Upload
dothu
View
214
Download
0
Embed Size (px)
Citation preview
The Influence of Prehospital Times on Trauma Mortality
James Paul Byrne
A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy
Institute of Health Policy, Management, and Evaluation University of Toronto
© Copyright by James Paul Byrne 2018
ii
The Influence of Prehospital Times on Trauma Mortality
James Paul Byrne Doctor of Philosophy
Institute of Health Policy, Management, and Evaluation University of Toronto
2018
Emergency medical services (EMS) are a crucial component of high performing systems.
The impact of EMS and prehospital trauma system factors on trauma mortality is
understudied. This thesis seeks to clarify the role of EMS prehospital times, one important
system-level factor, on trauma mortality at the hospital and population-levels.
In the first part of the thesis, the influence of prehospital times on trauma center
mortality is investigated. Because differences in prehospital times might influence the
likelihood that patients with negligible chance of survival (so-called “unsalvageable”
patients) will arrive at the emergency department (ED), a case definition for such patients is
established. A PROXY definition for the unsalvageable patient based on objective ED
metrics (SBP=0, HR=0, GCS=3) is tested. Construct validity and relevance to trauma center
benchmarking is demonstrated.
Applying the PROXY definition, the influence of prehospital times on trauma center
performance is examined. Using novel linkage of the National EMS Information System
(NEMSIS) database to trauma centers participating in the Trauma Quality Improvement
Program, shorter prehospital times are found to be strongly associated with ED mortality.
However, no association between prehospital time and overall trauma center mortality is
observed after risk-adjustment.
In the second part of the thesis, influence of EMS response times on mortality from
motor vehicle crashes (MVCs) is investigated. Using data derived from the Fatality Analysis
iii
Reporting System (FARS) of the National Highway Traffic Safety Administration, the role
of EMS response times on regional differences in prehospital death from MVCs is examined.
After adjusting for potential confounders at the occupant, vehicle, and crash-levels, EMS
response times are found to be significantly associated with risk of prehospital death.
Finally, using novel linkage of NEMSIS and FARS data, the relationship between
EMS response time and MVC mortality is examined at the population-level within 2,268
United States counties. After adjusting for age, gender, on-scene and transport times, access
to trauma resources, state traffic safety laws, and rurality, longer EMS response times are
found to be strongly associated with greater rates of MVC-related death.
iv
ACKNOWLEDGEMENTS
I am grateful to the following people who made this undertaking possible:
Foremost, to Avery Nathens, for his generosity of time, for protecting my interests, and for
providing incredible opportunities for personal and professional growth. Through Avery’s
seemingly innate ability to see potential in others, I gained a mentor and a role model. Any
achievements made during this time were only possible through his tireless support.
To Paul Karanicolas and Sandro Rizoli, for sharing their wealth of experience in research,
and advancing my critical-thinking skills through dedication to high methodological standards.
To N Clay Mann, for opening doors to working with National EMS Information System, and
for his generosity in sharing an unparalleled understanding of prehospital trauma care. To
Mengtao Dai, for patiently responding to dozens of questions and data requests.
To the team at TQIP, Melanie Neal, Christopher Hoeft, and Ryan Murphy. Through their
knowledge and collaboration, I went from statistically naïve to capable, and remain
enormously inspired about the possibilities for using data to improve trauma care.
To Wei Xiong, whose statistical insights shared over coffee during the early days of this work
were truly essential.
v
To the Clinical Epidemiology and Health Care Research program at the IHPME, under the
leadership of Robert Fowler, for providing a world class course. To Zoe Downie-Ross at the
program office for providing guidance along the way.
To Mary Stott, for being ever-present in her support and life advice.
And to my parents, Dr. Margaret A. Stott and Philip Byrne, who breathed an apprehensive
sigh at the mention of another degree, but nonetheless went all in, as is their nature.
vi
TABLE OF CONTENTS
ACKNOWLEDGEMENTS
TABLE OF CONTENTS
LIST OF ABBREVIATIONS
CHAPTER
1 BACKGROUND
1.1 INTRODUCTION TO TRAUMA EPIDEMIOLOGY AND RESEARCH
1.1.1 Trauma: Describing a Heterogeneous Disease
Definitions and Classification
Anatomic Site and Injury Severity
Timing and Location of Death Following Traumatic Injury
1.1.2 The Epidemiology of Trauma
The Global Burden of Injury
Overview of Trauma Epidemiology in North America
1.1.3 Public Health Approaches to Improving Trauma Care
The Haddon Matrix
The Donabedian Framework
1.2 CONTEMPORARY STRUCTURES AND PROCESSES OF TRAUMA CARE
1.2.1 Trauma Systems
Development of Regional Trauma Systems
Page
iv
vi
x
1
1
1
1
2
3
6
6
7
8
8
9
11
11
11
vii
Core Components of Regional Trauma Systems
The Impact of Trauma Systems on Care of the Injured Patient
1.2.2 Trauma Centers
The Regionalization of Trauma Care: Rationale and Impact on Outcomes
Trauma Center Levels and Resources
Inclusive Versus Exclusive Trauma Systems
Triage and Inter-facility Transfer
1.2.3 Prehospital Trauma Care
Organization of Emergency Medical Services
Emergency Medical Service Level of Care and Prehospital Interventions
Helicopter Emergency Medical Services
Prehospital Time Intervals
Total Prehospital Time
Response Time
On-scene Time
Transport Time
1.3 REGIONAL VARIATIONS IN STRUCTURES, PROCESSES, AND OUTCOMES OF TRAUMA CARE
1.4 RISK-ADJUSTED OUTCOMES AND TRAUMA CENTER PERFORMANCE
The Need to Evaluate Trauma Center Performance
TQIP Inclusion Criteria, Exclusion Criteria, and Data Management
Hierarchical Modelling in Benchmarking of Trauma Center Performance
12
14
15
15
16
17
18
21
21
22
24
28
28
30
31
31
32
33
33
35
36
viii
Aspects of TQIP Methodology Requiring Clarification
1.5 MOTOR VEHICLE CRASH MORTALITY
1.5.1 An Important Research Endpoint in Studies of Trauma System Effectiveness
1.5.2 Factors Influencing Motor Vehicle Crash Mortality
1.5.3 Rural-Urban Disparities in Motor Vehicle Crash Mortality
Prehospital Deaths from Motor Vehicle Crashes
1.6 SPECIFIC OBJECTIVES OF THIS THESIS
2 REDEFINING “DEAD ON ARRIVAL”: IDENTIFYING THE UNSALVAGEABLE PATIENT FOR THE PURPOSE OF PERFORMANCE IMPROVEMENT ABSTRACT
INTRODUCTION
METHODS
RESULTS
DISCUSSION
3 THE IMPACT OF SHORT PREHOSPITAL TIMES ON TRAUMA CENTER
PERFORMANCE BENCHMARKING: AN ECOLOGIC STUDY
ABSTRACT
INTRODUCTION
METHODS
RESULTS
37
38
38
39
41
43
44
47
47
49
50
53
56
62
62
64
65
70
ix
DISCUSSION
4 THE RELATIONSHIP BETWEEN EMERGENCY MEDICAL SERVICE RESPONSE
TIME AND PREHOSPITAL DEATH FROM MOTOR VEHICLE CRASHES: RURAL-URBAN DISPARITIES AND IMPLICATIONS FOR TRAUMA SYSTEM IMPROVEMENT
ABSTRACT
INTRODUCTION
METHODS
RESULTS
DISCUSSION
5 THE RELATIONSHIP BETWEEN EMERGENCY MEDICAL SERVICE RESPONSE
TIME AND MOTOR VEHICLE CRASH MORTALITY: AN ANALYSIS OF 2,268 UNITED STATES COUNTIES
ABSTRACT
INTRODUCTION
METHODS
RESULTS
DISCUSSION
6 SUMMARY AND FUTURE DIRECTIONS
FIGURES
TABLES
REFERENCES
72
77
77
79
80
84
86
91
91
93
94
98
101
105
111
125
147
x
LIST OF ABBREVIATIONS
AAAM
ACS
ADAMS
AIDS
AIS
ALS
AMPT
BAC
BLS
CDC
DC
ED
EMS
FARS
GCS
GIS
HIV
HR
ICD
ISS
LMICs
Association for the Advancement of Automotive Medicine
American College of Surgeons
Atlas & Database of Air Medical Services
acquired immunodeficiency syndrome
Abbreviate Injury Scale
Advanced Life Support
Air Medical Prehospital Triage
blood alcohol concentration
Basic Life Support
Centers for Disease Control and Prevention
District of Columbia
emergency department
emergency medical services
Fatality Analysis Reporting System
Glasgow Coma Scale
geographic information system
human immunodeficiency virus
heart rate
International Classification of Diseases
injury severity score
low and middle-income countries
xi
MVC
NEMSIS
NHTSA
SBP
TQIP
US
WHO
YLL
motor vehicle crash
National Emergency Medical Service Information System
National Highway Traffic Safety Administration
systolic blood pressure
Trauma Quality Improvement Program
United States
World Health Organization
years of life lost
1
CHAPTER 1: BACKGROUND
1.1 INTRODUCTION TO TRAUMA EPIDEMIOLOGY AND RESEARCH
1.1.1 Trauma: Describing a Heterogeneous Disease
Definitions and Classification
Trauma is defined as an injury to living tissue caused by an extrinsic agent or force [1]. In
humans, the potential causes of harm are many and can be mechanical, thermal, electrical,
chemical, or radiant in nature [2]. In some cases, injury may result from a lack of the
necessary substrates for sustaining life, as might occur during drowning or suffocation [3].
