- Email: [email protected]
International comparison of prehospital trauma care systems
Injury, Int. J. Care Injured (2007) 38, 993—1000
www.elsevier.com/locate/injury
International comparison of prehospital trauma care systems Bahman S. Roudsari a,*, Avery B. Nathens b, Peter Cameron c, Ian Civil d, Russel L. Gruen e, Thomas D. Koepsell f,g, Fiona E. Lecky h, Rolf L. Lefering i, Moishe Liberman j, Charles N. Mock k,f, ¨rg Oestern l, Thomas A. Schildhauer m, Hans-Jo Christian Waydhas n, Frederick P. Rivara o,f a
Department of Epidemiology, University of Texas, School of Public Health, Dallas, USA St. Michael’s Hospital, Toronto, Canada c Department of Epidemiology and Preventive Medicine, Monash School of Public Health, Alfred Hospital, Australia d University of Auckland, Auckland Hospital, Auckland, New Zealand e Department of Health Services Research and Policy, Menzies School of Health Research, Australia f Harborview Injury Prevention and Research Centre, Department of Epidemiology, University of Washington, Seattle, USA g Harborview Injury Prevention and Research Centre, Department of Health Services, University of Washington, Seattle, USA h University of Manchester, Department of Emergency Medicine, Manchester, United Kingdom i Biochemical & Experimental Division, University of Cologne, Cologne, Germany j Montreal General Hospital, McGill University Health Centre, Divisions of Surgery and Clinical Epidemiology, Montreal, Canada k Harborview Injury Prevention and Research Centre, Department of Surgery, University of Washington, Seattle, USA l Chefarzt der Klinik fu ¨ r, Unfall und Wiederherstellungschirurgie, Allgemeines Krankenhaus Celle, Germany m Chirurgische Klinik und Poliklinik, BG Kliniken Bergmannsheil, Ruhr-Universitat Bochum, Bochum, Germany n Klinik und Poliklinik fu ¨ r Unfallchirurgie, Universita ¨tsklinikum Essen, Hufelandstr., Germany o Harborview Injury Prevention and Research Centre, Department of Pediatrics, University of Washington, Seattle, USA b
Accepted 27 March 2007 * Corresponding author. Tel.: +1 214 648 1054; fax: +1 214 648 1081. E-mail addresses: [email protected] (B.S. Roudsari), [email protected] (A.B. Nathens), [email protected] (P. Cameron), [email protected] (I. Civil), [email protected] (R.L. Gruen), [email protected] (T.D. Koepsell), [email protected] (F.E. Lecky), [email protected] (R.L. Lefering), [email protected] (M. Liberman), [email protected] (C.N. Mock), [email protected] (H.-J. Oestern), [email protected] (T.A. Schildhauer), [email protected] (C. Waydhas), [email protected] (F.P. Rivara). 0020–1383/$ — see front matter # 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.injury.2007.03.028
994 KEYWORDS Emergency medical services (EMS); Trauma; Injury; International; Mortality; Clustered study; Shock rate; Intracluster correlation coefficient (ICC)
B.S. Roudsari et al. Summary Objective: Given the recent emphasis on developing prehospital trauma care globally, we embarked upon a multicentre study to compare trauma patients’ outcome within and between countries with technician-operated advanced life support (ALS) and physician-operated (Doc-ALS) emergency medical service (EMS) systems. These environments represent the continuum of prehospital care in high income countries with more advanced prehospital trauma care systems. Methods: Five countries with ALS-EMS system and four countries with Doc-ALS EMS system provided us with de-identified patient-level data from their national or local trauma registries. Generalised linear latent and mixed models was used in order to compare emergency department (ED) shock rate (systolic blood pressure (SBP) <90 mmHg) and early trauma fatality rate (i.e. death during the first 24 h after hospital arrival) between ALS and Doc-ALS EMS systems. Logistic regression was used to compare outcomes of interest among different countries, accounting for withinsystem correlation in patient outcomes. Results: After adjustment for patient age, sex, type and mechanism of injury, injury severity score and SBP at scene, the ED shock rate did not vary significantly between Doc-ALS and ALS systems (OR: 1.16, 95% CI: 0.73—1.91). However, the early trauma fatality rate was significantly lower in Doc-ALS EMS systems compared with ALS EMS systems (OR: 0.70, 95% CI: 0.54—0.91). Furthermore, we found a considerable heterogeneity in patient outcomes among countries even with similar type of EMS systems. Conclusion: These findings suggest that prehospital trauma care systems that dispatch a physician to the scene may be associated with lower early trauma fatality rates, but not necessarily with significantly better outcomes on other clinical measures. The reasons for these findings deserve further studies. # 2007 Elsevier Ltd. All rights reserved.
Introduction Prehospital care at the scene of injury and during transportation to a medical care facility is often provided by emergency medical service (EMS) systems and is the first step in managing the injured patient. Currently there are four different types of prehospital trauma care systems worldwide. In some environments, there is no organisation responsible for providing prehospital care to trauma patients. This is the pattern in most developing countries.39 Basic life support (BLS) EMS systems provide noninvasive supportive care to trauma patients. The major role of emergency medical technicians (EMTs) in these systems is to transport trauma patients rapidly to a medical care facility and to keep them alive during transport. Some developing countries and many small cities and rural areas in developed countries are served by this type of EMS system.39 Advanced life support (ALS) EMS systems provide more sophisticated care. In these systems, paramedics have intensive training programs for performing invasive procedures such as intubation and intravenous fluid therapy. In the most advanced form of prehospital trauma care (Doc-ALS EMS systems), physicians are responsible for providing prehospital care to trauma patients at the scene of injury and
during transport. The primary goal of ALS and DocALS EMS systems is to initiate the highest level of care at scene and during transport. The gold standard for the evaluation of the effectiveness of different levels of prehospital trauma care is a randomised trial. These randomised trials would almost certainly have to be group-randomised, since the exposure of interest, i.e. types of EMS system, varies at the population level and not at the individual level.33 However, ethical concerns and resource and infrastructure limitations preclude such trials. Therefore, observational studies have primarily been used to compare patients’ outcome across different systems of prehospital trauma care. However, these studies often suffer from a number of limitations. First, patients within the catchment area of a particular trauma centre often receive prehospital care from a single EMS system. Therefore for an evaluation of the association between type of EMS and patients’ outcome, other settings/areas with different types of EMS systems should be used as controls.33 This issue requires researchers to design and conduct multicentre studies, taking into consideration the correlation of the observations within centres or systems.1,9—11,14,15,26,33 Second, multicentre studies that evaluate different aspects of trauma care often use the in-hospital trauma fatality rate as
International comparison of prehospital trauma care systems the main outcome of interest.3,31,47 However, this outcome may be heavily influenced by the quality of hospital care rather than prehospital interventions.17,27,28,33,44 Considering these limitations, and given the recent emphasis on developing prehospital trauma care globally,39 we embarked upon a multicentre study to compare trauma patients’ outcome across different countries. We sought to test the null hypothesis that patient outcomes do not differ materially between EMS systems. Any superiority of patient outcomes under Doc-ALS compared to ALS EMS systems might be expected to be more prominent for patients with more serious injuries that require higher level of care and more advanced procedures. Due to limited access to the information from countries with BLS EMS system, the focus of this study was on the comparison of the patients’ outcome between ALS and Doc-ALS EMS systems.
Methods Source of data De-identified patient-level data were obtained from regions in five countries with ALS EMS system and four countries with DOC-ALS EMS system. In order to protect countries’ identity, we refer to these countries as ALS1 to ALS5 and Doc-ALS1 to DOC-ALS4. The trauma registries within different regions had dissimilar inclusion criteria. Therefore, to render the mix of studied patients more comparable across regions, we restricted our analysis to criteria common to all registries. Thus the analyses were limited to patients 15—55 years old, with severe injuries (defined as injury severity score6 (ISS) > 15) and directly transported from the scene of injury to a definitive trauma care facility. This study was approved by the institutional review board or other relevant authorities from each participating country and of the University of Washington.
