JOURNAL OF NEUROTRAUMA 31:1077–1082 (June 15, 2014) ª Mary Ann Liebert, Inc. DOI: 10.1089/neu.2013.3281
Original Articles
Restraints and Peripheral Nerve Injuries in Adult Victims of Motor Vehicle Crashes Kimon Bekelis,1,* Symeon Missios,2,* and Robert J. Spinner,3–5
Abstract
The pattern of injuries in restrained victims of motor vehicle crashes (MVCs) remains an issue of debate. We investigated the association of peripheral nerve injuries with the use of protective devices (seat belt and air bag) during MVCs. We performed a retrospective cohort study of 384,539 adult MVC victims who were registered in the National Trauma Data Bank (NTDB) between 2009 and 2011. Regression techniques were used to investigate the association of restraint use with the risk of peripheral nerve injury in patients hospitalized after an MVC. Of the study patients, 271,099 were using restraints and 113,440 were not. Overall, there were a total of 3086 peripheral nerve injuries. Multivariable logistic regression analysis demonstrated an association of protective device use with decreased risk of peripheral nerve injury (odds ratio [OR], 0.89; 95% confidence interval [CI], 0.82– 0.96; absolute risk reduction, 10.68%). This corresponds to 16 patients who needed to be restrained to prevent one nerve injury. The location of the patient in the vehicle did not seem to affect the risk of peripheral nerve injury, with drivers demonstrating no association with nerve injuries (OR, 0.94; 95% CI, 0.87–1.02) in comparison with non-drivers. On the contrary, alcohol consumption was associated with increased incidence of peripheral nerve injuries (OR, 1.10; 95% CI, 1.01–1.20). In summary, restraint use was associated with decreased risk of peripheral nerve injury in MVC victims, after controlling for confounders. Key words: air bag; motor vehicle crash; NTDB; peripheral nerve injury; seat belt
Introduction
M
otor vehicle crashes (MVCs) are the leading cause of death among young adults in the United States.1–3 In addition, they contribute to significant morbidity for nearly 3 million patients who sustain non-lethal injuries annually.1–3 Laws mandating the use of active (seat belts) and passive (air bags) restraints while penalizing drunken driving have had a major impact on the reduction of MVC-related fatalities in recent years.3,4 Despite widespread regulation and overall increases in safer behaviors, crash data1 show that more than 50% of fatalities were among unrestrained occupants and nearly 40% involved alcohol. Besides increasing MVC survival,5 restraints have decreased the incidence of serious injuries, including brain and facial trauma, intra-abdominal solid organ injuries, and long bone fractures.5 A new group of injuries, however, has been identified related to the use of restraints.5 These include hollow viscous injuries, aortic tears, abdominal wall hernias, and neck trauma. The association of peripheral nerve injuries (PNI) with the use of supplemental restraints during MVCs remains elusive. Although PNI are a major contributor to incapacitating pain and disability, their prevention in
MVCs has not been studied. This pathology is extremely rare, and often under-recognized during the initial stages of the accident. Most studies describing the phenomenon involve retrospective analyses of single institution or regional experiences,6–9 demonstrating results with limited generalization, given their inherent selection bias. These reports, however, have not examined the association of PNI with the use of restraints. We hypothesized that protective devices (seat belt and air bag) are associated with decreased risk of PNI for victims of MVCs. We tested this hypothesis using a regression model in a cohort of patients who were involved in MVCs and whose data were recorded in the 2009–2011 National Trauma Databank (NTDB). The advantage of using a comprehensive database, such as the NTDB, is that it provides an adequate event volume to study a rare entity, such as PNI. Methods NTDB All patients with MVCs who were registered in NTDB (Committee on Trauma, American College of Surgeons) between 2009
1
Section of Neurosurgery, Dartmouth-Hitchcock Medical Center, Lebanon, New Hampshire. Department of Neurosurgery, Cleveland Clinic, Cleveland, Ohio. Departments of 3Neurosurgery and 4Orthopedics, Mayo Clinic, Rochester, Minnesota. 5 Department of Anatomy, Mayo Medical School, Rochester, Minnesota. *The first two authors contributed equally. 2
1077
1078
BEKELIS ET AL.
and 2011 were included in the analysis. The NTDB is the largest collection of US trauma registry data ever assembled. It includes data collected from more than 900 institutions in the United States.10 We merged the available data sets for the variables of interest, using source codes provided by the NTDB.11 More information about the NTDB is available at http://www.facs.org/ trauma/ntdb/. Cohort definition To establish the cohort of patients, we used International Classification of Diseases, Ninth Revision (ICD-9) E-codes to identify patients in the registry who were involved in MVCs. E-codes 810, 810.1, 810.8, 811, 811.1, 811.8, 812, 812.1, 812.4, 812.8, 813, 813.1, 813.4, 813.8, 814, 814.1, 814.4, 814.8, 815, 815.1, 815.4, 815.8, 816, 816.1, 816.4, 816.8, 819, 819.1, 819.4, 819.8, 820, 820.1, 820.8, 821, 821.1, 821.4, 821.8, 822, 822.1, 822.8, 823, 823.1, 823.4, 823.8, 824, 824.1, 824.4, 824.8, 825, 825.1, 825.8 were eligible for inclusion. Patients older than 16 years who were drivers or passengers involved in MVCs were included for analysis. Exclusion criteria consisted of patients who were dead on arrival to the emergency department (ED) and incomplete record information on the use of restraints and age. Patients involved in motorcycle accidents were not included in the cohort creation. Primary outcome and exposure variables The primary outcome variable was the incidence of peripheral nerve injury. This was defined as an ICD-9 code of 953.x, 954.x, 955.x, 956.x, and 957.x. The effect on the outcome of the pertinent exposure variables was examined. Age, systolic blood pressure (SBP), heart rate (HR), temperature on arrival to the ED, sex, race (African American, Hispanic, Asian, Caucasian, American Indian,
FIG. 1.
and other), insurance (private insurance, self-pay, Medicaid, and Medicare, automobile, and other), use of protective devices (victims wearing their seat belts, cars equipped with air bags that deployed during the crash), location of the victim in the vehicle (driver, passenger), and alcohol use were categorical variables. Glasgow Coma Scale (GCS) score and the patient’s Injury Severity Score (ISS) were ordinal variables. The hospital characteristics used in the analysis as categorical variables included hospital region (West, South, Midwest, North), hospital bed size ( < 200, 200– 400, 400–600 and > 600 beds), and teaching status of the hospital (university, nonteaching, community). Variables were included in the final models after ensuring that the proportion of missing data points was less than 20%. Illicit drug use, medical comorbidities, respiratory rate, O2 saturation, total elapsed Emergency Medical Services (EMS) time from dispatch to ED, EMS time at the scene, and patient days in the intensive care unit and/or on a ventilator were excluded because they were found to have a prevalence of missing data greater than 20%. Statistical analysis The effect of the exposure variables on the primary outcome was examined using a multivariable logistic regression model. Multiple imputation was performed for each variable associated with missing values using the Amelia II package12 in the 64-bit version of R 3.0.2 (R Foundation for Statistical Computing). Imputation was used for the following missing data: sex, payer source, race, ISS score, SBP, HR, temperature, GCS score, use of alcohol, and region. First, the proportion of missing data for variables of interest was calculated. The Amelia II program was used to impute missing data based on the other available variables. This process was repeated five times, creating five separate imputed data sets. These five data sets were combined to create a full-pooled data set
Cohort selection for the study.