The net result is a disruption of normal anatomic or physiologic form and function that places
the individual on a spectrum of injury from which full recovery, permanent disability, or
death may follow. Both the exposure to, and the outcomes of, injury are influenced by
factors that may be within or beyond individual or societal control [4].
External causes of trauma are commonly classified by the etiology of injury as
mechanical (traffic-related injuries, falls, other blunt force trauma, firearm, or stab injuries)
or non-mechanical (thermal, electrical, or chemical causes) [5]. Mechanical causes of injury
are further categorized as blunt (traffic-related injuries, falls, or other blunt force trauma) or
penetrating (firearm and stab injuries). Trauma due to the mechanical transfer of forces
account for more than three-quarters of all injuries [5, 6].
Trauma may also be classified by intent [2, 6]. Most traffic-related injuries, falls,
other blunt force trauma, as well as exposures to thermal, electrical, or chemical insults are
unintentional. Injuries due to assault, self-inflicted violence, and war are considered
intentional.
2
It is common in studies of trauma outcomes for inclusion criteria to be limited to
patients with mechanical trauma (blunt or penetrating mechanisms). This practice is due to
fundamental differences in the nature and public health relevance of injuries due to non-
mechanical causes. For this reason, patients with International Classification of Diseases
(ICD) diagnosis codes [7] for injuries due to non-mechanical causes (electricity, thermal,
chemical) are often excluded [8, 9].
Anatomic Site and Injury Severity
Injuries are also defined by anatomic location and severity. The Abbreviated Injury Scale
(AIS), introduced by the Association for the Advancement of Automotive Medicine
(AAAM) in 1969, is an anatomically based, consensus derived, injury severity scoring
system that classifies an individual injury by body region [10]. The unique body regions
considered include the head, face, neck, chest, abdomen, spine, upper extremities, and lower
extremities. The AIS provides a standardized terminology to describe injuries and ranks
injuries by severity. Each injury with assigned AIS code is associated with a specific injury
score, ranging from 1 (minor injury with low probability of death) to 6 (unsurviveable injury
with 100% probability of death) [11, 12]. Anatomic regions with no injury receive an AIS
score of 0. The AIS is used by motor vehicle crash (MVC) investigators to identify
mechanisms of injury to improve automotive design, and by trauma researchers for the
purpose of risk-adjustment of patient case mix to generate robust estimates of outcomes, both
of which may influence health policy and clinical practice [10]. The AIS, as well as the
International Classification of Diseases (ICD) coding system developed by the World Health
3
Organization (WHO), are used for the coding of injuries in trauma registries worldwide. AIS
scores can also be derived from ICD codes using validated crosswalk algorithms [13].
The AIS is also the basis for the injury severity score (ISS) [10]. While the AIS
describes injuries and their severity by body region, the ISS is a measure of global injury
severity, particularly useful in patients with multiple injuries [14]. Calculation of the ISS is
performed by selecting the highest AIS scores from the three most-severely injured body
regions [15]. The three highest AIS scores are then squared, and then added together to yield
the ISS.
Timing and Location of Death Following Traumatic Injury
As described, trauma is a heterogeneous disease characterized by varying mechanisms and
severity. The timing and location of death from trauma is correspondingly heterogeneous
depending on the nature of the traumatic injury. This must be considered to understand how
and at what rate trauma patients succumb to their injuries, so that specific interventions might
be undertaken to favorably affect the disease course and outcome.
Trunkey first described a trimodal distribution of trauma deaths in the United States
(US) in 1983 [16]. This description was based on the observation that when the rate of death
from injury was plotted as a function of time, the resulting distribution showed three peaks
that corresponded to phases of maximal mortality.
The first peak in the trimodal distribution corresponds to “immediate deaths” – those
patients who die within minutes of injury. Invariably, these patients represent those who
have sustained catastrophic trauma to the central nervous system (CNS) or cardiovascular
system. It follows that the outcomes of these patients are the least modifiable, even with the
4
most expeditious and optimal medical care. Trunkey observed that deaths occurring within
the first hour comprised 45% of all fatalities.
The second peak in the trimodal distribution represent “early deaths” – those patients
with severe injuries who are at risk of death within hours of the insult. Patients in this cohort
often die from the evolution of hemorrhage or other immediately life-threatening injuries in
the head and torso. The risk of death in these patients is the most “time sensitive”, and
outcomes might be favorably modified with urgent intervention. Trunkey found that early
fatalities (1-4 hours) represented 24% of deaths.
The third peak in the trimodal distribution are “late deaths” – patients who die within
days to weeks of their injuries, often due to complications that arise in the setting of multiple
organ dysfunction. These deaths might be modified through optimal multidisciplinary care
in an environment that enhances recovery and minimizes the risk of potentially fatal
complications. Trunkey observed a late (>1 week) mortality rate of 20%.
While Trunkey’s trimodal distribution of deaths provides a useful model for
understanding how different patterns of injury translate to different mortality risks, and how
this risk evolves over time, subsequent studies failed to identify a similar trimodal
distribution [17, 18]. Sauaia and colleagues [17] examined all trauma fatalities in Denver
City and County during 1992, and found that 34% of deaths occurred in the prehospital
setting. A further 53% of deaths occurred during the first 48 hours in hospital, while only
14% of deaths occurred after 1 week. This distribution was similar for deaths from both
blunt and penetrating injuries. Sauaia et al. proposed that the accentuated peak in early in-
hospital deaths they observed was due to an improved emergency medical service (EMS)
response to trauma, leading to the prolonged survival of patients previously considered to be
5
unsalvageable. Demetriades and colleagues, in their study of trauma deaths in Los Angeles
County during 2000-2002, found that 50% of deaths occurred within the first hour (20%
deaths at the scene), 28% within 1-24 hours, and only 7.6% after 1 week [18]. Demetriades
et al. cited advancements in approaches to resuscitation and clinical care leading to a
minimization of late deaths to explain the apparent lack of the “third peak” described by
Trunkey two decades earlier. Both the Sauaia and Demetriades studies provide evidence that
a trimodal distribution of trauma deaths does not exist in mature trauma systems, where EMS
and trauma centers have evolved to provide specialized high-quality trauma care.
Because most trauma research has focused on hospital-based interventions, with the
endpoint of in-hospital mortality, it is also useful to briefly discuss the commonly observed
causes of in-hospital trauma death. Acosta et al. described the causes of deaths from trauma
that occurred at the University of California San Diego Medical Center over a ten-year period
during 1985-1995 [19]. The overall mortality rate for trauma was 7.3%. However, the rate
of mortality among patients with penetrating injuries was twice that of patients with blunt
trauma (11% vs. 5.8%). Most deaths occurred within the first 24 hours (70%). Deaths
within 15 minutes of hospital arrival account for 25% of all fatalities. Deaths from
penetrating trauma were more likely to occur within 15 minutes of arrival compared to blunt
trauma (33% vs. 19%). These very early deaths were most commonly due to major thoracic
vascular (61%) and CNS injuries (27%) in victims of penetrating trauma, whereas severe
multiple system trauma were more commonly the cause in patients with blunt injuries (24%).
Blunt CNS injuries represent 60% of all deaths that occurred between 24 and 74 hours,
reflecting the time taken in the critical care setting for full evaluation and prognostication of
these injuries. Most deaths beyond 72 hours were due to inflammatory sequelae (42%), CNS
6
injuries (26%), and pulmonary emboli (5%) in patients with blunt trauma. Fewer than 10%
of deaths due to penetrating trauma occurred beyond 72 hours.
1.1.2 The Epidemiology of Trauma
The Global Burden of Injury
Trauma is a global public health problem. Each day 14,000 people are killed as the result of
traumatic injuries [20]. More than 5 million trauma-related deaths occur annually,
accounting for 1-in-10 deaths worldwide, nearly 1.7 times the number of fatalities due to
malaria, tuberculosis, and HIV/AIDS combined. Because trauma disproportionately affects
young people, it is one of the leading causes of years of life lost (YLL) at approximately 250
million, second only to cardiovascular disease at 380 million (as a unified entity including
ischemic heart disease, stroke, rheumatic heart disease, cardiomyopathies, and other
circulatory diseases) [21]. For each person killed by traumatic injury, thousands more
survive, many with permanent disabilities. The burden on societies is therefore significant
and complex, with both economic and societal costs.
Violence (28% - 16% suicide, 10% homicide, 2% war), road traffic injuries (24%),
and falls (14%) are the leading causes of trauma fatalities worldwide [20]. Road traffic
injuries are therefore the leading single cause of injury-related death, and are expected to
become the 7th leading cause of all deaths globally by 2030 (currently the 9th leading cause of
death). Road traffic injuries cost most countries approximately 3% of their gross domestic
product [22].
The incidence of traumatic injury is unevenly distributed, both within and between
populations. Significant age and gender disparities exist. Road traffic injuries are the
7
leading cause of death in youth aged 15-29 years [20]. Three-quarters of deaths from road
traffic injuries, four-fifths of homicides, and nine-tenths of deaths from war occur in men
[20], while women suffer disproportionately from physical and sexual violence perpetrated
by an intimate partner.
Poverty is also strongly associated with burden of traumatic injury. Rates of trauma-
related death are significantly higher in low and middle-income counties (LMICs), where
90% of global deaths from injury occur, compared to high-income counties [2, 23]. Among
reasons for disparities in the burden of injury between LMICs and high-income countries are
a relative paucity of health care resources, poorly maintained roadways and infrastructure,
lack of road safety regulations, and stark differences between rural and urban regions with
respect to prehospital emergency services and access to care.
Overview of Trauma Epidemiology in North America
The present thesis utilizes data derived in the North American context, predominantly from
the US. Therefore, a brief description of trauma epidemiology in North America is useful.
