected the effect of transport time to trauma center on outcome. Out-of-hospital time information was too incomplete in this data set to facilitate meaningful analyses. In this analysis, we did not use matching techniques (such as conditional logistic regression) because there was adequate sample size to make accurate inferences using simple covariate adjustment. We were also concerned about the potential loss in sample size from unmatched pairs. For example, in the study by Murray et al,12 of the original cohort of 795 patients, only 57 matched pairs could be identified, a loss of more than 85% of the data set. The loss of such a large fraction of the data set would have reduced the statistical power and clinical utility of the current analysis.

DISCUSSION

Although numerous publications exist describing out-ofhospital endotracheal intubation, there are only limited reports linking out-of-hospital endotracheal intubation with inhospital outcomes in either medical or trauma patients.12-14,48-51 Our findings reinforce the importance of this potential connection. Although the patients in the out-of-hospital endotracheal intubation group appear to be more severely injured, our data suggest that an association between out-of-hospital endotracheal intubation and death exists even after adjustment for injury severity and other potential confounders. In fact, the odds of death increased after adjustment for potential confounders. Likewise, we also demonstrated adverse associations between out-of-hospital endotracheal intubation and neurologic and functional outcomes adjusted for potential confounders. Our study has many notable features. The analysis used a large-scale data set drawing on trauma patients from an entire state, not a single county or region. The Pennsylvania Trauma Outcome Study data set provided efficient access to vast amounts of clinical data, including outcomes. We used multivariate methods to account for potential confounders. Propensity scores accounted for the small but potentially influential effects of preexisting conditions, inhospital occurrences, and social variables. We also quantified functional outcome using the functional impairment score, which was more informative than the surrogate variable ‘‘discharge destination’’ used in

Table 4. Adjusted odds of functional impairment.*

Variable Endotracheal intubation (out-of-hospital endotracheal intubation vs ED endotracheal intubation) Age (ordinal) Sex (male vs female) Head/neck AIS score ISS (ordinal) Mechanism of injury (penetrating vs blunt) Systolic blood pressure on admission (ordinal) Mode of treatment/transport (ground only vs air or [air + ground]) Out-of-hospital neuromuscular blockade use Propensity score (survivors only, ordinal)

Moderate or Severe Functional Impairment (Functional Impairment Score 6–15), OR (95% CI)

Severe Functional Impairment (Functional Impairment Score 11–15), OR (95% CI)

1.92 (1.40–2.64)

1.80 (1.29–2.52)

1.41 0.82 2.18 1.39 1.18 0.93 1.13 0.77 1.49

1.43 0.87 2.66 1.19 0.91 0.91 1.19 0.96 1.43

(1.32–1.51) (0.64–1.05) (1.85–2.57) (1.21–1.58) (0.75–1.86) (0.86–1.01) (0.85–1.51) (0.56–1.06) (1.31–1.69)

(1.33–1.53) (0.67–1.13) (2.20–3.22) (1.02–1.38) (0.53–1.56) (0.84–0.99) (0.87–1.63) (0.69–1.34) (1.24–1.64)

*Includes only survivors with complete data where functional impairment score components were reported (n=1,886). ORs reflect adjusted odds of moderate-or-severe (functional impairment score 6–15) and severe (functional impairment score 11–15) functional impairment. Hosmer-Lemeshow goodness of fit P=.57 and .36, respectively. The propensity score used in this model was calculated according to survivors only.