Table 1A. Patient and Hospital Characteristics All patients Sample size
384,539 Mean 40.45 N
Age Sex F M Unreported data Region North South Midwest West Unreported data Payer Medicare Medicaid Private payer Self-payer Automobile Other Unreported data Race Caucasian African American Hispanic Asian American Indian Other Unreported data Hospital status Community Non-teaching University Bed size < 200 200–400 400–600 > 600 Number of neurosurgeons Number of trauma surgeons
Protective devices used
SD 19.33 %
271,099 Mean 42.25 N
164,292 220,126
42.74 57.26 121
128,383 142,639 77
47.37 52.63
55,293 152,620 91,699 83,544
14.43 39.83 23.93 21.80 1383
37,496 105,954 64,617 62,214 818
27,265 30,660 109,607 63,199 69,610 41,280
7.98 8.97 32.08 18.50 20.38 12.08 42,918
260,549 49,170 41,753 7313 3344 11,028
No protective devices used 113,440 Mean 36.15 N
SD 19.89 %
p value
SD 17.17 %
< 0.0001
35,909 77,487 44
31.67 68.33
< 0.0001 < 0.0001
13.87 39.20 23.91 23.02
17,797 46,666 27,082 21,330 565
15.77 41.34 23.99 18.90
< 0.0001 < 0.0001 0.80
20,923 18,880 78,151 41,576 53,261 29,321 28,987
8.64 7.80 32.28 17.17 22.00 12.11
6342 11,780 31,456 21,623 16,349 11,959 13,930
6.37 11.84 31.61 21.73 16.43 12.02
< 0.0001 < 0.0001 < 0.0001 < 0.0001 < 0.0001 0.012
69.82 13.18 11.19 1.96 0.90 2.96 11382
179,753 37,147 30,199 6227 1613 8448 7712
68.25 14.10 11.47 2.36 0.61 3.21
80,796 12,023 11,554 1086 1731 2580 3670
73.60 10.95 10.53 0.99 1.58 2.35
< 0.0001 < 0.0001 < 0.0001 < 0.0001 < 0.0001 < 0.0001
140,665 56,507 187,367
36.58 14.69 48.73
99,470 41,693 129,936
36.69 15.38 47.93
41,195 14,814 57,431
36.31 13.06 50.63
0.027 < 0.0001 < 0.0001
24,033 97,160 117,326 146,020 5.3 6.5
6.25 25.27 30.51 37.97 3.2 2.7
17,665 67,204 83,639 102,591 5.2 6.5
6.52 24.79 30.85 37.84 3.2 2.7
6368 29,956 33,687 43,429 5.3 6.6
5.61 26.41 29.70 38.28 3.1 2.8
< 0.0001 < 0.0001 < 0.0001 0.010 < 0.0001 < 0.0001
SD, standard deviation.
Table 1B. Patient Characteristics All patients
SBP mm Hg GCS score HR beats/min Temperature oC Driver Alcohol use 1–24 24–49 50–75
No protective devices used
Mean
SD
Mean
SD
Mean
SD
p value
137.0 14.0 91.8 36.4 N
25.4 2.9 19.6 2.3 %
137.9 14.3 91.1 36.4 N
25.3 2.4 19.1 2.1 %
134.9 13.3 93.7 36.3 N
20.6 3.7 20.6 2.6 %
< 0.0001 < 0.0001 < 0.0001 < 0.0001
290,772 78,109
75.62 22.57 38468 0.80 0.91 0.09 0.01
210,165 45,888 28,100 1974 250,619 19,054 1426
77.52 18.88
80,607 32,221 10,368 1112 98,124 14,085 1231
71.06 31.26
< 0.0001 < 0.0001
0.98 86.50 12.42 1.09
< 0.0001 < 0.0001 < 0.0001 < 0.0001
Unreported data Peripheral nerve injuries Injury Severity Score
Protective devices used
3086 348,743 33,139 2657
0.73 92.45 7.03 0.53
SD, standard deviation; SBP, systolic blood pressure; GCS, Glasgow Coma Scale; HR, heart rate.
1079
1080
BEKELIS ET AL.