In the US, unintentional injury (not including homicide or suicide) is the 4th leading
cause of death (after heart disease, cancer, and chronic respiratory illness), but the most
common cause in people between the ages of 1 and 44 [6]. The overall rate of deaths due to
unintentional injury was 45.8 per 100,000 people (146,571 deaths) in 2015. With a focus on
mechanical causes of injury (excluding poisoning, drowning, suffocation, or burns), the
leading causes of trauma death in descending order are MVCs (including motorcycle), falls,
and firearm injuries (including suicide and homicide). Cumulatively, trauma accounts for
8
more YLL in the US (34% of years of potential life lost – 21% unintentional injury, 8%
suicide, 5% homicide) than cancer (15%) and heart disease (12%) combined.
The toll of traumatic injury is similar in Canada, where unintentional injury is the 5th
leading cause of death (after cancer, heart disease, cerebrovascular disease, and chronic
respiratory illness), accounting for 32.7 deaths per 100,000 people in 2014 (11,724 deaths)
[24]. Based on data from Ontario, Canada’s most populated province (13.6 million residents,
37% of the total Canadian population), the leading causes of trauma death are falls (38%),
MVCs (19%), and firearm injuries (5%) when intent is not differentiated [25].
Rural populations are known to exhibit disproportionately higher trauma-related
mortality rates compared to urban populations [26]. Decreasing population density is one of
the strongest predictors of trauma mortality at the US county level [26, 27]. On average,
trauma deaths in rural areas involve injuries of lower severity than in urban areas [26, 28],
indicating that differences in timely access to medical resources are likely to contribute to
this disparity [26]. Patients injured in rural regions are also more likely to be declared dead-
at-the-scene, implicating delays in discovery and EMS response in rural environments [29].
Higher rates of trauma mortality in rural areas are largely attributable to greater MVC
mortality, with MVC mortality rates showing an inversely-proportional relationship with
population density [30]. Causes for rural-urban disparities in MVC-related deaths are an
important area of ongoing research and are discussed in subsequent sections.
1.1.3 Public Health Approaches to Improving Trauma Care
9
While not explicitly stated in the research articles that comprise this body of work, the
present thesis utilizes accepted approaches to conceptualizing the factors that influence
trauma patient outcomes. These approaches are described here.
The Haddon Matrix
In 1968, William Haddon Jr., a physician appointed by President Lyndon B. Johnson as
administrator of the National Traffic Safety Agency (precursor to the National Highway
Traffic Safety Administration [NHTSA]) [31], published a public health approach to
conceptualizing trauma for the purpose of research directed at ameliorating the burden of
injury on society [32]. This work described three phases of societal concern.
The first phase is pre-event, encompassing all exposures in the environment that
might lead to a traumatic event occurring. The focus in the first phase is countermeasures
and prevention. The second phase relates to the event itself, and the factors that influence the
type and severity of injury to persons involved. The focus in the second phase is minimizing
injury severity and the likelihood of death. The third phase relates to the post-event time
period. The focus in the third phase is achieving “maximum salvage” and involves the type
and timeliness of medical care delivered with the goal of improving patient outcomes. By
considering specific components of the traumatic event as they relate to the pre-event, event,
and post-event phases, the Haddon matrix provides a framework for identifying specific
factors that might be targeted to ameliorate the impact at the population-level.
A common application of the Haddon matrix is to MVCs. For example, targets for
improvement related to drivers include reducing alcohol-impaired driving (pre-crash),
seatbelt use (crash), and early arrival of prehospital medical care (post-crash). Similarly,
10
vehicles might be augmented to achieve improved control in adverse weather conditions
(pre-crash), protect occupants with deployment of airbags (crash), and notify EMS with
automatic crash notification (post-crash). Environmental/roadway interventions include
restrictive speed limits (pre-crash and crash), impact-absorbing barriers (crash), and
optimizing access to prehospital and in-hospital trauma care (post-crash).
In the present thesis, the studies of MVC mortality were conceptualized utilizing the
Haddon matrix to inform which variables at the occupant, vehicle, crash, and system-levels
should be considered in risk-adjusted analyses.
The Donabedian Framework
Avedis Donabedian provided a conceptual framework in which all data relating to quality of
care can be classified as pertaining to one of three categories: structure, process, or outcome
[33]. Structures of care relate to the material attributes and resources of the environment in
which healthcare is delivered. Processes of care describe the sequence of events that take
place during the delivery of care and reflect the activities and decision-making of both the
patient and healthcare provider. The outcome represents the cumulative effect of care on the
health status of both individual patients and the population. Characterization of these
domains is necessary because optimal outcomes are only possible in environments where
good processes are supported by good structures of care. Therefore, it is necessary to
recognize the causal linkage between these components when evaluating quality of care.
The Donabedian Framework is useful in designing studies for the evaluation of the
quality of trauma care [34]. The organization, availability of resources (material and human),
and training of personnel reflect the structures of care within trauma systems and hospitals.
11
The sequence of events following injury, including the type and timeliness if prehospital care
delivered by EMS, decision-making and interventions performed by hospital-based
providers, reflect the processes of trauma care. Both the relevant structures and processes of
care must be thoroughly considered as they pertain to patient outcomes, in order to
appropriately evaluate quality of trauma care.
The present thesis examines the relationship between specific processes of care and
trauma patient outcomes at the individual, hospital, and system-levels. The Donabedian
Framework provides a model for understanding the broader context of interacting factors in
which these relationships occur.
1.2 CONTEMPORARY STRUCTURES AND PROCESSES OF TRAUMA CARE
In order to design studies examining relationships between specific structures, processes of
trauma care, and patient outcomes, it is necessary to describe the current setting in which
these relationships exist. In doing so, gaps in knowledge and sources of variability can be
identified as potential targets for quality improvement, and important confounding factors
can be identified and accounted for in study design and risk-adjustment. The following
sections describe the key structures and processes that are integral to the research papers
included in this thesis: trauma systems, trauma centers, and prehospital care. While these
components are described in distinct sections for convenience in describing the evolution and
current state of the literature pertaining to each, they are not independent. Trauma centers
and effective prehospital trauma care are essential components of modern trauma systems. It
is synergy between these components that cumulatively provide opportunity to achieve high
quality care for injured patients.
12
1.2.1 Trauma Systems
Development of Regional Trauma Systems
In 1966, the National Research Council recognized trauma as the neglected disease of
modern society, causing temporary and permanent disability to more than 10 million and
400,000 US citizens respectively in 1965, and costing society $18 billion (equivalent to $140
billion in 2017) [35]. Underlying the state of affairs identified in this report were a lack of
development of, and coordination between, emergency services at the prehospital and in-
hospital levels to address the unique needs of severely injured patients. Recommendations
from this report provided the impetus for the development of trauma systems coordinated at
the regional level to ensure the resources necessary for early identification and triage of
patients with severe injury, to allow for timely transport to definitive trauma care where
resources exist to address their unique needs. The recognition that “the right patient needs to
get to the right place in the right time” [36] is the core principle behind the evolution of
regional trauma systems. The American College of Surgeons (ACS) Committee on Trauma
provides an outline of the core elements needed to sustain a regional trauma system [37, 38].
By 2017, most states had trauma systems supported by state-level legislation.
Core Components of Regional Trauma Systems
The leadership, statute authority, and funding necessary to drive the organization of trauma
systems is predominantly based at the state-level [37]. Leadership is necessary at multiple
levels, including not only at the state agency level (trauma system medical director), but also
at the level of prehospital agencies and hospital administration (trauma center medical
13
director). The coordination and integration of regional resources to achieve improved trauma
care is made possible by collaboration between the different agencies within the system. The
framework for this collaboration is supported by legislative mandates and funding allocated
by the state government.
Prevention and outreach is an important component of the function of a regional
trauma system. Each regional system is a distinct collection of populations and geographic
regions with unique patterns of injury and opportunities for prevention. It is the role of
leadership and providers within the trauma system to work with community organizations,
businesses, and members of the public to identify injury prevention strategies that are
prioritized based on local trauma epidemiologic data [37]. These strategies may leverage the
community outreach frameworks of existing organizations, such as those that aim to reduce
deaths from alcohol-impaired driving [39] or firearm violence [40].
EMS are integral components of high-performing trauma systems [37]. EMS provide
medical care during the prehospital time period, the critical interval between the moment of
injury and arrival at the definitive care facility. The organization and training of EMS varies
widely between agencies, ranging from volunteer services of providers with limited training,
to highly trained paramedics [41]. Multiple distinct EMS agencies often function within
overlapping geographic or municipal areas. Leadership at the trauma system, EMS, and
hospital-levels must work together closely to develop protocols and assess the quality of
services provided to meet the needs of the population. EMS also play an essential role in the
development and evaluation of disaster preparedness and response plans.
Definitive care facilities are central to regional trauma systems, providing in-hospital
care to injured patients from the surrounding population. Inclusive trauma systems are those
14
that utilize all healthcare facilities within the region to optimally allocate the necessary
material and human resources to meet the unique patterns of need within the system [37].
Such allocation of resources involves designating tertiary regional centers (level I or II
trauma centers) with the specialist capability of treating specific patterns of severe injury
such as burns, traumatic brain injury, spine, or multiple system trauma. Protocols for triage
and inter-facility transfer of patients to tertiary centers should be clearly defined.
Finally, essential to the optimal function of regional trauma systems is the
opportunity for research and evaluation of performance [37]. Central to this function is the
definition of objective metrics of performance and data collection at prehospital and in-
hospital phases of care. Incentives and supports at the system-level are necessary to provide
agencies and healthcare facilities with the motivation to continuously strive to reach
performance metrics. It is the role of system-level leadership to work with agencies and
healthcare facilities to ensure that performance metrics are aligned with system-level needs.