NOVEMBER 2004

44:5


4 4 7


previous studies.11,14 (We included discharge destination as a secondary outcome in this analysis strictly to facilitate comparisons with these previous studies.) We used a narrowly defined study population, including only those who were exposed to endotracheal intubation in either the out-of-hospital or ED setting. We excluded patients who were not intubated in either of these settings. We also excluded patients who did not have the opportunity to be exposed to out-of-hospital endotracheal intubation, for example, patients who were not treated by advanced life support out-of-hospital emergency personnel. Our goal was to assess only patients who were exposed to endotracheal intubation during the acute out-of-hospital or ED resuscitation phases. In post hoc analyses, we confirmed that the 5,622 patients who did not receive out-of-hospital or ED endotracheal intubation were either intubated later in the inhospital course (operating room, ICU, regular floor) or not intubated at all during their hospital stay. Although several studies have examined the effect of out-of-hospital endotracheal intubation on outcome in traumatic brain injury patients, these studies have important limitations that contrast with the current effort. Winchell and Hoyt11 found that out-of-hospital endotracheal intubation imparted a positive effect on traumatic brain injury survival and no effect on discharge destination, but their analysis used a smaller sample size (n=671) and applied only univariate analyses without controlling for severity of injury or clinical presentation. Bochicchio et al13 similarly examined intubated head-injured patients but used only univariate techniques applied to a limited series (n=191). Similar to our study, Murray et al12 used multivariate methods and found an adjusted odds of death of 4.18 for the out-of-hospital endotracheal intubation group. None of these studies examined functional outcomes or considered the concurrent effects of preexisting conditions, inhospital occurrences, or social variables, as we attempted in this effort. The most widely publicized data about out-of-hospital endotracheal intubation and traumatic brain injury outcome was reported by Davis et al,14 who found that out-ofhospital rapid sequence intubation was associated with an increased adjusted odds of death (OR 1.6). We emphasize that our data are not directly comparable with this series, which examined the prospective implementation of paramedic neuromuscular blockade–assisted intubation. The vast majority of out-of-hospital services in the United States do not have access to neuromuscular blocking agents to facilitate endotracheal intubation.52 Our series contains a broader range of patient encounters treated by a diverse

4 4 8

range of emergency personnel, which may improve the generalizability of our findings because it better represents the heterogeneity in the emergency medical services rescuer population seen in most regions. Although our multivariate models suggested that the use of neuromuscular blocking agents was associated with decreased mortality, this variable was used for covariate adjustment only; we cannot conclude that the use of paralytic agents reduces traumatic brain injury mortality. Likewise, our data suggest that the involvement of air medical crews (including flight paramedics, nurses, and physicians) improved outcome, which may be partially influenced by the fact that air medical crews in Pennsylvania carry neuromuscular blocking agents. However, because this variable was used for covariate adjustment, we again hesitate to make any similar inferences. The clinical application of our findings naturally presents important challenges. We emphasize that this is a retrospective study and thus identifies an association between out-of-hospital endotracheal intubation and traumatic brain injury outcome, not causality. Although we identified that out-of-hospital endotracheal intubation is deleterious in traumatic brain injury, we do not believe that the correct clinical interpretation is to defer out-ofhospital endotracheal intubation. Our results most likely signal the presence of underlying problems with the process of out-of-hospital endotracheal intubation. This interpretation is not unreasonable because there are multiple reports of complications and errors associated with out-of-hospital endotracheal intubation.7-9 Paramedics are known to have limited endotracheal intubation training and clinical skills and do not have full access to the drugs used for ED intubations.52 Dunford et al47 showed that desaturation and bradycardia are frequent events during out-of-hospital endotracheal intubation of traumatic brain injury patients. Katz and Falk23 showed that unrecognized endotracheal tube misplacement is a frequent occurrence in out-of-hospital endotracheal intubation. One of the goals of ED rapid sequence intubation is to attenuate theoretical intracranial pressure changes in patients with traumatic brain injury; it is plausible that conventional laryngoscopy by paramedics without the benefit of neuroprotective drugs may be deleterious in this subset.53 Therefore, our results are unlikely to simply represent spurious statistical observations. Our results likely reflect the plausible connection between the manner of execution of out-of-hospital endotracheal intubation and patient outcome. There are several potential directions for future efforts. Others may validate or refute our observations using