Table 2. Incidence of Peripheral Nerve Injuries in Different Subgroups Group + Alcohol use + Use of protective devices + Driver
Patients with peripheral injuries 2271/290,772 815/93,767 758/78,109 2031/267,962 1974/271,099 1112/113,440
%
Table 3. Models Demonstrating the Association of Protective Device Use with the Incidence of Peripheral Nerve Injury
p value
0.78 0.009 0.87 0.97 < 0.0001 0.76 0.73 < 0.0001 0.98
with no missing values, which was used in a multinomial logistic regression (logit) model using the Zeilig statistical package13,14 in the 64-bit version of R 3.0.2. The method of multivariable logistic regression was used to determine the association of use of restraints with the incidence of PNI controlling for: age, sex, race, payer, SBP, GCS, HR, temperature, hospital status, hospital bed size, alcohol use, region, and ISS. Regression diagnostics were performed for all logistic regression models. The Hosmer-Lemeshow and Pearson goodness-of-fit tests were used to confirm that the models adequately fit the data ( p > 0.10). Absolute Risk Reduction (ARR) calculations were made after calculating the number needed to treat (NNT) based on the adjusted odds ratios (ORs),15 using the formula: NNT = (1-[PEER*(1-OR)]) / ([1-PEER]*[PEER]*[1-OR]). PEER is the patient expected event rate (the incidence of PNI in the control [no-restraint use] group). ARR confidence intervals (CIs) were calculated as follows: ARR – 1.96 SQRT (PEER*[1-PEER]/ number control patients + EER* [1-EER]/ number experimental patients). EER is the experimental event rate—in our case, the incidence of PNI in the treatment (use of restraints) group. This study, based on 271,099 patients involved in motor vehicle collisions using protective devices and 113,440 patients not using protective devices has sufficient power (90%) at a 5% type a error rate to detect differences in incidence of PNI as small as 0.8% (corresponding to a number needed to treat as large as 1/0.008 = 125). All probability values are the results of two-sided tests, and the level of significance was set at p < 0.05. Results Patients From 2009 to 2011, there were 384,539 patients (Fig. 1) involved in MVCs who were registered in NTDB and met the inclusion criteria for the study. Of these patients, 271,099 were using restraints and 113,440 were not using restraints. The respective distribution of exposure variables between the two groups can be found in Tables 1A and 1B. Patients not using restraints were associated with significantly worse initial neurologic status, higher ISS, higher HR, lower SBP, and higher incidence of alcohol use. We further adjusted for these differences with regression analyses. Primary outcome In the 271,099 patients using restraints, PNI developed in 1974 (0.73%). On the contrary, from the 113,440 patients not using restraints, 1112 (0.98%) had a diagnosis of PNI. Table 2 demonstrates the incidence of nerve injuries in various subgroups. Overall, plexus injuries developed in 826 patients (722 in the brachial plexus, 91 in the lumbosacral plexus, and 13 unspecified) and injuries distal to the plexus developed in 2260 (1417 in the upper
OR (95% CI) Unadjusted 0.74 (0.69–0.79) Logistic 0.89 (0.82–0.96) regression
p value
ARR (95% CI)
< 0.0001 25.30 (25.24–25.37) 0.005 10.68 (10.61–10.75)
OR, odds ratio; CI, confidence interval; ARR, Absolute Risk Reduction.
extremity, 711 in the lower extremity, and 132 unspecified). Use of protective devices was associated (Table 3) with decreased unadjusted incidence of PNI in comparison with no use (OR, 0.74; 95% CI, 0.69–0.79). This relationship remained after logistic regression analysis (Fig. 2) (OR, 0.89; 95% CI, 0.82–0.96; ARR, 10.67%). This ARR corresponded to 16 patients who needed to be restrained to prevent one PNI. Other factors that are crucial in the outcomes of MVCs were also examined. The location of the patient in the vehicle did not demonstrate an association, with drivers having no difference in the incidence of PNI in comparison with non-drivers (OR, 0.94; 95% CI, 0.87–1.02). On the contrary, alcohol consumption was associated with a higher rate of nerve injuries (OR, 1.10; 95% CI, 1.01–1.20). Discussion Using a comprehensive national database, we demonstrated that the use of restraints is associated with a decreased incidence of PNI in victims of MVCs. Traumatic PNI have been reported16 to occur in up to 5% of patients in a level I trauma center, most of them in the setting of multi-trauma. About half of these patients are the drivers or passengers of motor vehicles.17,18 Although these injuries do not contribute in immediate mortality, they can impact the long-term quality of life of the affected persons.19 They generally occur in the most productive age group,19 resulting in major social and economic burden. Functional disability and neuropathic pain associated with nerve injuries can be devastating.19 Their repair frequently necessitates complicated operations, with variable functional improvement.20 Despite their negative impact on longterm population health, the prevention of PNI during MVCs has not been investigated. Several small series from the United States and other countries have described the etiology and extent of traumatic nerve injuries.6,9,17,19,21,22 They have recognized the significant contribution of MVCs in PNI. In addition, these analyses confirm the debilitating nature of the injuries, but do not focus on their potential prevention. On the contrary, McGovern and associates23 investigated the association of air bag use with the incidence of extremity injuries during MVCs. Although they concluded that air bags were associated with more extremity injuries, there was no specific mention of PNI. Their analysis was not methodologically sound, however, because it introduced significant survivorship bias and lacked regression techniques to adjust for confounders. These limitations are addressed in the present study. Alcohol and position of the patient in the motor vehicle have been recognized as significant contributing factors to MVC morbidity and mortality.2,24,25 Being in the driver’s seat was not associated with the incidence of PNI in our cohort. It appears that the effect of restraints was more significant than the relative position of the patient in the
MVCS: RESTRAINTS AND PERIPHERAL NERVE INJURIES
1081
FIG. 2. Multivariate regression model demonstrating the association of the incidence of peripheral nerve injuries with the exposure variables. The corresponding forest plot is displayed on the right.
motor vehicle. On the contrary, alcohol use was associated with more nerve injuries. This correlation is most likely secondary to the higher speed and more severe nature of MVCs involving alcohol. It appears that all insurance types were associated with higher rates of PNI in comparison with Medicare. This is probably the result of the fact that most Medicare beneficiaries are more than 65 years old. This population is more likely to demonstrate safe behaviors, like use of restraints, and avoid reckless high speed driving, preventing the development of PNI. Although our results are not expected to alter the policies regarding the use of restraints, they add to the large body of literature for injuries that can be prevented with the use of protective devices, even in non-fatal crashes. The potential detrimental functional and financial burden of PNI for the society and the individual underline the importance of the efforts2 of the National Highway Traffic Safety Administration to strictly regulate the use of restraints. The present study has several limitations. First, indication bias and residual confounding could account for some of the observed associations. Second, some coding inaccuracies will undoubtedly occur and can affect our estimates. This is no different from other studies involving the NTDB. Third, the NTDB during the years studied did not include all trauma centers from all states. It is expected to preferentially involve larger hospitals, as well as