The Impact of Trauma Systems on Care of the Injured Patient
Since the initial implementation of regional trauma systems, the effectiveness of trauma
systems to improve the care of injured patients and reduce mortality has been evaluated
extensively.
Illinois was the first state to implement a state-wide trauma system in 1971. Mullner
et al. examined the impact of trauma system implementation on patient survival in 1978 by
comparing rates of mortality from vehicular trauma during time periods before and after
implementation [42]. This report demonstrated that the case fatality rate decreased
15
substantially at designated trauma centers after trauma system implementation, while
mortality rates at non-trauma centers remained static.
Numerous other studies have subsequently demonstrated that trauma systems are
associated with reduced trauma mortality [43-57]. Examination of deaths due to injury
within regional systems have emphasized decreases in death due to MVCs [45, 46, 57],
among the geriatric population [44], and patients with severe injury requiring laparotomy
[43]. Other studies found that the institution of trauma systems reduced the number of
potentially preventable fatalities [50, 52, 53]. Inclusive trauma systems, those where a
greater number of hospitals are designated to provide a specific level of trauma care, were
found to be associated with lower mortality than those less inclusive [58], suggesting that the
beneficial effect of a regional system is associated with the extent of organization and
coordination of resources.
Studies that reported changes in specific processes of care found that trauma system
implementation was associated with a greater likelihood that patients with severe injuries
would be admitted to tertiary care facilities (level I or II trauma centers) [48, 49, 54], would
undergo transfer to designated trauma centers when appropriate [59], and receive necessary
investigations or interventions in a timely manner [60].
1.2.2 Trauma Centers
The Regionalization of Trauma Care: Rationale and Impact on Outcomes
Central to the function of regional trauma systems is a network of definitive care facilities, or
trauma centers, that are capable of delivering the full spectrum of care for injured patients in
the surrounding population [61]. Trauma centers are acute care hospitals with a defined set
16
of specialist material and human resources that allow them to provide a specific level of care
to injured patients within the broader trauma system [5, 36, 37, 62, 63]. Acute care hospitals
are designated by a lead agency (e.g. Department of Health) as trauma centers. The level of
trauma center designation (discussed below) speaks to the available resources at that center
and the anticipated population (with patterns of injury, severity, and volume) it is to serve.
Through collaboration between hospitals and other agencies within the system, patients are
triaged or transferred to trauma centers with the resources commensurate to their injuries.
The regionalization of trauma care to designated specialist centers is a concept that
has existed since the implementation of regional trauma systems. Significant benefits to
regionalization have been demonstrated. Patients with moderate-to-severe injuries who are
treated at trauma centers are 25% less likely to die [64] and have improved functional
outcomes [65] compared to those treated at non-trauma centers. Furthermore, the added cost
of treatment at trauma centers is acceptable at the societal-level, with greatest cost-
effectiveness in the treatment of younger patients with the most severe injuries [66].
At the population-level, the presence of trauma centers is associated with lower rates
of death due to injury [67-69]. This effect appears to extend to the prehospital environment,
where proximity to a trauma center is associated with lower rates of prehospital death [67] or
death at the scene due to MVC [70]. These data seem to suggest that trauma centers confer a
beneficial effect that extends beyond care delivered in-hospital, that may be related to
delivering a higher quality of preventative efforts or prehospital care. Trauma centers
typically have, as part of their mandate, injury prevention programs and robust performance
improvement processes that touch upon all patients arriving to that center, thus this
relationship is not surprising.
17
Trauma Center Levels and Resources
In the US, trauma centers are recognized according to their resources and roles within the
broader trauma system by one of two processes: designation or verification [63].
Designation is a process carried out by a state or local government to define an acute care
hospital as a trauma center of a specific level (I – V) defined by resource availability [71].
Similarly, the ACS Committee on Trauma provides a framework for trauma center
verification [72]. Verification is a voluntary process undertaken by a hospital to become
recognized as a trauma center of a specific level (I – IV) based on criteria outlined in
Resource for Optimal Care of the Injured Patient [61].
ACS verified level I trauma centers serve as the lead hospital for a regional trauma
system [61]. These are typically academic centers in more population-dense areas that admit
at least 1,200 patients annually. Level II trauma centers often serve as the lead trauma
facility to regions distant from level I trauma centers. Both level I and II trauma centers are
characterized by the rapid availability of specialist services such as critical care, general
surgery, neurosurgery, orthopedic surgery, in addition to more resource-intensive treatment
modalities such as massive transfusion and angiography. Level I and II trauma centers are
predominantly located within urban and suburban regions with high population density,
whereas rural regions with low population density are typically distant from level I and II
trauma centers (Figures 1.1A and B). Level III trauma centers typically serve communities
outside of urban or suburban areas, without access to level I or II trauma centers. These
facilities play the important role of providing high quality emergency and surgical care early
in the post-injury time period. Patients treated at level III trauma centers may remain there if
18
appropriate based on their injuries, while more severely-injured patients are stabilized and
transferred to higher-level facilities based on necessity and regional transfer agreements. The
presence and integration of level I, II, and III trauma centers within the broader system is a
key principle of inclusive trauma systems.
Inclusive Versus Exclusive Trauma Systems
Early in the evolution of trauma systems the focus was on the designation of a select few
highly-resourced acute care hospitals as level I and II trauma centers for the treatment of
patients with the most severe injuries [36]. However, this exclusive trauma system model
ignores the true epidemiology of injury within a region, whereby many acute care facilities
without a trauma center designation will treat patients with both minor and major injuries.
Nathens et al., in their resource-based assessment of trauma care in the US in 2004, found
that one-in-three patients with major trauma were treated at hospitals not designated for
trauma care [73]. The lack of trauma care preparation, or communication with level I or II
trauma centers, at non-designated facilities within exclusive trauma systems leads to
inadequate care and delays in transfer to definitive care for those patients most in need [5, 37,
61].
Therefore, the current emphasis is on the development of inclusive trauma systems, in
which every acute care facility with a 24-hour emergency department (ED) is prepared and
designated to treat trauma patients [36, 37, 61]. When all facilities within a system are
integrated in this manner, the broader trauma system can better allocate resources according
to system-level demands and patients with major trauma will be triaged or transferred to
higher-level facilities in a timelier fashion. The benefit of inclusive systems has been
19
demonstrated by Utter et al., who found that the odds of mortality for patients with severe
injury was significantly lower in the most inclusive trauma systems where the greatest
proportion of acute care facilities received a trauma center designation [58].
Triage and Inter-facility Transfer
For regional trauma systems to be effective at delivering optimal care, it is important that
each patient be transported or transferred to the level of definitive care commensurate with
the nature and severity of their injuries. The process by which this occurs is triage. Derived
from the French word meaning “to sort”, triage as applied to medical treatment on the basis
of injury severity was first implemented by Napoleon’s chief surgeon, Dominique Jean
Larrey, in the late 18th century [5]. In the context of modern trauma systems, triage occurs
first in the prehospital environment, where EMS personnel apply criteria to determine
whether a patient should be transported to a level I or II trauma center, which may not be the
nearest healthcare facility. For patients initially transported to a non-trauma center, it must
next be determined whether inter-facility transfer to a trauma center is necessary.
EMS apply field trauma triage criteria to determine whether injured patients should
first be transported to a trauma center for specialized care [41]. The ACS Committee on
Trauma and the Centers for Disease Control and Prevention (CDC) developed the Guidelines
for Field Triage of Injured Patients which is commonly used for this purpose [74]. This
guideline applies four steps in identifying patients with severe injuries or specific
characteristics who might benefit from trauma center care based on physiologic criteria,
anatomic criteria, mechanism of injury, or special patient characteristics. Physiologic criteria
represent states of extremis identified by decreased levels of consciousness, hypotension, or
20
respiratory compromise. Anatomic criteria reflect specific injuries that are more probable to
require surgical intervention or management by specialist trauma teams. Patients who meet
physiologic or anatomic criteria should be transported preferentially to trauma centers with
the highest available level of care. Patients who meet mechanism of injury criteria (for
example, significant falls or high-risk MVCs) or special characteristics (for example,
children or elders) may not necessarily require transport to the highest level of care, but
should be evaluated at designated facilities capable of determining the nature of their injuries.
In an ideal system, patients with severe injuries who require the specialist resources
of designated trauma centers will be identified and triaged to trauma centers 100% of the
time, while those who do not require trauma center care will be directed to other hospitals.
In other words, the application of field trauma triage criteria will be highly sensitive and
specific, to achieve optimal patient outcomes while limiting the potential for overwhelming
regional resources [36]. Unfortunately, as might be expected due to the challenging process
of evaluating patients in the field, often under difficult conditions with limited resources, the
performance of field trauma triage guidelines is variable and limited [75, 76]. Newgard et al.
found that the sensitivity of the CDC Field Trauma Triage Guidelines to predict severe injury
(ISS ≥ 16) was only 66%, while the sensitivity to predict need for early critical resource use
was 80% [77]. The specificity of the Field Trauma Triage Guidelines to identify patients
without severe injury, or those who did not require early critical resource use, was 88% and
87%, respectively.
Undertriage is the decision to transport a patient to a lower level of care, not
commensurate with the true nature of their injuries. Such undertriage is a system-level
problem that puts patients at risk of adverse outcome. Overtriage, the decision to transport a
21
patient to a trauma center when their injuries might have been sufficiently managed within a
lower level of care, results in an overutilization of finite human, material, and financial
trauma resources.
The undertriage of patients with severe injuries to non-trauma facilities results in a
significantly increased risk of death, compared to those patients transported directly to
trauma centers [78]. Such undertriage may occur more frequently in rural regions where
direct transport to trauma centers is not feasible. In such regions, initiatives such as the Rural
Trauma Team Development Course of the ACS Committee on Trauma may improve the
early care delivered to patients with severe injuries in these areas, and reduce delays in the
transfer process [79].