44:5

NOVEMBER 2004


another large data set. A prospective observational trial encompassing out-of-hospital and ED phases could yield considerable insight about the airway management of head-injured patients in these settings and how these techniques may be related to outcome. As discussed previously, however, a prospective data collection effort covering 2 phases of airway management would be difficult to perform. Finally, a logical—but risky and controversial—direction would be to conduct a controlled trial randomizing traumatic brain injury patients to either out-of-hospital endotracheal intubation or no out-ofhospital endotracheal intubation. In conclusion, we observed adverse associations between out-of-hospital endotracheal intubation, survival, and neurologic and functional outcomes after severe traumatic brain injury. Additional study is needed to distinguish whether it is the procedure (out-of-hospital endotracheal intubation) or its manner of performance that is the causative factor. We thank the Pennsylvania Trauma Systems Foundation, Mechanicsburg, PA, for their assistance with this study. We especially thank Mary Ann Spott, MPA, MSIS, Pennsylvania Trauma Systems Foundation, for making this study possible. We also thank Joseph P. Costantino, DrPH, Department of Biostatistics, University of Pittsburgh, for his invaluable statistical guidance and mentorship. Author contributions: HEW, ABP, LDC, PDA, and DMY conceived and designed the study. HEW performed the statistical analyses. HEW drafted the manuscript, and all authors contributed to its revision. HEW takes responsibility for the paper as a whole. Received for publication February 18, 2004. Revisions received March 31, 2004, and April 5, 2004. Accepted for publication April 9, 2004. Available online August 12, 2004. Presented at the National Association of EMS Physicians annual meeting, Tucson, AZ, January 10, 2004. Dr. Wang is supported by Clinical Research Training Award 1 K08 HSO13628 from the Agency for Healthcare Research and Quality. Reprints not available from the authors. Address for correspondence: Henry E. Wang, MD, MPH, Department of Emergency Medicine, University of Pittsburgh School of Medicine, 230 McKee Place, Suite 400, Pittsburgh, PA 15213; 412-647-4925; E-mail [email protected].

REFERENCES 1. Muizelaar JP, Marmarou A, Ward JD, et al. Adverse effects of prolonged hyperventilation in patients with severe head injury: a randomized clinical trial. J Neurosurg. 1991;75:731-739. 2. Pigula FA, Wald SL, Shackford SR, et al. The effect of hypotension and hypoxia on children with severe head injuries. J Pediatr Surg. 1993;28:310-314. 3. Chesnut RM, Marshall LF, Klauber MR, et al. The role of secondary brain injury in determining outcome from severe head injury. J Trauma. 1993;34:216-222.