Patient descriptors Baseline characteristics of subjects across the nine regions were age, gender, mechanism of injury, mode of transportation (ground ambulance, air ambulance, etc.), and whether the patient was transported directly from the scene to a centre providing definitive care or was transferred by means of another health care facility. Injury severity was evaluated using the injury severity score (ISS), Glasgow coma scale21 (GCS), and the presence of shock at the scene of injury (systolic blood pressure (SBP) <90 mmHg). Prehospital information included
995
vital signs at scene, prehospital time intervals and prehospital interventions (intravenous access and endotracheal intubation).
Outcomes of interest We defined the emergency department (ED) shock rate (proportion of patients with SBP <90 mmHg in the ED) and early trauma fatality rate (i.e. death during the first 24 h after hospital arrival) as the outcomes of interest, since they may be mainly influenced by the severity of injury and quality of prehospital care rather than the quality of hospital interventions.
Power calculation For the calculation of power, we used a formula for clustered studies15,29 (Appendix A). It is comprised of two factors, the standard sample size calculation and a ‘‘design effect’’ or ‘‘variance inflation factor’’.15,26,30 The design effect depends on the average number of subjects per cluster and the intracluster correlation coefficient (ICC) for the outcomes of interest.33 An estimate of the ICC came from a separate analysis of patient outcome data from trauma systems in the US.11 Considering an estimated frequency of ED shock in ALS EMS systems of 20%,33 an ICC of shock 0.009833 and an average of 1000 subjects per cluster, we estimated that our study had approximately 90% power for detecting an 8% absolute difference in the rate of shock between ALS and Doc-ALS EMS systems. We used the US National Trauma Databank4 in order to calculate the ICC for early trauma fatality (results not presented here). Accordingly, our study had 80% power for detecting 8% absolute difference in early trauma fatality between ALS and Doc-ALS systems, assuming an 18% early fatality rate in ALS EMS systems and ICC equal to 0.019.33
Data analysis We used generalised linear latent and mixed models (GLLAMM) to estimate the odds ratios for the outcomes of interest comparing ALS EMS (reference group) with Doc-ALS EMS system, adjusting for age, sex, type and mechanism of injury, ISS and SBP at scene. Since the exposure of interest (i.e. type of EMS system) varied only at the country level and not at the individual level, we only included random intercept and not random slope to the GLLAMM model. 32,45 To evaluate the hypothesis that any superiority of Doc-ALS over ALS would be most prominent for more severely injured patients, we conducted a sub-analysis that evaluated the association between type of
996
B.S. Roudsari et al.
EMS and patients’ outcome among patients with ISS > 25 using the GLLAMM model. Due to the differences in registry protocols, patients who were dead-on-arrival (DOA) might be under-represented in a particular database. In order to evaluate the potential influence that this difference in inclusion criteria might have had on our estimates, we conducted two other sub-analyses. First, we limited analyses to patients with a detectable blood pressure at the scene of injury. Later, we implemented the same approach focusing on subjects with a detectable blood pressure at the time of ED arrival. For the second approach, we compared ED shock and early trauma fatality among different countries, using logistic regression analysis. In this approach we treated ‘‘country’’ as a fixed effect variable16 and evaluated the heterogeneity in outcomes of interest within the countries with similar type of EMS system. We adjusted the results for patient age, sex, mechanism and type of injury, injury severity score and SBP at scene as potential confounding or precision variables.22 Doc-ALS3 had the largest sample size in our study. Therefore, we used this country as the baseline for calculation of odds ratios for ED shock and early trauma fatality in logistic regression models. This approach allowed us to estimate the ORs more precisely (i.e. with the narrowest possible confidence intervals).22 Finally, in order to quantify within and between region variability in the outcomes of interest, we estimated the intracluster correlation coefficient (ICC) of ED shock and early trauma fatality rate for ALS and Doc-ALS EMS systems.10,15,29,33 One of the obstacles in using trauma registries is the issue of missing data. Table 1 reflects the proportion of missing values for the main variables. Our evaluations showed that missing pattern was not completely at random (data not presented).36— 38,40—42 It has been shown that when missingness
is not completely at random, using complete case analysis could lead to biased estimates.36—38,40,41 Therefore, we used multiple imputation (MI) using predictive mean matching19,36—38,48 in order to deal with these missing data. Special emphasis was placed on the imputation of systolic blood pressure at the scene and at ED. The proportion of missing SBP at scene varied from 0% in ALS5 database to 58% in Doc-ALS2. We conducted 10 imputations. Since different datasets had different characteristics that could qualify for inclusion in the MI model by the above criteria, each dataset was imputed separately and then the datasets were combined. We included the following variables in the MI model12,46: (a) variables that were used in the final analysis that compared patients’ outcome across ALS and Doc-ALS EMS systems: age, sex, mechanism and type of injury, injury severity score, type of EMS, SBP at scene and at ED and death during the first 24 h of hospital arrival; (b) variables that potentially influenced the occurrence of missing data: mode of transportation (i.e. ground ambulance, air ambulance, etc.) and prehospital time intervals; and (c) variables that were correlated with SBP5,46: length of hospital stay, ICU admission and prehospital and hospital interventions. We used the ICE command written by Royston for STATA (STATA Corporation, College Station, Texas) in order to implement the multiple imputation process.34,35 Each imputed dataset was analysed separately. The results were averaged and calculation of the standard errors was based on Rubin’s formula.36—38,40
Results The mean age of the patients varied from 30 years in ALS2 to 35 years in Doc-ALS1. Men comprised 75% (Doc-ALS1) to 82% (Doc-ALS4) of the patients.
Table 1 Proportion of missing data for the main variables Country
Age
Sex
Type of injury
Mechanism
ISS a
SBPb (scene)
SBPb (EDc)
Outcome
LOS d
ALS1 ALS2 ALS3 ALS4 ALS5 Doc-ALS1 Doc-ALS2 Doc-ALS3 Doc-ALS4
0 0 0 2 0 0 0 1 0
0 0 0 0 0 0 0 0 0
4 0 0 0 0 0 0 1 0
3 0 0 0 0 0 0 0 0
1 1 0 2 1 1 0 15 0
16 18 52 52 0 25 58 22 29
3 4 5 26 0 10 7 8 6
2 0 0 0 0 0 1 0 0
1 1 0 1 0 1 7 8 0
a b c d
Injury severity score. Systolic blood pressure. Emergency department. Length of stay.
International comparison of prehospital trauma care systems Table 2 Measurements of injury severity
ALS1 ALS2 ALS3 ALS4 ALS5 Doc-ALS1 Doc-ALS2 Doc-ALS3 Doc-ALS4
Shock at scene (%)
ISSa (mean S.D.)
18 17 21 10 17 21 27 17 20
26 11 24 9 25 11 26 10 27 13 26 10 30 13 32 13 30 13
997
system. In order to quantify the heterogeneity in outcomes of interest within and between countries with similar type of EMS system, we calculated the ICC of ED shock and early trauma fatality rate.15,33 The ICCs of ED shock for ALS and Doc-ALS EMS systems were 0.007 (95% CI: 0.000—0.018) and 0.013 (0.000—0.038), respectively. The ICCs for early trauma fatality rate were 0.019 (95% CI: 0.000—0.049) for ALS and 0.036 (95% CI: 0.000— 0.101) for Doc-ALS EMS systems.
Discussion
Systolic blood pressure <90 mmHg. a Injury severity score.