The sensitivity of trauma triage criteria is greatest in youth, but fails most often in
elderly patients [77]. In response to this, trauma triage criteria have been modified to include
geriatric-specific criteria. These modifications aim to account for changes with age (for
example, changes in normal systolic blood pressure) and have led to significant
improvements in sensitivity to identify elders with severe injury [41, 80, 81]. Future
inclusion of objective data such as prehospital lactate measurement [82-84] or vehicular
crash data [85] might further improve the sensitivity of field trauma triage criteria.
1.2.3 Prehospital Trauma Care
While trauma centers occupy a central role in the structure of regional trauma systems, the
effectiveness of trauma systems to achieve optimal outcomes is limited without access to
high quality prehospital care [63]. EMS perform the essential function of responding rapidly
to the scene of injury, providing on-scene medical stabilization of injured patients, timely
22
triage, and transport to the appropriate level of care for further intervention [36]. There is
considerable heterogeneity with respect to the structures and processes of prehospital care
provided between regional trauma systems. These are discussed in the following sections.
Organization of Emergency Medical Services
There is significant regional variability in the structure and characteristics of EMS owing to
differences in geography and resource constraints [41, 63]. While regulatory authority for
EMS, including treatment protocols and licensure, exists at the state level [41] the
organization of EMS is predominantly at the local level. At present there are approximately
21,283 EMS agencies in the US [86]. Agencies may be operated by the municipality in more
densely-populated regions, or by the county in more rural areas [41]. EMS might be
provided by the local fire department, standalone agencies owned by the municipality, or by
private corporations on a contract-basis, and may be comprised of either paid professionals
or volunteers [41]. The structure of EMS is often dictated by the financial resources and
demand of the local system, with densely-populated municipalities with high call volumes
employing professional EMS personnel, and rural regions employing volunteer members.
Emergency Medical Service Level of Care and Prehospital Interventions
The level of training and scope of practice of prehospital personnel is highly variable
between EMS agencies. Emergency medical technicians are trained to provide Basic Life
Support (BLS), while paramedics provide Advance Life Support (ALS). BLS allows for
basic patient assessment including measurement of vital signs and non-invasive techniques
for airway and ventilation management and hemorrhage control. ALS allows for more
23
invasive methods for securing the airway, including endotracheal intubation, as well as
intravenous fluid resuscitation and delivery of vasoactive medications [41]. Another source
of heterogeneity between local systems, there are numerous levels of care along the spectrum
between BLS and ALS, with NHTSA identifying 39 licensure levels in 30 sampled states and
territories [87].
It is unclear as to whether higher levels of prehospital trauma care result in better
patient outcomes. The greatest controversy surrounds whether prehospital providers capable
of delivering ALS tend to perform a greater number of procedures in the prehospital
environment, therefore leading to delays in achieving definitive in-hospital care. There is
evidence that patients with penetrating trauma may experience significantly longer on-scene
time [88], and in-fact may be at increased risk of death [89], when treated by ALS units in
the field. Wandling et al. further found that private vehicle transport was associated with
improved survival compared to ground ambulance transport in patients with penetrating
trauma [90]. These data support those of Ivatury et al., that the “scoop-and-run” approach
where prehospital interventions are minimized may lead to better outcomes in this patient
population [91]. A systematic review of outcomes in trauma patients receiving prehospital
BLS and ALS is currently underway [92].
The prehospital interventions performed by EMS may vary between regions and by
level of training. One of the most controversial of prehospital interventions under
investigation is endotracheal intubation by ALS providers. Several studies have reported
greater mortality and adverse outcomes associated with endotracheal intubation in the
prehospital environment, for patients with both traumatic brain injury [93-97] and
hemorrhagic shock [98]. Conversely, other studies demonstrated improved neurologic
24
outcomes in patients with traumatic brain injury who underwent rapid sequence intubation in
the field [99, 100]. Proposed explanations for poor outcomes with intubation are secondary
neurologic injury due to hyperventilation [100] or hypoxia [97], or long scene and
prehospital times [97, 98]. Better outcomes in patients intubated by air medical crews
indicate that level of training may be an important factor affecting outcomes from this
intervention [100].
In addition to endotracheal intubation, ALS providers are capable of performing
needle chest decompression in hemodynamically unstable patients with suspected tension
pneumothorax. Recent studies have demonstrated that needle thoracostomy insertion at the
traditional site, the second intercostal space and the mid-clavicular line, is frequently
unsuccessful at puncturing the pleural space due to thickness of the chest wall at this location
[101-103]. For this reason, the fourth or fifth intercostal space at the anterior axillary line is
a more optimal position.
In patients with extremity bleeding, application of a tourniquet has been shown to be
effective in the prehospital setting at stemming hemorrhage where direct pressure is
insufficient [104-108].
ALS providers are also capable of establishing peripheral intravenous access for fluid
resuscitation. Where intravenous access is not readily achieved, rapid intraosseous access
has been demonstrated feasible in the prehospital environment [109]. Prehospital fluid
resuscitation has been associated with poor outcomes in bleeding patients [110, 111] and
those without hypotension [112], in part attributed to longer prehospital times [110].
However, benefit has been demonstrated in patients with traumatic brain injury or
25
hypotension [112-116]. Therefore, a goal-directed approach with fluid resuscitation in
selected patients is warranted [112, 114].
Helicopter Emergency Medical Services
The availability and use of helicopter EMS varies between regional trauma systems. The
first documented patient evacuation by helicopter was during World War II, in which a US
operated Sikorsky YR-4B helicopter was used to rescue an injured British soldier in Japanese
occupied territory in Burma in 1944 [117, 118]. The use of rotor-wing aircraft to provide
prehospital medical care and transport was greatly advanced by the US military during the
Korean War in the 1950s [119]. Helicopters were subsequently used routinely by the US as a
means to evacuate injured soldiers and provide medical treatment en-route to medical
treatment facilities during the Vietnam War, an integral process in recent conflicts in Iraq and
Afghanistan [120]. The development of prehospital medical care utilizing helicopter
transport has been credited with significant decreases in the case fatality rate of casualties in
recent conflicts, although this effect is difficult to isolate due to parallel changes in other
aspects of medical care, patterns of injury, and personal protective equipment [119, 121].
Through partnership with the US military, programs trialing the shared use of helicopters to
augment civilian EMS arose in Mississippi and Texas in the late 1960s [117, 122]. The first
permanent civilian helicopter ambulance service was established in Munich, Germany in
1970. This was soon followed by the first helicopter EMS in the US in Maryland and
Denver, Colorado in 1972 [117, 122, 123].
The benefits of helicopter transport for patients with severe injury are thought to
stem from a higher level of medical training and resources brought directly to the location of
26
injury, a greater speed of transport, and the opportunity to retrieve patients otherwise out of
reach of timely ground EMS care [119]. Baxt et al. found that helicopter aeromedical
transport of injured patients was associated with a 52% reduction in predicted mortality,
hypothesizing that this difference may be due to both the availability of advanced
resuscitation modalities and greater consistency in the high-level training of physician-nurse
teams [119]. Moylan et al. similarly found a distinct survival advantage for patients with
multisystem injuries transported by helicopter (83% vs. 54%), citing an extension of level I
trauma center resources and expertise by helicopter crews into rural areas [124]. In the latter
study, prehospital time intervals were similar between ground and helicopter transport
groups, however patients transported by helicopter received a greater number of potentially
life-saving interventions, including endotracheal intubation and blood transfusion.
Nonetheless, recent evidence for the benefit of helicopter transport in patients with
severe injury is conflicting. Several studies demonstrate a survival advantage for patients
transported by helicopter [125-130], while other studies found no survival advantage [131-
134]. Therefore, efforts to identify which patients might benefit from helicopter transport
have been prioritized. Brown et al. developed the Air Medical Prehospital Triage (AMPT)
score to identify patients who should be triaged to receive helicopter transport [135, 136].
The AMPT score is comprised of 7 criteria, allotting one point each for Glasgow Coma
Score <14, respiratory rate < 10 or > 29 breaths per minute, unstable chest wall fractures,
suspected hemothorax or pneumothorax, paralysis, multisystem trauma (defined as injury to
at least three body regions), and two points for the presence of both physiologic and
anatomic criteria as defined by ACS Committee on Trauma field triage guidelines [74].
Helicopter transport was associated with a survival advantage for patients with AMPT scores
27
≥ 2, while no difference in survival was found in patients with AMPT score < 2 [135].
Therefore, a selective approach to patient transport by helicopter EMS is needed,
incorporating patient factors, injury location relative to trauma center, as well as traffic and
weather conditions [41].
Access to helicopter EMS resources varies significantly between geographic regions
[137], with unclear consequences to trauma-related mortality. Rhinehart et al. examined the
relationship between distance from residence to helicopter EMS airbase and injury-related
mortality [138]. For patients residing > 20 miles from an airbase, increasing distance was
significantly associated with increasing risk of death. Furthermore, the finding that most
airbases are situated at or near a trauma center indicates that poor access to trauma center
care may be further compounded by lack of access to helicopter ambulance resources. A
recent analysis by NHTSA further examined the location of death for fatal MVCs relative to
helicopter ambulance 20-minute coverage areas serving level I or II trauma centers [139]. A
significantly greater number of fatalities occurred at the scene beyond 20-minute coverage
areas (64% vs. 55%), indicating that prompt access to helicopter services in the field may
improve access to in-hospital trauma care for critically-injured crash occupants.