younger and more severely injured patients.10 Although this introduces a selection bias, the NTDB is the most extensive trauma registry available. Fourth, there is a large number of missing data for some variables in the database. The values missing are not expected to be random.26 We addressed this issue by excluding variables with more than 20% missing values, as previously established for NTDB,27 and applying multiple imputation methods for the remaining missing data.28 This is still inferior to using real data, however. Fifth, the NTDB is lacking information on the speed of the MVC and the location of the passenger in the front or the back seat of the car, and therefore we could not control for this potential confounder.24 In addition, we did not have information on the size and type of the motor vehicle, preventing further analysis of these subgroups. Future studies should identify the possible effect of these factors on the development of PNI. Sixth, the NTDB does not provide any information on the post-acute course of the patients. Seventh, we used largely ecologic data, and therefore causality cannot be established solely based on them. Conclusions The effect of protective device use on the incidence of PNI remains elusive. We investigated this potential association using regression models. Among patients involved in MVCs, restraint use
1082 was associated with decreased risk of PNI, after controlling for known confounders. Author Disclosure Statement No competing financial interests exist. References 1. National Highway Traffic Safety Administration. 2004. Traffic Safety Facts 2004: A Compilation of Motor Vehicle Crash Data from the Fatality Analysis Reporting System and the General Estimates System. DOT HS 809 919. U.S. Department of Transportation: Washington, DC. 2. Subramanian R. (2005). Motor Vehicle Traffic Crashes as a Leading Cause of Death in the United States, 2002. Traffic Safety Facts Research Note. DOT HS 809 831. National Traffic Safety Administration, National Center for Statistics and Analysis: Washington, DC. 3. Williams, S.B., Whitlock, E.P., Edgerton, E.A., Smith, P.R., Beil, T.L.; U.S. Preventive Services Task Force. (2007). Counseling about proper use of motor vehicle occupant restraints and avoidance of alcohol use while driving: a systematic evidence review for the U.S. Preventive Services Task Force. Ann. Intern. Med. 147, 194–206. 4. Kahane CJ. (2000). Fatality reduction by safety belts for front-seat occupants of cars and light trucks: updates and expanded estimates based on 1986–1999. Department of Transportation. National Highway Traffic Safety Administration: Washington, DC. 5. Carter, P.R., and Maker, V.K. (2010). Changing paradigms of seat belt and air bag injuries: what we have learned in the past 3 decades. J. Am. Coll. Surg. 210, 240–252. 6. Coert, J.H., and Dellon, A.L. (1994). Peripheral nerve entrapment caused by motor vehicle crashes. J. Trauma 37, 191–194. 7. Lad, S.P., Nathan, J.K., Schubert, R.D., and Boakye, M. (2010). Trends in median, ulnar, radial, and brachioplexus nerve injuries in the United States. Neurosurgery 66, 953–960. 8. Noble, J., Munro, C.A., Prasad, V.S., and Midha, R. (1998). Analysis of upper and lower extremity peripheral nerve injuries in a population of patients with multiple injuries. J. Trauma 45, 116–122. 9. Taylor, C.A., Braza, D., Rice, J.B., and Dillingham, T. (2008). The incidence of peripheral nerve injury in extremity trauma. Am. J. Phys. Med. Rehabil. 87, 381–385. 10. American College of Surgeons. (2011). National Trauma Data Bank: NTDB Research Data Set Admission Year 2011, Annual Report. American College of Surgeons: Chicago, IL. 11. Haider, A.H., Saleem, T., Leow, J.J., Villegas, C.V., Kisat, M., Schneider, E.B., Haut, E.R., Stevens, K.A., Cornwell, E.E., III, MacKenzie, E.J., and Efron, D.T. (2012). Influence of the National Trauma Data Bank on the study of trauma outcomes: is it time to set research best practices to further enhance its impact? J. Am. Coll. Surg. 214, 756–768. 12. Honaker, J., King, G., and Blackwell, M. (2009). AMELIA II: A Program for Missing Data. 13. Imai, K., King, G., and Lau, O. (2008). Toward a common framework for statistical analysis and development. J. Comput. Graph. Stat. 17:892–913. 14. Imai, K., King, G., and Lau, O. (2009). Zelig: Everyone’s Statistical Software.
BEKELIS ET AL. 15. Lindenauer, P.K., Pekow, P., Wang, K., Gutierrez, B., and Benjamin, E.M. (2004). Lipid-lowering therapy and in-hospital mortality following major noncardiac surgery. JAMA 291, 2092–2099. 16. Robinson, L.R. (2000). Traumatic injury to peripheral nerves. Muscle Nerve 23, 863–873. 17. Kouyoumdjian, J.A. (2006). Peripheral nerve injuries: a retrospective survey of 456 cases. Muscle Nerve 34, 785–788. 18. Eser, F., Aktekin, L.A., Bodur, H., and Atan, C. (2009). Etiological factors of traumatic peripheral nerve injuries. Neurol. India 57, 434–437. 19. Ciaramitaro, P., Mondelli, M., Logullo, F., Grimaldi, S., Battiston, B., Sard, A., Scarinzi, C., Migliaretti, G., Faccani, G., Cocito, D.; Italian Network for Traumatic Neuropathies. (2010). Traumatic peripheral nerve injuries: epidemiological findings, neuropathic pain and quality of life in 158 patients. J. Peripher. Nerv. Syst. 15, 120–127. 20. Dubuisson, A.S., and Kline, D.G. (2002). Brachial plexus injury: a survey of 100 consecutive cases from a single service. Neurosurgery 51, 673–683. 21. Stone, L., and Keenan, M.A. (1988). Peripheral nerve injuries in the adult with traumatic brain injury. Clin. Orthop. Relat. Res. 233, 136– 144. 22. Midha, R. (1997). Epidemiology of brachial plexus injuries in a multitrauma population. Neurosurgery 40, 1182–1189. 23. McGovern, M.K., Murphy, R.X. Jr., Okunski, W.J., and Wasser, T.E. (2000). The influence of air bags and restraining devices on extremity injuries in motor vehicle collisions. Ann. Plast. Surg. 44, 481–485. 24. Siegel, J.H., Loo, G., Dischinger, P.C., Burgess, A.R., Wang, S.C., Schneider, L.W., Grossman, D., Rivara, F., Mock, C., Natarajan, G.A., Hutchins, K.D., Bents, F.D., McCammon, L., Leibovich, E., and Tenenbaum, N. (2001). Factors influencing the patterns of injuries and outcomes in car versus car crashes compared to sport utility, van, or pick-up truck versus car crashes: Crash Injury Research Engineering Network Study. J. Trauma 51, 975–990. 25. Waller, P.F. (2002). Challenges in motor vehicle safety. Annu. Rev. Public Health 23, 93–113. 26. Roudsari, B., Field, C., and Caetano, R. (2008). Clustered and missing data in the US National Trauma Data Bank: implications for analysis. Inj. Prev. 14, 96–100. 27. Galvagno, S.M., Jr., Haut, E.R., Zafar, S.N., Millin, M.G., Efron, D.T., Koenig, G.J., Jr., Baker, S.P., Bowman, S.M., Pronovost, P.J., and Haider, A.H. (2012). Association between helicopter vs ground emergency medical services and survival for adults with major trauma. JAMA 307, 1602–1610. 28. Oyetunji, T.A., Crompton, J.G., Ehanire, I.D., Stevens, K.A., Efron, D.T., Haut, E.R., Chang, D.C,. Cornwell, E.E., III, Crandall, M.L., and Haider, A.H. (2011). Multiple imputation in trauma disparity research. J. Surg. Res. 165, e37–e41.