Shock at the scene was more common in Doc-ALS2 (27%), Doc-ALS1 (21%) and ALS3 (21%). DOC-ALS3 (32 13), DOC-ALS2 (30 13) and Doc-ALS3 (30 13) reported the highest mean ISS (Table 2). The results of the GLLAMM model showed that after adjustment for patient age, sex, type and mechanism of injury, ISS and SBP at scene, the observed difference in ED shock rate (OR: 1.16, 95% CI: 0.73—1.91) did not reach a statistically significant level. However, early trauma fatality was less common in Doc-ALS EMS systems compared with ALS EMS systems (OR: 0.70, 95% CI: 0.54—0.91). Changes in the ORs after restricting the analyses to patients with more severe injuries (ISS > 25) and detectable SBP at scene or at ED have been summarised in Table 3. This shows that the OR for mortality decreased even further with Doc-ALS compared with ALS for more severe injuries. This pattern pertained whether or not analysis was restricted to detectable SBP. Fig. 1a and b demonstrate the adjusted results of between-countries comparison of outcomes, modeling ‘‘country’’ as a fixed-effect variable in the logistic regression model. As presented, there is a considerable heterogeneity in outcomes of interest even among countries with similar type of EMS
Our report is the first study that aims at a comparison of patients’ outcome within and between ALS and Doc-ALS EMS systems using a broad sample of such systems. The following characteristics differentiate this study from prior related reports. First, we used ED shock and early trauma fatality rate, and not in-hospital trauma fatality rate,2,18,20,23,25,43 as the outcomes of interest. As previously mentioned, the rationale behind this decision was the potential dependency of late hospital trauma fatality on the quality of hospital care rather than prehospital care. Ideally, even earlier assessment of mortality status (within the first hour of ED arrival) would provide greater insight into prehospital care. Unfortunately, trauma registries do not record the time of death with this degree of resolution. Second, when an EMS system adopts a particular model of care (i.e. ALS or Doc-ALS) almost all of the patients within the jurisdiction of that EMS receive that level of care if required. For example, in 1990, 98.5% of the 200 largest cities in the US had ALS EMS system.8 In 82% of these cities, ALS units were present at all dispatches.8 To address the possible consequences of this policy decision, we considered type of EMS as a system-level varying exposure of interest. Accordingly, we designed a clustered study, taking into consideration the correlation of observations within each cluster.33
Table 3 Adjusteda odds ratios of ED shock and early trauma fatality rate, comparing Doc-ALS with ALS EMS system (baseline group) for different levels of injury severity and SBP at scene and at ED OR of ED shock b All patients, regardless of their SBP at scene or at ED Patients with detectable SBP at scene Patients with detectable SBP at ED a b
OR of early trauma fatality a
ISS > 15
ISS > 25
ISS > 15
ISS > 25
1.18 (0.73—1.92)
1.00 (0.72—1.39)
0.70 (0.54—0.91)
0.57 (0.42—0.77)
1.09 (0.82—1.43)
1.00 (0.71—1.41)
0.64 (0.49—0.85)
0.55 (0.40—0.75)
1.03 (0.76—1.46)
1.00 (0.68—1.45)
0.61 (0.46—0.80)
0.53 (0.39—0.73)
Adjusted for age, sex, type and mechanism of injury, injury severity score and SBP at scene. Comparing Doc-ALS EMS system to ALS (reference group).
998
Figure 1 Odds ratio ED shock (a) and of early trauma fatality (b) adjusted for age, sex, mechanism and type of injury, injury severity score and SBP at the scene of injury, comparing different countries with DOC-ALS3 as the baseline for comparison.
To date several studies have evaluated the association between type of prehospital care and patients’ outcome and systematic reviews have nicely summarised the results of those studies.7,23,25 These reports have concluded that the evidence supporting advanced prehospital trauma care is lacking. However, almost all of these studies treated prehospital care (i.e. ALS or BLS) as a patient-level
B.S. Roudsari et al. varying exposure and used hospital trauma fatality as the main outcome of interest. Furthermore, with the exception of a single study,24 no comparison between ALS and Doc-ALS EMS systems has been published. Liberman and his colleagues compared Doc-ALS EMS system in Montreal with ALS EMS system in Toronto.24 They reported that after adjustment for age, injured body region, mechanism of injury and ISS, being treated in the Doc-ALS EMS system was not associated with a reduction in risk of in-hospital death (OR: 1.14, 95% CI: 0.97—1.33).24 We observed considerable heterogeneity in both outcomes even among countries with similar type of EMS system. Nonetheless, while the observed differences in ED shock rate in our study did not reach a statistically significant level, early trauma fatality was significantly less common in Doc-ALS EMS systems. Unfortunately our current data do not allow us to identify those characteristics of Doc-ALS EMS systems that have been associated with lower early trauma fatality rate. However, the magnitude of difference in trauma fatality rate among different countries even with similar types of EMS system was much higher than that can be explained only by the differences in the quality of prehospital care. The differences in mortality between EMS systems may represent a real difference in outcome. However, several measurement issues might also account for the perceived differences. For example, there are differences in registry and EMS protocols in dealing with dead at scene, dead-on-arrival (DOA) and death during the first few minutes after hospital arrival.13 Some trauma registries might not record DOA into their databases which can lead to underestimation of the trauma fatality rate. Pronouncement of death at scene or en route might also vary from system to system. For example, while physicians are qualified to announce a patient ‘‘dead en route’’, EMTs might require a physician’s approval for such a pronouncement. Under these circumstances, ‘‘dead en route’’ subjects might be counted as ED fatalities in an ALS EMS system and as DOA in a Doc-ALS EMS system. Deaths before an adequate survey of injuries might also impact estimation of the ISS, leading to a potential for bias in the types of patients included in the two types of systems’ registries. By the same token, these differences in inclusion—exclusion criteria might influence ED shock rate estimation. However, since only a fraction of patients with shock at ED die during the first few minutes after ED arrival, the magnitude of bias using ED shock as the outcome might be significantly lower. There are some potential limitations to this analysis. First, apart from the number of subjects, the
International comparison of prehospital trauma care systems power of a clustered study is highly dependent on the number of regions or clusters. The small number of the regions restricted our ability to detect small but clinically important differences in distribution of ED shock rate between ALS and Doc-ALS EMS systems, if in reality such differences exist. Second, establishment of comprehensive trauma registries requires long term planning and sufficient financial support. This need can explain why all the collaborating countries in this research project are high income countries with more advanced EMS systems, while no countries with BLS EMS system were represented. Third, although multiple imputation is currently a preferred method for dealing with missing information, it cannot remove all of the uncertainty resulting from the missing data. The method thus yields larger standard deviations and wider confidence intervals than would be expected if there were no missing data. Finally, there were potential differences in registry protocol across countries and EMS systems. In spite of these limitations, our study provided a reasonable basis for comparison of patient outcomes within and between EMS systems. Our findings raise a few important issues. First, there is significant heterogeneity even within a particular type of EMS system. Therefore the ability to generalise across EMS systems is somewhat limited. Second (and closely related), calculation of the ICCs especially for early trauma fatality rate showed a substantial correlation of observations within countries with the same type of EMS system. These large ICCs should be incorporated in sample size calculation of international multicentre studies that use trauma fatality as the main outcome of interest. Third, due to potential differences in registry protocols, study populations should be as comparable as possible. Prospective multicentre studies that follow similar data collection protocols and use uniform inclusion criteria might be the best approach for this purpose.13
Conflicts of interest None of the authors has any conflict of interest with this research project.
Acknowledgements The authors would like to thank the following individuals and organisations for their invaluable support throughout this research project: Dr. Patrick Haegarty, Christopher Mack, Dr. Steve Bowman, Michele Plorde, Victorian State Trauma Registry (Australia), The Quebec Trauma Registry (Canada),
999
German Trauma Registry and German Society of Trauma Surgery (Germany, Austria and the Netherlands), Auckland City Hospital Trauma Registry (New Zealand), Trauma Audit and Research Network and University of Manchester (UK), and King County EMS and Washington State Trauma Registry (USA).