For studies of trauma system effectiveness, an important question is how best to
operationalize measures of system or population access to helicopter emergency medical
system resources. The Atlas & Database of Air Medical Services (ADAMS) provides a
catalog of approximately 95% of rotor wing air medical services, including the locations of
air medical bases with associated numbers of rotor wing aircraft [137]. In 2017, there were
908 air medical bases with 1,049 active helicopter ambulances in the US. Ten-minute fly
circles are used by ADAMS to estimate the proportion of populations and roadways within
28
15-20 minute response areas, assuming an estimate 5-10 minute delay between activation and
launch. Based on a 2014 census of active helicopter ambulances in the US, the cruising
speeds of the ten most common helicopters in use (approximately 90% of all helicopter
ambulances) ranged from 150-178 miles per hour, with median 152 miles per hour
(interquartile range 152-158 miles per hour) [140]. Therefore, ten-minute flight circles
correspond to a distance of 25-30 miles (Figure 1.2), depending on the helicopter. The US
Census Bureau provides shapefiles for census blocks, granular subdivisions of population
last updated in 2010, for use in mapping with geographic information systems (GIS) software
[141]. Similarly, the Federal Highway Administration provides a shapefile for mapping of
interstate, principal arterial, and minor arterial highways [142]. Therefore, the proportion of
populations and major roadways within specific distances of air medical bases can be
estimated for counties and states.
Prehospital Time Intervals
Severe injury is a time-sensitive condition, and the time elapsed from the moment of injury
to arrival at definitive trauma care might influence patient outcomes. For purpose of trauma
system performance improvement, a distinction should be made between those time intervals
that are relevant to EMS quality improvement, and those that are not. Notification time, the
time elapsed between the moment of injury and the call to EMS dispatch, is largely out of
control of the EMS system. Therefore, while notification times are of importance to public
health and the broader trauma system, they are not included in the measurement of EMS time
intervals. EMS time intervals include response time, on-scene time, and transport time. The
total prehospital time is the sum of these time intervals and is a measure of how promptly
29
EMS respond to, provide on-scene care, and transport patients to hospital. The system-level
implications of these time intervals and the current evidence for their impact on patient
outcomes is outlined in the following sections.
Total Prehospital Time
That longer total prehospital times would be associated with greater risk of death in patients
suffering from severe trauma has biological rationale. However, the results of studies
examining this relationship are conflicting. Several studies demonstrate a significant
relationship. Swaroop et al. demonstrated that longer total prehospital times were associated
with greater odds of death in patients presenting with hypotension from penetrating thoracic
trauma [143]. Work from the same group in Chicago, Illinois found that suffering a gunshot
wound more than 5 miles from a trauma center, in so-called “trauma deserts”, was associated
with both increased mean transport time and increased odds of death [144]. These studies
emphasize the strong correlation between total prehospital time and the distance from trauma
center at which injury occurs, as well as the time-critical nature of penetrating torso injuries.
Other researchers examined the importance of prehospital time in patients with traumatic
brain injury. Tien et al. found an that survival was more likely in patients with subdural
hematoma who underwent craniotomy when prehospital time was shorter [145], while Dinh
et al. demonstrated a similar relationship in patients with severe head injury (AIS ≥ 3) [146].
Patients with penetrating torso trauma and traumatic brain injuries are populations in whom
timely surgical intervention and critical care are most likely to make a difference to survival.
Therefore, it stands to reason that delays to definitive care would be observed to adversely
affect outcomes.
30
Examining a mixed trauma population, Feero et al. studied all patients transported to
trauma centers in Portland, Oregon during 1990 irrespective of injury pattern and found that
“unexpected survivors” were associated with significantly shorter total prehospital time
[147]. However, most studies that examined mixed trauma populations have found no
relationship between prehospital time and mortality [148-150]. Since most trauma patients
will not require immediate intervention or critical care upon arrival at hospital, it might be
expected that any incremental relationship between time and outcome in patients with time-
critical conditions would be diluted in analyses that include the broader heterogeneous
trauma population.
Response Time
EMS response time is defined as the time elapsed between EMS notification and ambulance
arrival at the scene of injury. Feero et al., in the study discussed previously, found that
response times were significantly shorter in “unexpected survivors” [147]. Gonzalez et al. in
an analysis of MVCs in the state of Alabama during 2001-2002 found that longer EMS
response times were associated with greater mortality in rural crashes [151]. Other studies
found no relationship between response time and trauma mortality [143, 149, 152, 153].
These studies have considerable limitations, including a lack of risk-adjustment for important
confounders, that hinder their extrapolation to the trauma system level. However, the
findings of Gonzalez et al. in particular indicate that the role of EMS response times in
mortality from MVCs in rural regions should be further investigated.
EMS response time is dependent upon the number of available ambulances and the
efficiency with which units are allocated to the scene of injury. Therefore, this time interval
31
might be shortened by increasing the number of available units [154] and adopting systems
for allocation that are dynamic and adaptable to predicted system demands [155-157].
Because EMS response time is a trauma system measure with the potential to be modified,
and reflects the time to first medical contact with implications for triage and access to trauma
care, it has been identified as a potential quality indicator of prehospital trauma care [158,
159].
On-scene Time
EMS on-scene time is the time elapsed from ambulance arrival at the scene of injury and
departure from the scene en-route to hospital. Most studies that have examined this time
interval have found no relationship with mortality [147, 149, 152, 153, 160, 161]. In
contrast, Gonzalez et al. reported that on-scene times were significantly longer for non-
survivors of rural MVCs [151], although they were unable to adjust for measures of crash
severity. Most recently, Brown et al. found that prolonged on-scene time relative to other
prehospital time intervals was associated with increased mortality in a mixed trauma
population [162]. Need for extrication from MVCs and prehospital intubation, themselves
indicators of injury severity, were found to mediate this relationship.
At the trauma system level, on-scene times might be modified by establishing
prehospital care protocols that limit prehospital interventions (“scoop-and-run” versus “stay-
and-play”). However, at the broader scale the impact of such efforts to shorten on-scene
times might be marginal due to the dominant influence of factors related to the mechanism of
injury (for example, the degree of deformation of the crashed vehicle as an indicator of the
forces imparted upon occupants).
32
Transport Time
EMS transport time is the time elapsed between ambulance departure from the scene of
injury and arrival at hospital. Most studies have found no relationship between transport
time and trauma mortality [143, 147, 149, 150, 153, 161]. One factor hindering the
measurement of the relationship between transport time and mortality at the patient-level is
confounding by indication. Specifically, transport times to hospital are often quicker for
patients identified as suffering life-threatening injuries, and therefore the counterintuitive
result of greater mortality with shorter times may be observed [151]. Confounding by
indication is less problematic for studies of EMS response time since the extent or severity of
patient injuries are unknown prior to ambulance arrival on-scene.
Transport time is most related to the distance between the injury location and the
destination hospital. Therefore, this measure can only be truly modified by changing the
configuration of trauma centers within a region, or by utilizing helicopter EMS where
appropriate.
1.3 REGIONAL VARIATIONS IN STRUCTURES, PROCESSES, AND
OUTCOMES OF TRAUMA CARE
Despite the development of organized systems of trauma care, significant variation exists
between regions in access to trauma care and injury-related mortality. Champion et al. found
that two-thirds of Americans were confident that they would receive “the best medical care”
should they sustain a serious injury [163]. Nonetheless, the likelihood of death following
injury continues to depend on the location at which injury occurs. Minei et al. found that the
33
adjusted mortality rate from severe injury ranged from 3.8 to 29.2 deaths per 100,000 people
between 9 North American sites, with case fatality rates ranging from 14% to 51% [164].
Differences in mortality appear strongly correlated with rural location and access to
trauma care. In a landmark study by Branas et al. in 2005, it was found that while four-fifths
of Americans live within one hour of a level I or II trauma center, by ground ambulance or
helicopter, nearly 50 million persons had no such access [165]. The availability of trauma
resources tends to cluster within densely-populated regions, such that the mean number of
trauma centers accessible within one hour across all Americans is 10. However, this value
does not acknowledge the reality of stark regional disparities, particularly between rural and
urban populations. An updated study by Carr et al. in 2017 found that lack of access to
trauma care is a rural problem, with only 40% of census blocks in rural regions having access
to trauma center care within one hour, compared to 100% in urban areas [166]. Gomez et al.
demonstrated that such lack of access impacts mortality, with highest adjusted rates of
injury-related death in regions with poor access to trauma center care [167]. In this
Canadian, population-based study, census regions with the least access had greater rates of
death at the scene, as well as greater risk of death in the ED, implicating disparities in the
quality of prehospital and in-hospital trauma care in underserved areas.
Such regional variability in access to care and trauma mortality is of great public
health concern, and is increasingly in the public eye [168]. Increasing access to trauma care
is a Health People 2020 priority of the Office of Disease Prevention and Health Promotion,
as is reducing mortality due to MVCs at the population-level [169]. Therefore, further
investigation is needed to determine the differences in structures and processes of care that
contribute to observed disparities in trauma mortality between rural and urban regions. This
34
will serve to identify important targets for system-level quality improvement and is a priority
of this thesis.
1.4 RISK-ADJUSTED OUTCOMES AND TRAUMA CENTER PERFORMANCE
The Need to Evaluate Trauma Center Performance
Trauma centers are the core component of inclusive trauma systems. Despite efforts to
achieve a standardized process for trauma center verification, ensuring that similar specialist
resources and processes of care are available to treat populations, practices and outcomes
continue to vary between institutions [170-173]. Shafi et al. found that adherence to
commonly recommended clinical practices varied widely between trauma centers and was
associated with significant differences in patient mortality [174].