Address correspondence to: Kimon Bekelis, MD Dartmouth-Hitchcock Medical Center One Medical Center Drive Lebanon, NH 03756 E-mail: [email protected]
Original Articles
Restraints and Peripheral Nerve Injuries in Adult Victims of Motor Vehicle Crashes Kimon Bekelis,1,* Symeon Missios,2,* and Robert J. Spinner,3–5
Abstract
The pattern of injuries in restrained victims of motor vehicle crashes (MVCs) remains an issue of debate. We investigated the association of peripheral nerve injuries with the use of protective devices (seat belt and air bag) during MVCs. We performed a retrospective cohort study of 384,539 adult MVC victims who were registered in the National Trauma Data Bank (NTDB) between 2009 and 2011. Regression techniques were used to investigate the association of restraint use with the risk of peripheral nerve injury in patients hospitalized after an MVC. Of the study patients, 271,099 were using restraints and 113,440 were not. Overall, there were a total of 3086 peripheral nerve injuries. Multivariable logistic regression analysis demonstrated an association of protective device use with decreased risk of peripheral nerve injury (odds ratio [OR], 0.89; 95% confidence interval [CI], 0.82– 0.96; absolute risk reduction, 10.68%). This corresponds to 16 patients who needed to be restrained to prevent one nerve injury. The location of the patient in the vehicle did not seem to affect the risk of peripheral nerve injury, with drivers demonstrating no association with nerve injuries (OR, 0.94; 95% CI, 0.87–1.02) in comparison with non-drivers. On the contrary, alcohol consumption was associated with increased incidence of peripheral nerve injuries (OR, 1.10; 95% CI, 1.01–1.20). In summary, restraint use was associated with decreased risk of peripheral nerve injury in MVC victims, after controlling for confounders. Key words: air bag; motor vehicle crash; NTDB; peripheral nerve injury; seat belt
Introduction
M
otor vehicle crashes (MVCs) are the leading cause of death among young adults in the United States.1–3 In addition, they contribute to significant morbidity for nearly 3 million patients who sustain non-lethal injuries annually.1–3 Laws mandating the use of active (seat belts) and passive (air bags) restraints while penalizing drunken driving have had a major impact on the reduction of MVC-related fatalities in recent years.3,4 Despite widespread regulation and overall increases in safer behaviors, crash data1 show that more than 50% of fatalities were among unrestrained occupants and nearly 40% involved alcohol. Besides increasing MVC survival,5 restraints have decreased the incidence of serious injuries, including brain and facial trauma, intra-abdominal solid organ injuries, and long bone fractures.5 A new group of injuries, however, has been identified related to the use of restraints.5 These include hollow viscous injuries, aortic tears, abdominal wall hernias, and neck trauma. The association of peripheral nerve injuries (PNI) with the use of supplemental restraints during MVCs remains elusive. Although PNI are a major contributor to incapacitating pain and disability, their prevention in
MVCs has not been studied. This pathology is extremely rare, and often under-recognized during the initial stages of the accident. Most studies describing the phenomenon involve retrospective analyses of single institution or regional experiences,6–9 demonstrating results with limited generalization, given their inherent selection bias. These reports, however, have not examined the association of PNI with the use of restraints. We hypothesized that protective devices (seat belt and air bag) are associated with decreased risk of PNI for victims of MVCs. We tested this hypothesis using a regression model in a cohort of patients who were involved in MVCs and whose data were recorded in the 2009–2011 National Trauma Databank (NTDB). The advantage of using a comprehensive database, such as the NTDB, is that it provides an adequate event volume to study a rare entity, such as PNI. Methods NTDB All patients with MVCs who were registered in NTDB (Committee on Trauma, American College of Surgeons) between 2009
1
Section of Neurosurgery, Dartmouth-Hitchcock Medical Center, Lebanon, New Hampshire. Department of Neurosurgery, Cleveland Clinic, Cleveland, Ohio. Departments of 3Neurosurgery and 4Orthopedics, Mayo Clinic, Rochester, Minnesota. 5 Department of Anatomy, Mayo Medical School, Rochester, Minnesota. *The first two authors contributed equally. 2
1077
1078
BEKELIS ET AL.
and 2011 were included in the analysis. The NTDB is the largest collection of US trauma registry data ever assembled. It includes data collected from more than 900 institutions in the United States.10 We merged the available data sets for the variables of interest, using source codes provided by the NTDB.11 More information about the NTDB is available at http://www.facs.org/ trauma/ntdb/. Cohort definition To establish the cohort of patients, we used International Classification of Diseases, Ninth Revision (ICD-9) E-codes to identify patients in the registry who were involved in MVCs. E-codes 810, 810.1, 810.8, 811, 811.1, 811.8, 812, 812.1, 812.4, 812.8, 813, 813.1, 813.4, 813.8, 814, 814.1, 814.4, 814.8, 815, 815.1, 815.4, 815.8, 816, 816.1, 816.4, 816.8, 819, 819.1, 819.4, 819.8, 820, 820.1, 820.8, 821, 821.1, 821.4, 821.8, 822, 822.1, 822.8, 823, 823.1, 823.4, 823.8, 824, 824.1, 824.4, 824.8, 825, 825.1, 825.8 were eligible for inclusion. Patients older than 16 years who were drivers or passengers involved in MVCs were included for analysis. Exclusion criteria consisted of patients who were dead on arrival to the emergency department (ED) and incomplete record information on the use of restraints and age. Patients involved in motorcycle accidents were not included in the cohort creation. Primary outcome and exposure variables The primary outcome variable was the incidence of peripheral nerve injury. This was defined as an ICD-9 code of 953.x, 954.x, 955.x, 956.x, and 957.x. The effect on the outcome of the pertinent exposure variables was examined. Age, systolic blood pressure (SBP), heart rate (HR), temperature on arrival to the ED, sex, race (African American, Hispanic, Asian, Caucasian, American Indian,
FIG. 1.
and other), insurance (private insurance, self-pay, Medicaid, and Medicare, automobile, and other), use of protective devices (victims wearing their seat belts, cars equipped with air bags that deployed during the crash), location of the victim in the vehicle (driver, passenger), and alcohol use were categorical variables. Glasgow Coma Scale (GCS) score and the patient’s Injury Severity Score (ISS) were ordinal variables. The hospital characteristics used in the analysis as categorical variables included hospital region (West, South, Midwest, North), hospital bed size ( < 200, 200– 400, 400–600 and > 600 beds), and teaching status of the hospital (university, nonteaching, community). Variables were included in the final models after ensuring that the proportion of missing data points was less than 20%. Illicit drug use, medical comorbidities, respiratory rate, O2 saturation, total elapsed Emergency Medical Services (EMS) time from dispatch to ED, EMS time at the scene, and patient days in the intensive care unit and/or on a ventilator were excluded because they were found to have a prevalence of missing data greater than 20%. Statistical analysis The effect of the exposure variables on the primary outcome was examined using a multivariable logistic regression model. Multiple imputation was performed for each variable associated with missing values using the Amelia II package12 in the 64-bit version of R 3.0.2 (R Foundation for Statistical Computing). Imputation was used for the following missing data: sex, payer source, race, ISS score, SBP, HR, temperature, GCS score, use of alcohol, and region. First, the proportion of missing data for variables of interest was calculated. The Amelia II program was used to impute missing data based on the other available variables. This process was repeated five times, creating five separate imputed data sets. These five data sets were combined to create a full-pooled data set
Cohort selection for the study.