Appendix A
n¼
ðZ a=2 þ Z b Þ2 ½P 1 ð1 P 1 Þ þ P 2 ð1 P 2 Þ ðP 1 P2 Þ2 |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} I
½1 þ ðm0 1Þ ICC |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
(1)
II
References 1. Ahn C, Jung SH, Donner A. Application of an adjusted chi2 statistic to site-specific data in observational dental studies. J Clin Periodontol 2002;29:79—82. 2. Ali JAR, Gana TJ, Bedaysie H, Williams JI. Effect of the prehospital trauma life support program (PHTLS) on prehospital trauma care. J Trauma 1997;42:786—90. 3. American College of Surgeons-Committee on Trauma. Resources for optimal care of the injured patients; 1999. 4. American College of Surgeons COT. NTDB 2004 Fact Sheet; 2004. 5. Arnold AM, Kronmal RA. Multiple imputation of baseline data in the cardiovascular health study. Am J Epidemiol 2003;157:74—84. 6. Baker SP, O’Neill B, Haddon Jr W, Long WB. The injury severity score: a method for describing patients with multiple injuries and evaluating emergency care. J Trauma 1974;14:187—96. 7. Bunn F, Kwan I, Roberts I, R W. Effectiveness of prehospital trauma care. Cochrane Injuries Group; 2001. 8. Cady G, Scott T. EMS in the United States. 1994 survey of providers in the 200 most populous cities. Jems 1994;19:88— 94. 6—7, 100. 9. Campbell M, Thomson S, Ramsay C, et al. Sample size calculator for cluster randomised trials. Comput Biol Med 2004;34:113—25. 10. Campbell MK, Grimshaw JM, Elbourne DR. Intracluster correlation coefficients in cluster randomised trials: empirical insights into how should they be reported. BMC Med Res Methodol 2004;4. 11. Campbell MK, Mollison J, Grimshaw JM. Cluster trials in implementation research: estimation of intracluster correlation coefficients and sample size. Stat Med 2001;20:391—9. 12. Clark TG, Altman DG. Developing a prognostic model in the presence of missing data: an ovarian cancer case study. J Clin Epidemiol 2003;56:28—37. 13. Dick WF, Baskett PJ. Recommendations for uniform reporting of data following major trauma-the Utstein style. A report of a working party of the International Trauma Anaesthesia and Critical Care Society (ITACCS). Resuscitation 1999;42:81—100. 14. Donner A, Klar N. Pitfalls of and controversies in cluster randomization trials. Am J Public Health 2004;94:416—22.
1000 15. Donner A, Klar NS. Design and analysis of cluster randomization trials: in health research. New York: Oxford University Press; 2000. 16. Fitzmaurice GM, Larid NM, Ware J. Applied longitudinal analysis. New Jersey: Wiley; 2004. 17. 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. 18. Gold CR. Prehospital advanced life support vs. ‘‘scoop and run’’ in trauma management. Ann Emerg Med 1987;16:797— 801. 19. Heitjan DF, Little RJA. Multiple imputation for the fatal accident reporting system. Appl Stat 1991;40:13—29. 20. Isenberg DL, Bissell R. Does advanced life support provide benefits to patients?: a literature review. Prehospital Disaster Med 2005;20:265—70. 21. Jennett B, Teasdale G, Braakman R, et al. Predicting outcome in individual patients after severe head injury. Lancet 1976;1:1031—4. 22. Koepsell TD, Weiss NS. Epidemiologic methods: studying the occurrence of illness. Oxford University Press; 2003. 23. Liberman M, Branas C, Mulder D, et al. Advanced versus basic life support in the prehospital setting- the controversy between the scoop and run and the stay and play approach to the care of the injured patient. Int J Disaster Med 2004;2:1—9. 24. Liberman M, Mulder D, Lavoie A, et al. Multicentre Canadian study of prehospital trauma care. Ann Surg 2003;237: 153—60. 25. Liberman M, Mulder D, Sampalis J. Advanced or basic life support for trauma: meta-analysis and critical review of the literature. J Trauma 2000;49:584—99. 26. Localio AR, Berlin JA, Ten Have TR, Kimmel SE. Adjustments for centre in multicentre studies: an overview. Ann Intern Med 2001;135:112—23. 27. Maio RF, Garrison HG, Spaite DW, et al. Emergency medical services outcomes project I (EMSOP I): prioritizing conditions for outcomes research. Ann Emerg Med 1999;33:423—32. 28. Maio RF, Garrison HG, Spaite DW, et al. Emergency medical services outcomes project (EMSOP) IV: pain measurement in out-of-hospital outcomes research. Ann Emerg Med 2002;40:172—9. 29. Murray DM. Design and analysis of group randomised trials. New York, NY: Oxford University Press Inc.; 1998. 30. Murray DM, Varnell SP, Blitstein JL. Design and analysis of group-randomised trials: a review of recent methodological developments. Am J Public Health 2004;94:423—32.
B.S. Roudsari et al. 31. Nathens AB, Jurkovich GJ, Rivara FP, Maier RV. Effectiveness of state trauma systems in reducing injury-related mortality: a national evaluation. J Trauma 2000;48:25—30. 32. Rabe-Hesketh S, Skrondal A, A P. GLLAMM Manual. U.C. Berkeley Division of Biostatistics Working Paper Series; 2004. 33. Roudsari B, Nathens A, Koepsell T, et al. Analysis of clustered data in multicentre trauma studies. Injury 2006. 34. Royston P. Multiple imputation of missing values. Stata J 2004;4:227—41. 35. Royston P. Multiple imputation of missing values:update. Stata J 2005;5:1—14. 36. Rubin DB. Inference and missing data. Biometrika 1976;63: 581—92. 37. Rubin DB. Multiple imputation after 18+ years. J Am Stat Assoc 1996;91:434. 38. Rubin DB, Schenker N. Multiple imputation in health-care databases: an overview and some applications. Stat Med 1991;10:585—98. 39. Sasser S, Varghese M, Kellermann A, JD. L. Prehospital trauma care systems. Geneva: World Health Organisation; 2005. 40. Schafer JL. Multiple imputation: a primer. Stat Meth Med Res 1999;8:3—15. 41. Schafer JL, N S. Inference with imputed conditional means. J Am Stat Assoc 2000 1995; 95, 1144—54. 42. Schiattino I, Villegas R, Cruzat A, et al. Multiple imputation procedures allow the rescue of missing data: an application to determine serum tumor necrosis factor (TNF) concentration values during the treatment of rheumatoid arthritis patients with anti-TNF therapy. Biol Res 2005;38:7—12. 43. Spaite DW, Valenzuela TD, Criss EA, et al. A prospective infield comparison of intravenous line placement by urban and nonurban emergency medical services personnel. Ann Emerg Med 1994;24:209—14. 44. Spaite DW, Maio R, Garrison HG, et al. Emergency medical services outcomes project (EMSOP) II: developing the foundation and conceptual models for out-of-hospital outcomes research. Ann Emerg Med 2001;37:657—63. 45. Twisk JWR. Applied longitudinal data analysis for epidemiology; a practical guide. Cambridge: Cambridge University Press; 2003. 46. van Buuren S, Boshuizen HC, Knook DL. Multiple imputation of missing blood pressure covariates in survival analysis. Stat Med 1999;18:681—94. 47. Vles W, Steyerberg E, Meeuwis J, L L. Pre-hospital trauma care: a proposal for more efficient evaluation. Injury 2004; 35, 725—33. 48. Zhou XH, Eckert GJ, Tierney WM. Multiple imputation in public health research. Stat Med 2001;20:1541—9.