Recognition of the need to monitor and learn from differences in patient outcomes
between trauma centers was the impetus for the creation of the ACS Trauma Quality
Improvement Program (TQIP) [175-178]. In the TQIP pilot study, reported in 2010,
Hemmila et al. identified significant differences in risk-adjusted mortality between hospitals,
despite the presumption of comparable resources at the level I and II trauma centers studied
[177]. To address this disparity, TQIP provides a means for informing local quality
improvement through external benchmarking of trauma center performance [179], built upon
the methodological approaches utilized by the National Surgical Quality Improvement
Program [180, 181]. Specifically, hierarchical regression models are used to determine each
trauma center’s performance (measured as observed-versus-expected ratios, or odds ratios,
with 95% confidence intervals for specific outcomes) relative to other hospitals. The results
of these analyses are fed back to the institution in the form of a benchmarking report, such
35
that specific areas of high or low performance (relative to peers) are highlighted to inform
quality improvement endeavors. At present, more than 400 level I and II trauma centers in
the US participate in TQIP, and a means for benchmarking of performance is mandatory to
achieve verification by the ACS [61].
TQIP Inclusion Criteria, Exclusion Criteria, and Data Management
To ensure that the results of TQIP analyses are useful to participating institutions, inclusion
and exclusion criteria are applied to identify a relevant patient population. Specifically, adult
patients at least 16 years of age with severe injuries caused by blunt or penetrating
mechanisms are included. Relevant injuries are identified by the presence of at least one
ICD-9 code in the range of 800 to 959.9, excluding late effects (905-909.9), superficial
injuries (910-924.9), and foreign bodies (930-930.9) [9]. Severe blunt injuries are defined as
those causing injury with AIS ≥ 3 in at least one of the head, face, neck, thorax, abdomen,
spine, upper extremity, or lower extremity. Severe penetrating injuries are those with AIS ≥
3 in at least one of the neck, thorax, or abdomen. Patients with pre-existing medical
directives to withhold life-sustaining treatment are excluded from TQIP cohorts, as are those
with isolated hip fractures [182].
Quality of the data collected are ensured through multiple mechanisms. Data
abstractors receive dedicated training. Data reliability audits are performed to ensure the
coding accuracy of the variables collected. TQIP also provides data quality reports to
institutions to inform efforts to decrease rates of missing data, incorrect application of
inclusion or exclusion criteria, or variability in the quality of specific data fields [9].
36
Missing data are a common problem among health care registries [183, 184]. The
easiest approach to managing missing data are to exclude patients with missing variables
from analysis. However, if these data are not missing at random, as is often the case,
exclusion of such cases introduces bias into adjusted results [183]. Therefore, TQIP uses a
multiple imputation methodology to deal with missing variables [184-186]. Specifically, a
multiple imputation algorithm is used to estimate the missing values based on non-missing
variables, and the final imputed value is the average of multiple estimates.
Hierarchical Modelling in Benchmarking of Trauma Center Performance
To achieve its stated aims, TQIP utilizes hierarchical logistic regression models to estimate
each trauma center’s risk-adjusted odds of mortality [9]. These models are mixed multilevel
logistic regression models that adjust for patient characteristics as “fixed effects”, while
random-intercept terms are used to account for the non-independence of observations due to
clustering of patients within trauma centers. The advantage of mixed multilevel models is
the ability to model the “cluster level” effect – that of the specific trauma center [187]. In
doing so, each trauma center’s unique odds ratio with 95% confidence interval for mortality,
adjusted for patient case-mix, is calculated from the resulting point estimates. Trauma
centers with lower limits of the 95% confidence intervals that are greater than 1 are identified
as high outliers - those where the odds of mortality are significantly higher than average.
Conversely, trauma centers with upper limits of the 95% confidence intervals that are lower
than 1 are identified as low outliers - those where the odds of mortality are significantly
lower than average. By plotting each trauma center’s odds ratio for mortality with 95%
confidence interval on the same graph, a caterpillar plot is obtained, providing visual
37
representation of each trauma center’s performance relative to peers. In this way, TQIP
benchmarking reports provide participating trauma centers with an objective measure of their
performance relative to peers.
Aspects of TQIP Methodology Requiring Clarification
Specific aspects of TQIP methodology that are the focus of ongoing research include factors
that affect that accuracy of risk-adjustment models to measure trauma center mortality
relative to peers. Such factors include case ascertainment and variables included in the
hierarchical risk-adjustment models.
Case ascertainment is important because it determines whether the performance
measures analyzed apply to the population of trauma patients with severe injury that is
relevant to participating hospitals, and that the selected patient population is appropriately
comparable between trauma centers. Case ascertainment is influenced by inclusion and
exclusion criteria. One exclusion criterion frequently cited by papers in trauma research, and
in TQIP methodology, is patients that are “dead on arrival” [175, 177, 188]. The rationale
used for this exclusion is the patients documented as “dead on arrival” are more likely to be
“unsalvageable” – that their outcome is unlikely to be modified by differences in care – and
therefore their inclusion will only serve to bias study results. Furthermore, hospitals do not
receive the same numbers of such patients, and therefore trauma center performance may be
unfairly biased towards higher mortality at those receiving more “dead on arrival” patients
than the average center. Risk-adjustment models are unlikely to be able to account for these
differences. Unfortunately, definitions for “dead on arrival” are highly variable between
registries and frequently based on subjective or difficult-to-measure criteria [189, 190].
38
Some studies exclude all early deaths, while others make exclusions based on prehospital
cardiac arrest, mechanism of injury, or interventions performed in the ED [191-194]. In an
effort to manage the potential impact of variable case ascertainment of “dead on arrival”
patients on the measurement of trauma center performance in TQIP analyses, TQIP has
previously performed analyses both including and excluding these patients [9]. However, an
improved and validated definition of “dead on arrival”, based on objective and reliable
criteria, is required to ensure accurate exclusion of patients who have negligible opportunity
for survival irrespective of differences in care.
Variables included in risk-adjustment models might also significantly influence the
measurement of trauma center performance. At present, variables that are included in TQIP
risk-adjustment models include those that describe patient demographics, injury mechanism
and severity, as well as ED vital signs. Factors related to the environment in which the
trauma center exists, such as those related to the surrounding trauma system, are often not
measured or included. This has been the focus of concern for some trauma centers where
higher proportions of penetrating violence or shorter EMS prehospital times might lead to
higher rates of “dead on arrival” or critically-injured patients arriving in the ED [195]. This
phenomenon might lead to a profile of higher mortality risk among patients cared for at these
centers which, if not accounted for in risk-adjustment models, could unfairly penalize such
centers with high-than-expected mortality rates. Therefore, there is need to determine
whether the inclusion of such prehospital trauma system factors, such as EMS prehospital
times, significantly influences trauma center performance benchmarking.
1.5 MOTOR VEHICLE CRASH MORTALITY
39
1.5.1 An Important Research Endpoint in Studies of Trauma System Effectiveness
As discussed, examining factors that contribute to variations in trauma mortality between
regions is a priority of this thesis. To achieve this goal, studies in this thesis will use MVC
fatalities as an endpoint to afford certain advantages. First, MVCs are the most common
cause of unintentional trauma death in the US [6], and reducing in MVC mortality is a
national priority [169]. Therefore, this endpoint is of great importance. Second, regional
variations in trauma mortality are exemplified by the extreme differences seen in rates of
MVC deaths between rural and urban regions [30]. Therefore, the endpoint is an ideal
measure for studies examining differences in specific aspects of trauma care across the
spectrum of rurality and access. Finally, MVCs require the full mobilization of trauma
system resources [196, 197], from the prompt response of EMS, to timely triage and inter-
facility transfer to a level I trauma center, and synergism of specialised in-hospital trauma
care. Therefore, MVC fatalities are a useful measure of the net effectiveness of regional
trauma systems.
1.5.2 Factors Influencing Motor Vehicle Crash Mortality
Multiple factors are known to influence MVC mortality by mitigating the risk of crash
fatality during the pre-crash, crash, and post-crash phases of concern. These factors are
important to account for in studies of the impact of trauma systems on MVC mortality.
State laws to limit speed, alcohol-impaired driving, and improve seatbelt use are
known to result in reduced crash fatalities. In the US, the national maximum speed limit of
55 miles per hour, instituted during the 1970s to conserve energy during the Middle East oil
embargo, was associated with a decrease in MVC deaths [198]. This law was repealed in
40
1987, allowing states to raise the maximum posted speed limit on rural interstate highways to
65 miles per hour. As a result, rates of MVC mortality increased in states that increased
speed limits [198-201], largely due to an increase in fatal single vehicle crashes [199]. A
concordant increase in the proportion of vehicles exceeding the 65 mile per hour speed limit
was observed [199], and fatalities also increased on regional roads where speed limits were
not increased, suggesting a spillover effect of driving behavior [200]. High vehicular speed
is significantly more likely in rural environments, and is associated with greater risk of fatal
MVC in rural jurisdictions [202].
Alcohol-impairment significantly increases the likelihood that a fatal crash will occur.
The relative risk of fatal single-vehicle crash injury significantly increases with increasing
values of blood alcohol concentration [203]. In response to overwhelming evidence for their
effectiveness [203-207], all states have now passed laws setting the legal blood alcohol
concentration limit at 0.08% [208]. The 0.08% blood alcohol concentration laws are most
effective in states that also have administrative license revocation laws, allowing law
enforcement officers to revoke licenses when alcohol impairment is suspected [204]. At
present, all but nine states have administrative license revocation laws. Greater enforcement
of alcohol-impaired driving laws, as measured by per capita arrests for driving under the
influence, is also associated with reduction of crash fatalities [209]. Despite the existence of
laws to reduce the risk of alcohol-related crash deaths, more than 25% of drivers killed in
MVCs have blood alcohol concentrations above the legal limit [208]. Alcohol intoxication
also appears to explain a proportion of the significantly higher rates of MVC deaths observed
in rural, compared to non-rural, environments [210].