Table 1A. Patient and Hospital Characteristics All patients Sample size
384,539 Mean 40.45 N
Age Sex F M Unreported data Region North South Midwest West Unreported data Payer Medicare Medicaid Private payer Self-payer Automobile Other Unreported data Race Caucasian African American Hispanic Asian American Indian Other Unreported data Hospital status Community Non-teaching University Bed size < 200 200–400 400–600 > 600 Number of neurosurgeons Number of trauma surgeons
Protective devices used
SD 19.33 %
271,099 Mean 42.25 N
164,292 220,126
42.74 57.26 121
128,383 142,639 77
47.37 52.63
55,293 152,620 91,699 83,544
14.43 39.83 23.93 21.80 1383
37,496 105,954 64,617 62,214 818
27,265 30,660 109,607 63,199 69,610 41,280
7.98 8.97 32.08 18.50 20.38 12.08 42,918
260,549 49,170 41,753 7313 3344 11,028
No protective devices used 113,440 Mean 36.15 N
SD 19.89 %
p value
SD 17.17 %
< 0.0001
35,909 77,487 44
31.67 68.33
< 0.0001 < 0.0001
13.87 39.20 23.91 23.02
17,797 46,666 27,082 21,330 565
15.77 41.34 23.99 18.90
< 0.0001 < 0.0001 0.80
20,923 18,880 78,151 41,576 53,261 29,321 28,987
8.64 7.80 32.28 17.17 22.00 12.11
6342 11,780 31,456 21,623 16,349 11,959 13,930
6.37 11.84 31.61 21.73 16.43 12.02
< 0.0001 < 0.0001 < 0.0001 < 0.0001 < 0.0001 0.012
69.82 13.18 11.19 1.96 0.90 2.96 11382
179,753 37,147 30,199 6227 1613 8448 7712
68.25 14.10 11.47 2.36 0.61 3.21
80,796 12,023 11,554 1086 1731 2580 3670
73.60 10.95 10.53 0.99 1.58 2.35
< 0.0001 < 0.0001 < 0.0001 < 0.0001 < 0.0001 < 0.0001
140,665 56,507 187,367
36.58 14.69 48.73
99,470 41,693 129,936
36.69 15.38 47.93
41,195 14,814 57,431
36.31 13.06 50.63
0.027 < 0.0001 < 0.0001
24,033 97,160 117,326 146,020 5.3 6.5
6.25 25.27 30.51 37.97 3.2 2.7
17,665 67,204 83,639 102,591 5.2 6.5
6.52 24.79 30.85 37.84 3.2 2.7
6368 29,956 33,687 43,429 5.3 6.6
5.61 26.41 29.70 38.28 3.1 2.8
< 0.0001 < 0.0001 < 0.0001 0.010 < 0.0001 < 0.0001
SD, standard deviation.
Table 1B. Patient Characteristics All patients
SBP mm Hg GCS score HR beats/min Temperature oC Driver Alcohol use 1–24 24–49 50–75
No protective devices used
Mean
SD
Mean
SD
Mean
SD
p value
137.0 14.0 91.8 36.4 N
25.4 2.9 19.6 2.3 %
137.9 14.3 91.1 36.4 N
25.3 2.4 19.1 2.1 %
134.9 13.3 93.7 36.3 N
20.6 3.7 20.6 2.6 %
< 0.0001 < 0.0001 < 0.0001 < 0.0001
290,772 78,109
75.62 22.57 38468 0.80 0.91 0.09 0.01
210,165 45,888 28,100 1974 250,619 19,054 1426
77.52 18.88
80,607 32,221 10,368 1112 98,124 14,085 1231
71.06 31.26
< 0.0001 < 0.0001
0.98 86.50 12.42 1.09
< 0.0001 < 0.0001 < 0.0001 < 0.0001
Unreported data Peripheral nerve injuries Injury Severity Score
Protective devices used
3086 348,743 33,139 2657
0.73 92.45 7.03 0.53
SD, standard deviation; SBP, systolic blood pressure; GCS, Glasgow Coma Scale; HR, heart rate.
1079
1080
BEKELIS ET AL.
Table 2. Incidence of Peripheral Nerve Injuries in Different Subgroups Group + Alcohol use + Use of protective devices + Driver
Patients with peripheral injuries 2271/290,772 815/93,767 758/78,109 2031/267,962 1974/271,099 1112/113,440
%
Table 3. Models Demonstrating the Association of Protective Device Use with the Incidence of Peripheral Nerve Injury
p value
0.78 0.009 0.87 0.97 < 0.0001 0.76 0.73 < 0.0001 0.98
with no missing values, which was used in a multinomial logistic regression (logit) model using the Zeilig statistical package13,14 in the 64-bit version of R 3.0.2. The method of multivariable logistic regression was used to determine the association of use of restraints with the incidence of PNI controlling for: age, sex, race, payer, SBP, GCS, HR, temperature, hospital status, hospital bed size, alcohol use, region, and ISS. Regression diagnostics were performed for all logistic regression models. The Hosmer-Lemeshow and Pearson goodness-of-fit tests were used to confirm that the models adequately fit the data ( p > 0.10). Absolute Risk Reduction (ARR) calculations were made after calculating the number needed to treat (NNT) based on the adjusted odds ratios (ORs),15 using the formula: NNT = (1-[PEER*(1-OR)]) / ([1-PEER]*[PEER]*[1-OR]). PEER is the patient expected event rate (the incidence of PNI in the control [no-restraint use] group). ARR confidence intervals (CIs) were calculated as follows: ARR – 1.96 SQRT (PEER*[1-PEER]/ number control patients + EER* [1-EER]/ number experimental patients). EER is the experimental event rate—in our case, the incidence of PNI in the treatment (use of restraints) group. This study, based on 271,099 patients involved in motor vehicle collisions using protective devices and 113,440 patients not using protective devices has sufficient power (90%) at a 5% type a error rate to detect differences in incidence of PNI as small as 0.8% (corresponding to a number needed to treat as large as 1/0.008 = 125). All probability values are the results of two-sided tests, and the level of significance was set at p < 0.05. Results Patients From 2009 to 2011, there were 384,539 patients (Fig. 1) involved in MVCs who were registered in NTDB and met the inclusion criteria for the study. Of these patients, 271,099 were using restraints and 113,440 were not using restraints. The respective distribution of exposure variables between the two groups can be found in Tables 1A and 1B. Patients not using restraints were associated with significantly worse initial neurologic status, higher ISS, higher HR, lower SBP, and higher incidence of alcohol use. We further adjusted for these differences with regression analyses. Primary outcome In the 271,099 patients using restraints, PNI developed in 1974 (0.73%). On the contrary, from the 113,440 patients not using restraints, 1112 (0.98%) had a diagnosis of PNI. Table 2 demonstrates the incidence of nerve injuries in various subgroups. Overall, plexus injuries developed in 826 patients (722 in the brachial plexus, 91 in the lumbosacral plexus, and 13 unspecified) and injuries distal to the plexus developed in 2260 (1417 in the upper
OR (95% CI) Unadjusted 0.74 (0.69–0.79) Logistic 0.89 (0.82–0.96) regression
p value
ARR (95% CI)
< 0.0001 25.30 (25.24–25.37) 0.005 10.68 (10.61–10.75)
OR, odds ratio; CI, confidence interval; ARR, Absolute Risk Reduction.