www.elsevier.com/locate/injury
International comparison of prehospital trauma care systems Bahman S. Roudsari a,*, Avery B. Nathens b, Peter Cameron c, Ian Civil d, Russel L. Gruen e, Thomas D. Koepsell f,g, Fiona E. Lecky h, Rolf L. Lefering i, Moishe Liberman j, Charles N. Mock k,f, ¨rg Oestern l, Thomas A. Schildhauer m, Hans-Jo Christian Waydhas n, Frederick P. Rivara o,f a
Department of Epidemiology, University of Texas, School of Public Health, Dallas, USA St. Michael’s Hospital, Toronto, Canada c Department of Epidemiology and Preventive Medicine, Monash School of Public Health, Alfred Hospital, Australia d University of Auckland, Auckland Hospital, Auckland, New Zealand e Department of Health Services Research and Policy, Menzies School of Health Research, Australia f Harborview Injury Prevention and Research Centre, Department of Epidemiology, University of Washington, Seattle, USA g Harborview Injury Prevention and Research Centre, Department of Health Services, University of Washington, Seattle, USA h University of Manchester, Department of Emergency Medicine, Manchester, United Kingdom i Biochemical & Experimental Division, University of Cologne, Cologne, Germany j Montreal General Hospital, McGill University Health Centre, Divisions of Surgery and Clinical Epidemiology, Montreal, Canada k Harborview Injury Prevention and Research Centre, Department of Surgery, University of Washington, Seattle, USA l Chefarzt der Klinik fu ¨ r, Unfall und Wiederherstellungschirurgie, Allgemeines Krankenhaus Celle, Germany m Chirurgische Klinik und Poliklinik, BG Kliniken Bergmannsheil, Ruhr-Universitat Bochum, Bochum, Germany n Klinik und Poliklinik fu ¨ r Unfallchirurgie, Universita ¨tsklinikum Essen, Hufelandstr., Germany o Harborview Injury Prevention and Research Centre, Department of Pediatrics, University of Washington, Seattle, USA b
Accepted 27 March 2007 * Corresponding author. Tel.: +1 214 648 1054; fax: +1 214 648 1081. E-mail addresses: [email protected] (B.S. Roudsari), [email protected] (A.B. Nathens), [email protected] (P. Cameron), [email protected] (I. Civil), [email protected] (R.L. Gruen), [email protected] (T.D. Koepsell), [email protected] (F.E. Lecky), [email protected] (R.L. Lefering), [email protected] (M. Liberman), [email protected] (C.N. Mock), [email protected] (H.-J. Oestern), [email protected] (T.A. Schildhauer), [email protected] (C. Waydhas), [email protected] (F.P. Rivara). 0020–1383/$ — see front matter # 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.injury.2007.03.028
994 KEYWORDS Emergency medical services (EMS); Trauma; Injury; International; Mortality; Clustered study; Shock rate; Intracluster correlation coefficient (ICC)
B.S. Roudsari et al. Summary Objective: Given the recent emphasis on developing prehospital trauma care globally, we embarked upon a multicentre study to compare trauma patients’ outcome within and between countries with technician-operated advanced life support (ALS) and physician-operated (Doc-ALS) emergency medical service (EMS) systems. These environments represent the continuum of prehospital care in high income countries with more advanced prehospital trauma care systems. Methods: Five countries with ALS-EMS system and four countries with Doc-ALS EMS system provided us with de-identified patient-level data from their national or local trauma registries. Generalised linear latent and mixed models was used in order to compare emergency department (ED) shock rate (systolic blood pressure (SBP) <90 mmHg) and early trauma fatality rate (i.e. death during the first 24 h after hospital arrival) between ALS and Doc-ALS EMS systems. Logistic regression was used to compare outcomes of interest among different countries, accounting for withinsystem correlation in patient outcomes. Results: After adjustment for patient age, sex, type and mechanism of injury, injury severity score and SBP at scene, the ED shock rate did not vary significantly between Doc-ALS and ALS systems (OR: 1.16, 95% CI: 0.73—1.91). However, the early trauma fatality rate was significantly lower in Doc-ALS EMS systems compared with ALS EMS systems (OR: 0.70, 95% CI: 0.54—0.91). Furthermore, we found a considerable heterogeneity in patient outcomes among countries even with similar type of EMS systems. Conclusion: These findings suggest that prehospital trauma care systems that dispatch a physician to the scene may be associated with lower early trauma fatality rates, but not necessarily with significantly better outcomes on other clinical measures. The reasons for these findings deserve further studies. # 2007 Elsevier Ltd. All rights reserved.
Introduction Prehospital care at the scene of injury and during transportation to a medical care facility is often provided by emergency medical service (EMS) systems and is the first step in managing the injured patient. Currently there are four different types of prehospital trauma care systems worldwide. In some environments, there is no organisation responsible for providing prehospital care to trauma patients. This is the pattern in most developing countries.39 Basic life support (BLS) EMS systems provide noninvasive supportive care to trauma patients. The major role of emergency medical technicians (EMTs) in these systems is to transport trauma patients rapidly to a medical care facility and to keep them alive during transport. Some developing countries and many small cities and rural areas in developed countries are served by this type of EMS system.39 Advanced life support (ALS) EMS systems provide more sophisticated care. In these systems, paramedics have intensive training programs for performing invasive procedures such as intubation and intravenous fluid therapy. In the most advanced form of prehospital trauma care (Doc-ALS EMS systems), physicians are responsible for providing prehospital care to trauma patients at the scene of injury and
during transport. The primary goal of ALS and DocALS EMS systems is to initiate the highest level of care at scene and during transport. The gold standard for the evaluation of the effectiveness of different levels of prehospital trauma care is a randomised trial. These randomised trials would almost certainly have to be group-randomised, since the exposure of interest, i.e. types of EMS system, varies at the population level and not at the individual level.33 However, ethical concerns and resource and infrastructure limitations preclude such trials. Therefore, observational studies have primarily been used to compare patients’ outcome across different systems of prehospital trauma care. However, these studies often suffer from a number of limitations. First, patients within the catchment area of a particular trauma centre often receive prehospital care from a single EMS system. Therefore for an evaluation of the association between type of EMS and patients’ outcome, other settings/areas with different types of EMS systems should be used as controls.33 This issue requires researchers to design and conduct multicentre studies, taking into consideration the correlation of the observations within centres or systems.1,9—11,14,15,26,33 Second, multicentre studies that evaluate different aspects of trauma care often use the in-hospital trauma fatality rate as
International comparison of prehospital trauma care systems the main outcome of interest.3,31,47 However, this outcome may be heavily influenced by the quality of hospital care rather than prehospital interventions.17,27,28,33,44 Considering these limitations, and given the recent emphasis on developing prehospital trauma care globally,39 we embarked upon a multicentre study to compare trauma patients’ outcome across different countries. We sought to test the null hypothesis that patient outcomes do not differ materially between EMS systems. Any superiority of patient outcomes under Doc-ALS compared to ALS EMS systems might be expected to be more prominent for patients with more serious injuries that require higher level of care and more advanced procedures. Due to limited access to the information from countries with BLS EMS system, the focus of this study was on the comparison of the patients’ outcome between ALS and Doc-ALS EMS systems.
Methods Source of data De-identified patient-level data were obtained from regions in five countries with ALS EMS system and four countries with DOC-ALS EMS system. In order to protect countries’ identity, we refer to these countries as ALS1 to ALS5 and Doc-ALS1 to DOC-ALS4. The trauma registries within different regions had dissimilar inclusion criteria. Therefore, to render the mix of studied patients more comparable across regions, we restricted our analysis to criteria common to all registries. Thus the analyses were limited to patients 15—55 years old, with severe injuries (defined as injury severity score6 (ISS) > 15) and directly transported from the scene of injury to a definitive trauma care facility. This study was approved by the institutional review board or other relevant authorities from each participating country and of the University of Washington.