NOVEMBER 2004

44:5


4. Biros MH, Heegaard W. Prehospital and resuscitative care of the head-injured patient. Curr Opin Crit Care. 2001;7:444-449. 5. Bullock R, Chesnut RM, Clifton G, et al. Guidelines for the management of severe head injury: Brain Trauma Foundation. Eur J Emerg Med. 1996;3:109-127. 6. Brain Trauma Foundation.Guidelines for prehospital management of traumatic brain injury. Available at: http://www.braintrauma.org. Accessed March 24, 2004. 7. Wang HE, Kupas DF, Paris PM, et al. Multivariate predictors of failed out-of-hospital endotracheal intubation. Acad Emerg Med. 2003;10:717-724. 8. Doran JV, Tortella BJ, Drivet WJ, et al. Factors influencing successful intubation in the prehospital setting. Prehospital Disaster Med. 1995;10:259-264. 9. Krisanda TJ, Eitel DR, Hess D, et al. An analysis of invasive airway management in a suburban emergency medical services system. Prehospital Disaster Med. 1992;7: 121-126. 10. Ochs M, Davis D, Hoyt D, et al. Paramedic-performed rapid sequence intubation of patients with severe head injuries. Ann Emerg Med. 2002;40:159-167. 11. Winchell RJ, Hoyt DB. Endotracheal intubation in the field improves survival in patients with severe head injury: Trauma Research and Education Foundation of San Diego. Arch Surg. 1997;132:592-597. 12. Murray JA, Demetriades D, Berne TV, et al. Prehospital intubation in patients with severe head injury. J Trauma. 2000;49:1065-1070. 13. Bochicchio GV, Ilahi O, Joshi M, et al. Endotracheal intubation in the field does not improve outcome in trauma patients who present without an acutely lethal traumatic brain injury. J Trauma. 2003;54:307-311. 14. Davis DP, Hoyt DB, Ochs M, et al. The effect of paramedic rapid sequence intubation on outcome in patients with severe traumatic brain injury. J Trauma. 2003;54: 444-453. 15. Harbrecht BG, Zenati MS, Ochoa JB, et al. Management of adult blunt splenic injuries: comparison between Level I and Level II trauma centers. J Am Coll Surg. 2004; 198:232-239. 16. Page RB, Spott MA, Krishnamurthy S, et al. Head injury and pulmonary embolism: a retrospective report based on the Pennsylvania Trauma Outcomes study. Neurosurgery. 2004;54:143-148. 17. Potoka DA, Schall LC, Ford HR. Improved functional outcome for severely injured children treated at pediatric trauma centers. J Trauma. 2001;51:824-832. 18. Reilly PM, Schwab CW, Haut ER, et al. Training in trauma surgery: quantitative and qualitative aspects of a new paradigm for fellowship. Ann Surg. 2003; 238:596-603. 19. Schall LC, Potoka DA, Ford HR. A new method for estimating probability of survival in pediatric patients using revised TRISS methodology based on age-adjusted weights. J Trauma. 2002;52:235-241. 20. Davis EG, MacKenzie EJ, Sacco WJ, et al. A new ‘‘TRISS-like’’ probability of survival model for intubated trauma patients. J Trauma. 2003;55:53-61. 21. Association for the Advancement of Automative Medicine. Abbreviated Injury Scale 1990. Barrington, IL: Association for the Advancement of Automative Medicine; 1998. 22. Wang HE, Kupas DF, Paris PM, et al. Preliminary experience with a prospective, multi-centered evaluation of out-of-hospital endotracheal intubation. Resuscitation. 2003;58:49-58. 23. Katz SH, Falk JL. Misplaced endotracheal tubes by paramedics in an urban emergency medical services system. Ann Emerg Med. 2001;37:32-37. 24. Jemmett ME, Kendal KM, Fourre MW, et al. Unrecognized misplacement of endotracheal tubes in a mixed urban to rural emergency medical services setting. Acad Emerg Med. 2003;10:961-965. 25. Hsiao AK, Michelson SP, Hedges JR. Emergent intubation and CT scan pathology of blunt trauma patients with Glasgow Coma Scale scores of 3-13. Prehospital Disaster Med. 1993;8:229-236. 26. Cassidy LD, Potoka DA, Adelson PD, et al. Development of a novel method to predict disability after head trauma in children. J Pediatr Surg. 2003;38:482-485. 27. McCauley SR, Hannay HJ, Swank PR. Use of the Disability Rating Scale Recovery curve as a predictor of psychosocial outcome following closed-head injury. J Int Neuropsychol Soc. 2001;7:457-467. 28. Rappaport M, Hall KM, Hopkins K, et al. Disability rating scale for severe head trauma: coma to community. Arch Phys Med Rehabil. 1982;63:118-123. 29. Choi SC, Bullock R. Design and statistical issues in multicenter trials of severe head injury. Neurol Res. 2001;23:190-192. 30. Zink BJ. Traumatic brain injury outcome: concepts for emergency care. Ann Emerg Med. 2001;37:318-332. 31. Baker SP, O’Neill B, Haddon W Jr, et al. The injury severity score: a method for describing patients with multiple injuries and evaluating emergency care. J Trauma. 1974;14:187-196. 32. Boyd CR, Tolson MA, Copes WS. Evaluating trauma care: the TRISS method: Trauma Score and the Injury Severity Score. J Trauma. 1987;27:370-378.