extremity, 711 in the lower extremity, and 132 unspecified). Use of protective devices was associated (Table 3) with decreased unadjusted incidence of PNI in comparison with no use (OR, 0.74; 95% CI, 0.69–0.79). This relationship remained after logistic regression analysis (Fig. 2) (OR, 0.89; 95% CI, 0.82–0.96; ARR, 10.67%). This ARR corresponded to 16 patients who needed to be restrained to prevent one PNI. Other factors that are crucial in the outcomes of MVCs were also examined. The location of the patient in the vehicle did not demonstrate an association, with drivers having no difference in the incidence of PNI in comparison with non-drivers (OR, 0.94; 95% CI, 0.87–1.02). On the contrary, alcohol consumption was associated with a higher rate of nerve injuries (OR, 1.10; 95% CI, 1.01–1.20). Discussion Using a comprehensive national database, we demonstrated that the use of restraints is associated with a decreased incidence of PNI in victims of MVCs. Traumatic PNI have been reported16 to occur in up to 5% of patients in a level I trauma center, most of them in the setting of multi-trauma. About half of these patients are the drivers or passengers of motor vehicles.17,18 Although these injuries do not contribute in immediate mortality, they can impact the long-term quality of life of the affected persons.19 They generally occur in the most productive age group,19 resulting in major social and economic burden. Functional disability and neuropathic pain associated with nerve injuries can be devastating.19 Their repair frequently necessitates complicated operations, with variable functional improvement.20 Despite their negative impact on longterm population health, the prevention of PNI during MVCs has not been investigated. Several small series from the United States and other countries have described the etiology and extent of traumatic nerve injuries.6,9,17,19,21,22 They have recognized the significant contribution of MVCs in PNI. In addition, these analyses confirm the debilitating nature of the injuries, but do not focus on their potential prevention. On the contrary, McGovern and associates23 investigated the association of air bag use with the incidence of extremity injuries during MVCs. Although they concluded that air bags were associated with more extremity injuries, there was no specific mention of PNI. Their analysis was not methodologically sound, however, because it introduced significant survivorship bias and lacked regression techniques to adjust for confounders. These limitations are addressed in the present study. Alcohol and position of the patient in the motor vehicle have been recognized as significant contributing factors to MVC morbidity and mortality.2,24,25 Being in the driver’s seat was not associated with the incidence of PNI in our cohort. It appears that the effect of restraints was more significant than the relative position of the patient in the
MVCS: RESTRAINTS AND PERIPHERAL NERVE INJURIES
1081
FIG. 2. Multivariate regression model demonstrating the association of the incidence of peripheral nerve injuries with the exposure variables. The corresponding forest plot is displayed on the right.
motor vehicle. On the contrary, alcohol use was associated with more nerve injuries. This correlation is most likely secondary to the higher speed and more severe nature of MVCs involving alcohol. It appears that all insurance types were associated with higher rates of PNI in comparison with Medicare. This is probably the result of the fact that most Medicare beneficiaries are more than 65 years old. This population is more likely to demonstrate safe behaviors, like use of restraints, and avoid reckless high speed driving, preventing the development of PNI. Although our results are not expected to alter the policies regarding the use of restraints, they add to the large body of literature for injuries that can be prevented with the use of protective devices, even in non-fatal crashes. The potential detrimental functional and financial burden of PNI for the society and the individual underline the importance of the efforts2 of the National Highway Traffic Safety Administration to strictly regulate the use of restraints. The present study has several limitations. First, indication bias and residual confounding could account for some of the observed associations. Second, some coding inaccuracies will undoubtedly occur and can affect our estimates. This is no different from other studies involving the NTDB. Third, the NTDB during the years studied did not include all trauma centers from all states. It is expected to preferentially involve larger hospitals, as well as
younger and more severely injured patients.10 Although this introduces a selection bias, the NTDB is the most extensive trauma registry available. Fourth, there is a large number of missing data for some variables in the database. The values missing are not expected to be random.26 We addressed this issue by excluding variables with more than 20% missing values, as previously established for NTDB,27 and applying multiple imputation methods for the remaining missing data.28 This is still inferior to using real data, however. Fifth, the NTDB is lacking information on the speed of the MVC and the location of the passenger in the front or the back seat of the car, and therefore we could not control for this potential confounder.24 In addition, we did not have information on the size and type of the motor vehicle, preventing further analysis of these subgroups. Future studies should identify the possible effect of these factors on the development of PNI. Sixth, the NTDB does not provide any information on the post-acute course of the patients. Seventh, we used largely ecologic data, and therefore causality cannot be established solely based on them. Conclusions The effect of protective device use on the incidence of PNI remains elusive. We investigated this potential association using regression models. Among patients involved in MVCs, restraint use