Patient descriptors Baseline characteristics of subjects across the nine regions were age, gender, mechanism of injury, mode of transportation (ground ambulance, air ambulance, etc.), and whether the patient was transported directly from the scene to a centre providing definitive care or was transferred by means of another health care facility. Injury severity was evaluated using the injury severity score (ISS), Glasgow coma scale21 (GCS), and the presence of shock at the scene of injury (systolic blood pressure (SBP) <90 mmHg). Prehospital information included
995
vital signs at scene, prehospital time intervals and prehospital interventions (intravenous access and endotracheal intubation).
Outcomes of interest We defined the emergency department (ED) shock rate (proportion of patients with SBP <90 mmHg in the ED) and early trauma fatality rate (i.e. death during the first 24 h after hospital arrival) as the outcomes of interest, since they may be mainly influenced by the severity of injury and quality of prehospital care rather than the quality of hospital interventions.
Power calculation For the calculation of power, we used a formula for clustered studies15,29 (Appendix A). It is comprised of two factors, the standard sample size calculation and a ‘‘design effect’’ or ‘‘variance inflation factor’’.15,26,30 The design effect depends on the average number of subjects per cluster and the intracluster correlation coefficient (ICC) for the outcomes of interest.33 An estimate of the ICC came from a separate analysis of patient outcome data from trauma systems in the US.11 Considering an estimated frequency of ED shock in ALS EMS systems of 20%,33 an ICC of shock 0.009833 and an average of 1000 subjects per cluster, we estimated that our study had approximately 90% power for detecting an 8% absolute difference in the rate of shock between ALS and Doc-ALS EMS systems. We used the US National Trauma Databank4 in order to calculate the ICC for early trauma fatality (results not presented here). Accordingly, our study had 80% power for detecting 8% absolute difference in early trauma fatality between ALS and Doc-ALS systems, assuming an 18% early fatality rate in ALS EMS systems and ICC equal to 0.019.33
Data analysis We used generalised linear latent and mixed models (GLLAMM) to estimate the odds ratios for the outcomes of interest comparing ALS EMS (reference group) with Doc-ALS EMS system, adjusting for age, sex, type and mechanism of injury, ISS and SBP at scene. Since the exposure of interest (i.e. type of EMS system) varied only at the country level and not at the individual level, we only included random intercept and not random slope to the GLLAMM model. 32,45 To evaluate the hypothesis that any superiority of Doc-ALS over ALS would be most prominent for more severely injured patients, we conducted a sub-analysis that evaluated the association between type of
996
B.S. Roudsari et al.
EMS and patients’ outcome among patients with ISS > 25 using the GLLAMM model. Due to the differences in registry protocols, patients who were dead-on-arrival (DOA) might be under-represented in a particular database. In order to evaluate the potential influence that this difference in inclusion criteria might have had on our estimates, we conducted two other sub-analyses. First, we limited analyses to patients with a detectable blood pressure at the scene of injury. Later, we implemented the same approach focusing on subjects with a detectable blood pressure at the time of ED arrival. For the second approach, we compared ED shock and early trauma fatality among different countries, using logistic regression analysis. In this approach we treated ‘‘country’’ as a fixed effect variable16 and evaluated the heterogeneity in outcomes of interest within the countries with similar type of EMS system. We adjusted the results for patient age, sex, mechanism and type of injury, injury severity score and SBP at scene as potential confounding or precision variables.22 Doc-ALS3 had the largest sample size in our study. Therefore, we used this country as the baseline for calculation of odds ratios for ED shock and early trauma fatality in logistic regression models. This approach allowed us to estimate the ORs more precisely (i.e. with the narrowest possible confidence intervals).22 Finally, in order to quantify within and between region variability in the outcomes of interest, we estimated the intracluster correlation coefficient (ICC) of ED shock and early trauma fatality rate for ALS and Doc-ALS EMS systems.10,15,29,33 One of the obstacles in using trauma registries is the issue of missing data. Table 1 reflects the proportion of missing values for the main variables. Our evaluations showed that missing pattern was not completely at random (data not presented).36— 38,40—42 It has been shown that when missingness
is not completely at random, using complete case analysis could lead to biased estimates.36—38,40,41 Therefore, we used multiple imputation (MI) using predictive mean matching19,36—38,48 in order to deal with these missing data. Special emphasis was placed on the imputation of systolic blood pressure at the scene and at ED. The proportion of missing SBP at scene varied from 0% in ALS5 database to 58% in Doc-ALS2. We conducted 10 imputations. Since different datasets had different characteristics that could qualify for inclusion in the MI model by the above criteria, each dataset was imputed separately and then the datasets were combined. We included the following variables in the MI model12,46: (a) variables that were used in the final analysis that compared patients’ outcome across ALS and Doc-ALS EMS systems: age, sex, mechanism and type of injury, injury severity score, type of EMS, SBP at scene and at ED and death during the first 24 h of hospital arrival; (b) variables that potentially influenced the occurrence of missing data: mode of transportation (i.e. ground ambulance, air ambulance, etc.) and prehospital time intervals; and (c) variables that were correlated with SBP5,46: length of hospital stay, ICU admission and prehospital and hospital interventions. We used the ICE command written by Royston for STATA (STATA Corporation, College Station, Texas) in order to implement the multiple imputation process.34,35 Each imputed dataset was analysed separately. The results were averaged and calculation of the standard errors was based on Rubin’s formula.36—38,40
Results The mean age of the patients varied from 30 years in ALS2 to 35 years in Doc-ALS1. Men comprised 75% (Doc-ALS1) to 82% (Doc-ALS4) of the patients.
Table 1 Proportion of missing data for the main variables Country
Age
Sex
Type of injury
Mechanism
ISS a
SBPb (scene)
SBPb (EDc)
Outcome
LOS d
ALS1 ALS2 ALS3 ALS4 ALS5 Doc-ALS1 Doc-ALS2 Doc-ALS3 Doc-ALS4
0 0 0 2 0 0 0 1 0
0 0 0 0 0 0 0 0 0
4 0 0 0 0 0 0 1 0
3 0 0 0 0 0 0 0 0
1 1 0 2 1 1 0 15 0
16 18 52 52 0 25 58 22 29
3 4 5 26 0 10 7 8 6
2 0 0 0 0 0 1 0 0
1 1 0 1 0 1 7 8 0
a b c d
Injury severity score. Systolic blood pressure. Emergency department. Length of stay.
International comparison of prehospital trauma care systems Table 2 Measurements of injury severity
ALS1 ALS2 ALS3 ALS4 ALS5 Doc-ALS1 Doc-ALS2 Doc-ALS3 Doc-ALS4
Shock at scene (%)
ISSa (mean S.D.)
18 17 21 10 17 21 27 17 20
26 11 24 9 25 11 26 10 27 13 26 10 30 13 32 13 30 13
997
system. In order to quantify the heterogeneity in outcomes of interest within and between countries with similar type of EMS system, we calculated the ICC of ED shock and early trauma fatality rate.15,33 The ICCs of ED shock for ALS and Doc-ALS EMS systems were 0.007 (95% CI: 0.000—0.018) and 0.013 (0.000—0.038), respectively. The ICCs for early trauma fatality rate were 0.019 (95% CI: 0.000—0.049) for ALS and 0.036 (95% CI: 0.000— 0.101) for Doc-ALS EMS systems.
Discussion
Systolic blood pressure <90 mmHg. a Injury severity score.