4 4 9


33. Champion HR, Copes WS, Sacco WJ, et al. Improved predictions from a severity characterization of trauma (ASCOT) over Trauma and Injury Severity Score (TRISS): results of an independent evaluation. J Trauma. 1996;40:42-48. 34. Rutledge R, Osler T, Emery S, et al. The end of the Injury Severity Score (ISS) and the Trauma and Injury Severity Score (TRISS): ICISS, an International Classification of Diseases, Ninth Revision–based prediction tool, outperforms both ISS and TRISS as predictors of trauma patient survival, hospital charges, and hospital length of stay. J Trauma. 1998;44:41-49. 35. Rutledge R. The Injury Severity Score is unable to differentiate between poor care and severe injury. J Trauma. 1996;40:944-950. 36. Hoyt DB. Is it time for a new injury score? Lancet. 1998;352:920-921. 37. Hosmer DW, Lemeshow S. Applied Logistic Regression. New York, NY: John Wiley & Sons; 2000. 38. D’Agostino RB Jr. Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group. Stat Med. 1998;17:2265-2281. 39. Previdi JK, Cayten CG, Byrne DW. Early predictors of sepsis in the motor-vehicle crash trauma victim. Prehospital Disaster Med. 1996;11:27-36. 40. Ferrera PC, Bartfield JM, D’Andrea CC. Outcomes of admitted geriatric trauma victims. Am J Emerg Med. 2000;18:575-580. 41. Hannan EL, Mendeloff J, Farrell LS, et al. Multivariate models for predicting survival of patients with trauma from low falls: the impact of gender and pre-existing conditions. J Trauma. 1995;38:697-704. 42. Morris JA Jr, MacKenzie EJ, Damiano AM, et al. Mortality in trauma patients: the interaction between host factors and severity. J Trauma. 1990;30:1476-1482. 43. Marcin JP, Schembri MS, He J, et al. A population-based analysis of socioeconomic status and insurance status and their relationship with pediatric trauma hospitalization and mortality rates. Am J Public Health. 2003;93:461-466. 44. Blondell RD, Looney SW, Krieg CL, et al. A comparison of alcohol-positive and alcohol-negative trauma patients. J Stud Alcohol. 2002;63:380-383. 45. Langdorf MI, Rudkin SE, Dellota K, et al. Decision rule and utility of routine urine toxicology screening of trauma patients. Eur J Emerg Med. 2002;9:115-121. 46. Ichino A, Becker S. Estimation of average treatment effects based on propensity scores. Stata J. 2002;2:358-377. 47. Dunford JV, Davis DP, Ochs M, et al. Incidence of transient hypoxia and pulse rate reactivity during paramedic rapid sequence intubation. Ann Emerg Med. 2003;42: 721-728. 48. Gausche M, Lewis RJ, Stratton SJ, et al. Effect of out-of-hospital pediatric endotracheal intubation on survival and neurological outcome: a controlled clinical trial. JAMA. 2000;283:783-790. 49. Brice JH, Garrison HG, Evans AT. Study design and outcomes in out-of-hospital emergency medicine research: a ten-year analysis. Prehospital Emerg Care. 2000;4: 144-150. 50. Cone DC. Outcomes research and emergency medical services: the time has come. Acad Emerg Med. 2000;7:188-191. 51. Garrison HG, Maio RF, Spaite DW, et al. Emergency Medical Services Outcomes Project III (EMSOP III): the role of risk adjustment in out-of-hospital outcomes research. Ann Emerg Med. 2002;40:79-88. 52. McDonald CC, Bailey B. Out-of-hospital use of neuromuscular-blocking agents in the United States. Prehospital Emerg Care. 1998;2:29-32. 53. Walls RM. Rapid-sequence intubation in head trauma. Ann Emerg Med. 1993;22: 1008-1013.

4 5 0


44:5

NOVEMBER 2004