1082 was associated with decreased risk of PNI, after controlling for known confounders. Author Disclosure Statement No competing financial interests exist. References 1. National Highway Traffic Safety Administration. 2004. Traffic Safety Facts 2004: A Compilation of Motor Vehicle Crash Data from the Fatality Analysis Reporting System and the General Estimates System. DOT HS 809 919. U.S. Department of Transportation: Washington, DC. 2. Subramanian R. (2005). Motor Vehicle Traffic Crashes as a Leading Cause of Death in the United States, 2002. Traffic Safety Facts Research Note. DOT HS 809 831. National Traffic Safety Administration, National Center for Statistics and Analysis: Washington, DC. 3. Williams, S.B., Whitlock, E.P., Edgerton, E.A., Smith, P.R., Beil, T.L.; U.S. Preventive Services Task Force. (2007). Counseling about proper use of motor vehicle occupant restraints and avoidance of alcohol use while driving: a systematic evidence review for the U.S. Preventive Services Task Force. Ann. Intern. Med. 147, 194–206. 4. Kahane CJ. (2000). Fatality reduction by safety belts for front-seat occupants of cars and light trucks: updates and expanded estimates based on 1986–1999. Department of Transportation. National Highway Traffic Safety Administration: Washington, DC. 5. Carter, P.R., and Maker, V.K. (2010). Changing paradigms of seat belt and air bag injuries: what we have learned in the past 3 decades. J. Am. Coll. Surg. 210, 240–252. 6. Coert, J.H., and Dellon, A.L. (1994). Peripheral nerve entrapment caused by motor vehicle crashes. J. Trauma 37, 191–194. 7. Lad, S.P., Nathan, J.K., Schubert, R.D., and Boakye, M. (2010). Trends in median, ulnar, radial, and brachioplexus nerve injuries in the United States. Neurosurgery 66, 953–960. 8. Noble, J., Munro, C.A., Prasad, V.S., and Midha, R. (1998). Analysis of upper and lower extremity peripheral nerve injuries in a population of patients with multiple injuries. J. Trauma 45, 116–122. 9. Taylor, C.A., Braza, D., Rice, J.B., and Dillingham, T. (2008). The incidence of peripheral nerve injury in extremity trauma. Am. J. Phys. Med. Rehabil. 87, 381–385. 10. American College of Surgeons. (2011). National Trauma Data Bank: NTDB Research Data Set Admission Year 2011, Annual Report. American College of Surgeons: Chicago, IL. 11. Haider, A.H., Saleem, T., Leow, J.J., Villegas, C.V., Kisat, M., Schneider, E.B., Haut, E.R., Stevens, K.A., Cornwell, E.E., III, MacKenzie, E.J., and Efron, D.T. (2012). Influence of the National Trauma Data Bank on the study of trauma outcomes: is it time to set research best practices to further enhance its impact? J. Am. Coll. Surg. 214, 756–768. 12. Honaker, J., King, G., and Blackwell, M. (2009). AMELIA II: A Program for Missing Data. 13. Imai, K., King, G., and Lau, O. (2008). Toward a common framework for statistical analysis and development. J. Comput. Graph. Stat. 17:892–913. 14. Imai, K., King, G., and Lau, O. (2009). Zelig: Everyone’s Statistical Software.
BEKELIS ET AL. 15. Lindenauer, P.K., Pekow, P., Wang, K., Gutierrez, B., and Benjamin, E.M. (2004). Lipid-lowering therapy and in-hospital mortality following major noncardiac surgery. JAMA 291, 2092–2099. 16. Robinson, L.R. (2000). Traumatic injury to peripheral nerves. Muscle Nerve 23, 863–873. 17. Kouyoumdjian, J.A. (2006). Peripheral nerve injuries: a retrospective survey of 456 cases. Muscle Nerve 34, 785–788. 18. Eser, F., Aktekin, L.A., Bodur, H., and Atan, C. (2009). Etiological factors of traumatic peripheral nerve injuries. Neurol. India 57, 434–437. 19. Ciaramitaro, P., Mondelli, M., Logullo, F., Grimaldi, S., Battiston, B., Sard, A., Scarinzi, C., Migliaretti, G., Faccani, G., Cocito, D.; Italian Network for Traumatic Neuropathies. (2010). Traumatic peripheral nerve injuries: epidemiological findings, neuropathic pain and quality of life in 158 patients. J. Peripher. Nerv. Syst. 15, 120–127. 20. Dubuisson, A.S., and Kline, D.G. (2002). Brachial plexus injury: a survey of 100 consecutive cases from a single service. Neurosurgery 51, 673–683. 21. Stone, L., and Keenan, M.A. (1988). Peripheral nerve injuries in the adult with traumatic brain injury. Clin. Orthop. Relat. Res. 233, 136– 144. 22. Midha, R. (1997). Epidemiology of brachial plexus injuries in a multitrauma population. Neurosurgery 40, 1182–1189. 23. McGovern, M.K., Murphy, R.X. Jr., Okunski, W.J., and Wasser, T.E. (2000). The influence of air bags and restraining devices on extremity injuries in motor vehicle collisions. Ann. Plast. Surg. 44, 481–485. 24. Siegel, J.H., Loo, G., Dischinger, P.C., Burgess, A.R., Wang, S.C., Schneider, L.W., Grossman, D., Rivara, F., Mock, C., Natarajan, G.A., Hutchins, K.D., Bents, F.D., McCammon, L., Leibovich, E., and Tenenbaum, N. (2001). Factors influencing the patterns of injuries and outcomes in car versus car crashes compared to sport utility, van, or pick-up truck versus car crashes: Crash Injury Research Engineering Network Study. J. Trauma 51, 975–990. 25. Waller, P.F. (2002). Challenges in motor vehicle safety. Annu. Rev. Public Health 23, 93–113. 26. Roudsari, B., Field, C., and Caetano, R. (2008). Clustered and missing data in the US National Trauma Data Bank: implications for analysis. Inj. Prev. 14, 96–100. 27. Galvagno, S.M., Jr., Haut, E.R., Zafar, S.N., Millin, M.G., Efron, D.T., Koenig, G.J., Jr., Baker, S.P., Bowman, S.M., Pronovost, P.J., and Haider, A.H. (2012). Association between helicopter vs ground emergency medical services and survival for adults with major trauma. JAMA 307, 1602–1610. 28. Oyetunji, T.A., Crompton, J.G., Ehanire, I.D., Stevens, K.A., Efron, D.T., Haut, E.R., Chang, D.C,. Cornwell, E.E., III, Crandall, M.L., and Haider, A.H. (2011). Multiple imputation in trauma disparity research. J. Surg. Res. 165, e37–e41.
Address correspondence to: Kimon Bekelis, MD Dartmouth-Hitchcock Medical Center One Medical Center Drive Lebanon, NH 03756 E-mail: [email protected]