Shock at the scene was more common in Doc-ALS2 (27%), Doc-ALS1 (21%) and ALS3 (21%). DOC-ALS3 (32 13), DOC-ALS2 (30 13) and Doc-ALS3 (30 13) reported the highest mean ISS (Table 2). The results of the GLLAMM model showed that after adjustment for patient age, sex, type and mechanism of injury, ISS and SBP at scene, the observed difference in ED shock rate (OR: 1.16, 95% CI: 0.73—1.91) did not reach a statistically significant level. However, early trauma fatality was less common in Doc-ALS EMS systems compared with ALS EMS systems (OR: 0.70, 95% CI: 0.54—0.91). Changes in the ORs after restricting the analyses to patients with more severe injuries (ISS > 25) and detectable SBP at scene or at ED have been summarised in Table 3. This shows that the OR for mortality decreased even further with Doc-ALS compared with ALS for more severe injuries. This pattern pertained whether or not analysis was restricted to detectable SBP. Fig. 1a and b demonstrate the adjusted results of between-countries comparison of outcomes, modeling ‘‘country’’ as a fixed-effect variable in the logistic regression model. As presented, there is a considerable heterogeneity in outcomes of interest even among countries with similar type of EMS
Our report is the first study that aims at a comparison of patients’ outcome within and between ALS and Doc-ALS EMS systems using a broad sample of such systems. The following characteristics differentiate this study from prior related reports. First, we used ED shock and early trauma fatality rate, and not in-hospital trauma fatality rate,2,18,20,23,25,43 as the outcomes of interest. As previously mentioned, the rationale behind this decision was the potential dependency of late hospital trauma fatality on the quality of hospital care rather than prehospital care. Ideally, even earlier assessment of mortality status (within the first hour of ED arrival) would provide greater insight into prehospital care. Unfortunately, trauma registries do not record the time of death with this degree of resolution. Second, when an EMS system adopts a particular model of care (i.e. ALS or Doc-ALS) almost all of the patients within the jurisdiction of that EMS receive that level of care if required. For example, in 1990, 98.5% of the 200 largest cities in the US had ALS EMS system.8 In 82% of these cities, ALS units were present at all dispatches.8 To address the possible consequences of this policy decision, we considered type of EMS as a system-level varying exposure of interest. Accordingly, we designed a clustered study, taking into consideration the correlation of observations within each cluster.33
Table 3 Adjusteda odds ratios of ED shock and early trauma fatality rate, comparing Doc-ALS with ALS EMS system (baseline group) for different levels of injury severity and SBP at scene and at ED OR of ED shock b All patients, regardless of their SBP at scene or at ED Patients with detectable SBP at scene Patients with detectable SBP at ED a b
OR of early trauma fatality a
ISS > 15
ISS > 25
ISS > 15
ISS > 25
1.18 (0.73—1.92)
1.00 (0.72—1.39)
0.70 (0.54—0.91)
0.57 (0.42—0.77)
1.09 (0.82—1.43)
1.00 (0.71—1.41)
0.64 (0.49—0.85)
0.55 (0.40—0.75)
1.03 (0.76—1.46)
1.00 (0.68—1.45)
0.61 (0.46—0.80)
0.53 (0.39—0.73)
Adjusted for age, sex, type and mechanism of injury, injury severity score and SBP at scene. Comparing Doc-ALS EMS system to ALS (reference group).
998
Figure 1 Odds ratio ED shock (a) and of early trauma fatality (b) adjusted for age, sex, mechanism and type of injury, injury severity score and SBP at the scene of injury, comparing different countries with DOC-ALS3 as the baseline for comparison.
To date several studies have evaluated the association between type of prehospital care and patients’ outcome and systematic reviews have nicely summarised the results of those studies.7,23,25 These reports have concluded that the evidence supporting advanced prehospital trauma care is lacking. However, almost all of these studies treated prehospital care (i.e. ALS or BLS) as a patient-level
B.S. Roudsari et al. varying exposure and used hospital trauma fatality as the main outcome of interest. Furthermore, with the exception of a single study,24 no comparison between ALS and Doc-ALS EMS systems has been published. Liberman and his colleagues compared Doc-ALS EMS system in Montreal with ALS EMS system in Toronto.24 They reported that after adjustment for age, injured body region, mechanism of injury and ISS, being treated in the Doc-ALS EMS system was not associated with a reduction in risk of in-hospital death (OR: 1.14, 95% CI: 0.97—1.33).24 We observed considerable heterogeneity in both outcomes even among countries with similar type of EMS system. Nonetheless, while the observed differences in ED shock rate in our study did not reach a statistically significant level, early trauma fatality was significantly less common in Doc-ALS EMS systems. Unfortunately our current data do not allow us to identify those characteristics of Doc-ALS EMS systems that have been associated with lower early trauma fatality rate. However, the magnitude of difference in trauma fatality rate among different countries even with similar types of EMS system was much higher than that can be explained only by the differences in the quality of prehospital care. The differences in mortality between EMS systems may represent a real difference in outcome. However, several measurement issues might also account for the perceived differences. For example, there are differences in registry and EMS protocols in dealing with dead at scene, dead-on-arrival (DOA) and death during the first few minutes after hospital arrival.13 Some trauma registries might not record DOA into their databases which can lead to underestimation of the trauma fatality rate. Pronouncement of death at scene or en route might also vary from system to system. For example, while physicians are qualified to announce a patient ‘‘dead en route’’, EMTs might require a physician’s approval for such a pronouncement. Under these circumstances, ‘‘dead en route’’ subjects might be counted as ED fatalities in an ALS EMS system and as DOA in a Doc-ALS EMS system. Deaths before an adequate survey of injuries might also impact estimation of the ISS, leading to a potential for bias in the types of patients included in the two types of systems’ registries. By the same token, these differences in inclusion—exclusion criteria might influence ED shock rate estimation. However, since only a fraction of patients with shock at ED die during the first few minutes after ED arrival, the magnitude of bias using ED shock as the outcome might be significantly lower. There are some potential limitations to this analysis. First, apart from the number of subjects, the
International comparison of prehospital trauma care systems power of a clustered study is highly dependent on the number of regions or clusters. The small number of the regions restricted our ability to detect small but clinically important differences in distribution of ED shock rate between ALS and Doc-ALS EMS systems, if in reality such differences exist. Second, establishment of comprehensive trauma registries requires long term planning and sufficient financial support. This need can explain why all the collaborating countries in this research project are high income countries with more advanced EMS systems, while no countries with BLS EMS system were represented. Third, although multiple imputation is currently a preferred method for dealing with missing information, it cannot remove all of the uncertainty resulting from the missing data. The method thus yields larger standard deviations and wider confidence intervals than would be expected if there were no missing data. Finally, there were potential differences in registry protocol across countries and EMS systems. In spite of these limitations, our study provided a reasonable basis for comparison of patient outcomes within and between EMS systems. Our findings raise a few important issues. First, there is significant heterogeneity even within a particular type of EMS system. Therefore the ability to generalise across EMS systems is somewhat limited. Second (and closely related), calculation of the ICCs especially for early trauma fatality rate showed a substantial correlation of observations within countries with the same type of EMS system. These large ICCs should be incorporated in sample size calculation of international multicentre studies that use trauma fatality as the main outcome of interest. Third, due to potential differences in registry protocols, study populations should be as comparable as possible. Prospective multicentre studies that follow similar data collection protocols and use uniform inclusion criteria might be the best approach for this purpose.13
Conflicts of interest None of the authors has any conflict of interest with this research project.
Acknowledgements The authors would like to thank the following individuals and organisations for their invaluable support throughout this research project: Dr. Patrick Haegarty, Christopher Mack, Dr. Steve Bowman, Michele Plorde, Victorian State Trauma Registry (Australia), The Quebec Trauma Registry (Canada),
999
German Trauma Registry and German Society of Trauma Surgery (Germany, Austria and the Netherlands), Auckland City Hospital Trauma Registry (New Zealand), Trauma Audit and Research Network and University of Manchester (UK), and King County EMS and Washington State Trauma Registry (USA).
Appendix A
n¼
ðZ a=2 þ Z b Þ2 ½P 1 ð1 P 1 Þ þ P 2 ð1 P 2 Þ ðP 1 P2 Þ2 |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} I
½1 þ ðm0 1Þ ICC |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
(1)
II
References 1. Ahn C, Jung SH, Donner A. Application of an adjusted chi2 statistic to site-specific data in observational dental studies. J Clin Periodontol 2002;29:79—82. 2. Ali JAR, Gana TJ, Bedaysie H, Williams JI. Effect of the prehospital trauma life support program (PHTLS) on prehospital trauma care. J Trauma 1997;42:786—90. 3. American College of Surgeons-Committee on Trauma. Resources for optimal care of the injured patients; 1999. 4. American College of Surgeons COT. NTDB 2004 Fact Sheet; 2004. 5. Arnold AM, Kronmal RA. Multiple imputation of baseline data in the cardiovascular health study. Am J Epidemiol 2003;157:74—84. 6. Baker SP, O’Neill B, Haddon Jr W, Long WB. The injury severity score: a method for describing patients with multiple injuries and evaluating emergency care. J Trauma 1974;14:187—96. 7. Bunn F, Kwan I, Roberts I, R W. Effectiveness of prehospital trauma care. Cochrane Injuries Group; 2001. 8. Cady G, Scott T. EMS in the United States. 1994 survey of providers in the 200 most populous cities. Jems 1994;19:88— 94. 6—7, 100. 9. Campbell M, Thomson S, Ramsay C, et al. Sample size calculator for cluster randomised trials. Comput Biol Med 2004;34:113—25. 10. Campbell MK, Grimshaw JM, Elbourne DR. Intracluster correlation coefficients in cluster randomised trials: empirical insights into how should they be reported. BMC Med Res Methodol 2004;4. 11. Campbell MK, Mollison J, Grimshaw JM. Cluster trials in implementation research: estimation of intracluster correlation coefficients and sample size. Stat Med 2001;20:391—9. 12. Clark TG, Altman DG. Developing a prognostic model in the presence of missing data: an ovarian cancer case study. J Clin Epidemiol 2003;56:28—37. 13. Dick WF, Baskett PJ. Recommendations for uniform reporting of data following major trauma-the Utstein style. A report of a working party of the International Trauma Anaesthesia and Critical Care Society (ITACCS). Resuscitation 1999;42:81—100. 14. Donner A, Klar N. Pitfalls of and controversies in cluster randomization trials. Am J Public Health 2004;94:416—22.
1000 15. Donner A, Klar NS. Design and analysis of cluster randomization trials: in health research. New York: Oxford University Press; 2000. 16. Fitzmaurice GM, Larid NM, Ware J. Applied longitudinal analysis. New Jersey: Wiley; 2004. 17. 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. 18. Gold CR. Prehospital advanced life support vs. ‘‘scoop and run’’ in trauma management. Ann Emerg Med 1987;16:797— 801. 19. Heitjan DF, Little RJA. Multiple imputation for the fatal accident reporting system. Appl Stat 1991;40:13—29. 20. Isenberg DL, Bissell R. Does advanced life support provide benefits to patients?: a literature review. Prehospital Disaster Med 2005;20:265—70. 21. Jennett B, Teasdale G, Braakman R, et al. Predicting outcome in individual patients after severe head injury. Lancet 1976;1:1031—4. 22. Koepsell TD, Weiss NS. Epidemiologic methods: studying the occurrence of illness. Oxford University Press; 2003. 23. Liberman M, Branas C, Mulder D, et al. Advanced versus basic life support in the prehospital setting- the controversy between the scoop and run and the stay and play approach to the care of the injured patient. Int J Disaster Med 2004;2:1—9. 24. Liberman M, Mulder D, Lavoie A, et al. Multicentre Canadian study of prehospital trauma care. Ann Surg 2003;237: 153—60. 25. Liberman M, Mulder D, Sampalis J. Advanced or basic life support for trauma: meta-analysis and critical review of the literature. J Trauma 2000;49:584—99. 26. Localio AR, Berlin JA, Ten Have TR, Kimmel SE. Adjustments for centre in multicentre studies: an overview. Ann Intern Med 2001;135:112—23. 27. Maio RF, Garrison HG, Spaite DW, et al. Emergency medical services outcomes project I (EMSOP I): prioritizing conditions for outcomes research. Ann Emerg Med 1999;33:423—32. 28. Maio RF, Garrison HG, Spaite DW, et al. Emergency medical services outcomes project (EMSOP) IV: pain measurement in out-of-hospital outcomes research. Ann Emerg Med 2002;40:172—9. 29. Murray DM. Design and analysis of group randomised trials. New York, NY: Oxford University Press Inc.; 1998. 30. Murray DM, Varnell SP, Blitstein JL. Design and analysis of group-randomised trials: a review of recent methodological developments. Am J Public Health 2004;94:423—32.
B.S. Roudsari et al. 31. Nathens AB, Jurkovich GJ, Rivara FP, Maier RV. Effectiveness of state trauma systems in reducing injury-related mortality: a national evaluation. J Trauma 2000;48:25—30. 32. Rabe-Hesketh S, Skrondal A, A P. GLLAMM Manual. U.C. Berkeley Division of Biostatistics Working Paper Series; 2004. 33. Roudsari B, Nathens A, Koepsell T, et al. Analysis of clustered data in multicentre trauma studies. Injury 2006. 34. Royston P. Multiple imputation of missing values. Stata J 2004;4:227—41. 35. Royston P. Multiple imputation of missing values:update. Stata J 2005;5:1—14. 36. Rubin DB. Inference and missing data. Biometrika 1976;63: 581—92. 37. Rubin DB. Multiple imputation after 18+ years. J Am Stat Assoc 1996;91:434. 38. Rubin DB, Schenker N. Multiple imputation in health-care databases: an overview and some applications. Stat Med 1991;10:585—98. 39. Sasser S, Varghese M, Kellermann A, JD. L. Prehospital trauma care systems. Geneva: World Health Organisation; 2005. 40. Schafer JL. Multiple imputation: a primer. Stat Meth Med Res 1999;8:3—15. 41. Schafer JL, N S. Inference with imputed conditional means. J Am Stat Assoc 2000 1995; 95, 1144—54. 42. Schiattino I, Villegas R, Cruzat A, et al. Multiple imputation procedures allow the rescue of missing data: an application to determine serum tumor necrosis factor (TNF) concentration values during the treatment of rheumatoid arthritis patients with anti-TNF therapy. Biol Res 2005;38:7—12. 43. Spaite DW, Valenzuela TD, Criss EA, et al. A prospective infield comparison of intravenous line placement by urban and nonurban emergency medical services personnel. Ann Emerg Med 1994;24:209—14. 44. Spaite DW, Maio R, Garrison HG, et al. Emergency medical services outcomes project (EMSOP) II: developing the foundation and conceptual models for out-of-hospital outcomes research. Ann Emerg Med 2001;37:657—63. 45. Twisk JWR. Applied longitudinal data analysis for epidemiology; a practical guide. Cambridge: Cambridge University Press; 2003. 46. van Buuren S, Boshuizen HC, Knook DL. Multiple imputation of missing blood pressure covariates in survival analysis. Stat Med 1999;18:681—94. 47. Vles W, Steyerberg E, Meeuwis J, L L. Pre-hospital trauma care: a proposal for more efficient evaluation. Injury 2004; 35, 725—33. 48. Zhou XH, Eckert GJ, Tierney WM. Multiple imputation in public health research. Stat Med 2001;20:1541—9.