41
Seatbelt use reduces the likelihood of severe injury or fatality when MVCs occur
[211], as do state laws to enforce compliance with seatbelt laws [212]. Primary enforcement
seatbelt laws are those that allow law enforcement officers to proactively stop vehicles solely
for the reason of non-compliance with seatbelt laws, while secondary enforcement laws only
allow citation for non-compliance when the vehicle has been stopped for a different reason.
Primary enforcement laws have been shown to be significantly more effective at reducing the
risk of a fatal crash [213]. Despite evidence for their effectiveness, only 34 states and the
District of Columbia (DC) have primary enforcement seatbelt laws, while New Hampshire is
unique in having no seatbelt law at all [208]. Crash occupants with incapacitating or fatal
injuries are significantly less likely to be unrestrained in rural compared to urban
environments [214], indicating that lack of seatbelt use may contribute to known higher rates
of crash fatalities in rural regions.
The development of organized trauma systems has been shown to significantly reduce
rates of MVC fatalities at the state population level [196]. In keeping with these findings,
MVC fatalities have been shown to cluster in regions distant from trauma resources such as
trauma centers and helicopter EMS bases [197]. Crash fatalities are also significantly less
likely to occur at the scene with increased proximity to level I or II trauma centers of
helicopter EMS [70, 139]. These findings indicate that the presence of structures and
processes that increase access of crashed vehicle occupants to trauma care are important for
reducing MVC mortality.
1.5.3 Rural-Urban Disparities in Motor Vehicle Crash Mortality
42
As outlined in previous sections, regional variations in access to care and trauma mortality
are an important public health problem [63], and an important focus of research with the goal
of identifying targets for system-level quality improvement. One of the greatest regional
disparities is seen in the association between rural environments and higher rates of MVC
mortality. While the annual rate of MVC fatalities in the US have declined to 10.3 per
100,000 people in 2015 (Figure 1.3) [215], this number remains above 30 per 100,000
people in the most rural areas. Baker et al., in their landmark study published in the New
England Journal of Medicine in 1987, reported a strong inverse correlation between
population density and rate of MVC deaths [30]. This relationship has been repeatedly
confirmed [216-222].
Several factors contribute to the greater risk of crash fatality in rural regions.
Alcohol-impairment is more commonly implicated in serious crashes leading to fatality in
rural jurisdictions [210]. Occupants involved in serious crashes in rural areas are
significantly more likely to be unrestrained or ejected from the vehicle [214, 218, 221]. High
speeds and adverse road conditions contribute to a greater proportion of single vehicle
crashes with severe vehicle deformation [202, 214, 218, 221]. As a result, crashes in rural
regions are more likely to result in incapacitating injury and death [214, 223].
However, higher rates of MVC death in rural areas are not explained by speed [202]
or greater crash severity alone [224]. Vehicle occupants who suffer severe injury are at
significantly greater risk of dying in rural regions [214, 216, 217, 223, 224], suggesting that
differences in the care patients receive during the post-crash time period likely contribute.
Post-crash factors include differences in prehospital care provided by EMS as well as access
to high-quality in-hospital trauma care.
43
EMS are required to cover significantly greater distances [225], and therefore
prehospital times are significantly prolonged in rural environments [29, 221, 226, 227].
Grossman et al., in a study of counties in Washington State, found that response times to
trauma were twice as long to rural locations (median 12 versus 6 minutes) [29]. Scene times
were slightly longer for rural trauma, in keeping with greater need for extrication from
serious MVCs, while transport times to hospital were twice as long. Injury severity did not
significantly influence scene or transport times to hospital indicating that factors related to
the incident, as well geographic and system constraints, contribute to delays in rural areas
more than patient factors. As expected, delays in EMS care are proportional to the degree of
rurality, with frontier regions the most affected [221]. Furthermore, delays in response
appear to affect outcome, with death in the prehospital environment 7 times more likely
among patients with response times greater than 30 minutes [29]. In a recent research letter
by Mell et al., an analysis of data from 485 EMS agencies found that response time to trauma
are consistently more than twice as long in rural jurisdictions [227]. The contributing role of
prehospital times in known disparities in trauma mortality between rural and non-rural
regions requires further investigation.
Access to trauma center care is sparse in rural areas [166], and MVC fatality rates are
inversely proportional to the density of medical treatment facilities [220]. Victims of trauma
are significantly more likely to be transported initially to non-trauma facilities, despite
requiring later inter-facility transfer to level I or II trauma centers [226]. Gomez et al. found
that death in the ED was significantly more likely among severely injured patients
transported to rural facilities [167], indicating that greater access to specialized trauma care is
44
needed. Initial management of severe injury at non-trauma centers and delays in transfer are
an important cause of potentially preventable trauma death [78].
Prehospital Deaths from Motor Vehicle Crashes
As the proportion of MVC fatalities that occur in-hospital has steadily declined, there has
been a commensurate increase in the proportion of deaths that occur in the prehospital
environment to 58% in 2015 (Figure 1.4) [215]. Furthermore, prehospital mortality is
strongly related to rurality, with the proportion of deaths occurring in the prehospital
environment inversely correlated with population density at the county-level (Figure 1.5A
and B). While many of these prehospital deaths are likely to occur on impact, or
immediately following the crash, due to destructive neurologic, vascular, or airway trauma, it
is possible that many may be potentially preventable with timely medical intervention [228].
It is unclear to what degree delays in medical treatment, whether EMS care or access to
definitive trauma care, contribute to rural prehospital mortality [229]. Therefore, there is
need to study the impact of prehospital factors on patterns of prehospital trauma mortality.
1.6 SPECIFIC OBJECTIVES OF THIS THESIS
In providing the background fund of knowledge to this thesis, we have discussed the role of
contemporary structures and processes of trauma care, namely organized trauma systems,
trauma centers, and prehospital trauma care. There is a relative lack of evidence examining
the impact of prehospital trauma system factors on outcomes. Therefore, this thesis will
explore the importance of one specific measure of prehospital trauma systems: prehospital
times. The thesis will be comprised of four research papers that contribute to two primary
45
domains of trauma research: (1) the measurement of trauma center performance, and (2)
regional variation in trauma mortality.
The first part of this thesis aims to improve the measurement of trauma center
performance by fulfilling the following objectives:
OBJECTIVE 1: To determine the optimal case definition for the unsalvageable trauma
patient, that patient who has minimal opportunity for survival, to allow appropriate exclusion
from trauma center performance improvement analyses.
Because a significant proportion of in-hospital trauma deaths occur within minutes of
arrival, attainment of Objective 1 is necessary to ensure that subsequent analyses of the
impact of prehospital times on mortality (Objective 2) are not unfairly biased by the
inclusion of patients with negligible chance for survival.
OBJECTIVE 2: To determine if EMS prehospital times impact the measurement of trauma
center risk-adjusted mortality and should therefore be included in statistical models
evaluating trauma center performance.
The second part of this thesis aims to estimate the contribution of prehospital times to
observed regional variations in trauma mortality by determining the impact of EMS response
times on death from MVCs. EMS response time is the prehospital time interval studied
because ambulance response times should be unbiased by the status of the patient, as injury
severity is generally unknown at the time of EMS activation. EMS response times are
therefore a measure of local prehospital system capabilities – one that is potentially
46
modifiable with resource allocation and system improvements. The influence of EMS
response times on MVC mortality are examined through the following objectives:
OBJECTIVE 3: To determine if EMS response times influence the risk of prehospital
death among fatally-injured crash occupants.
OBJECTIVE 4: To determine if EMS response times significantly influence MVC
mortality at the population level.
47
CHAPTER 2:
REDEFINING “DEAD ON ARRIVAL”: IDENTIFYING THE UNSALVAGEABLE
PATIENT FOR THE PURPOSE OF TRAUMA PERFORMANCE IMPROVEMENT
ABSTRACT
BACKGROUND: Significant variation exists across registries in the criteria used to identify
patients with no chance of survival, with potential for profound impact on trauma center
mortality. The purpose of this study was to identify the optimal case definition for the
unsalvageable patient, for the purpose of exclusion from performance improvement (PI)
endeavors.
METHODS: Data were derived from ACS TQIP for 2012 to 2013. We proposed three potential
case definitions for the unsalvageable patient: (1) no signs of life as determined by local
providers (NSOL), (2) prehospital cardiac arrest (PHCA), and (3) a proxy definition (PROXY)
based on presenting vital signs, defined as ED heart rate = 0, ED systolic blood pressure = 0,
and GCS score motor component = 1. Case definitions were compared using standard
predictive tests to determine specificity and positive predictive value (PPV) for in-hospital
mortality. After the optimal definition was identified, hierarchical logistic regression was used
to assess the impact of including unsalvageable patients on trauma center risk-adjusted
mortality. The impact on trauma center performance was determined as change in outlier status
and performance decile after exclusion of patients who met the optimal case definition.
RESULTS: During the study period, 223,643 patients met inclusion criteria across 192 trauma
centers. Overall in-hospital mortality was 7.2%. The PROXY definition had excellent PPV for
48
death, with less than 1% of patients meeting the PROXY criterion surviving. By contrast,
NSOL and PHCA had PPVs low enough such that many of these patients went on to live (33%
and 10%, respectively). After exclusion of patients who met the PROXY definition, 7% of
trauma centers changed performance decile. This change was greatest for patients with
penetrating injury and shock, with change in performance decile at 23% and 33% of centers,
respectively.
CONCLUSION: The PROXY case definition has excellent predictive utility to identify
patients who, based on presenting vital signs, will go on to die. PROXY should be used to
exclude unsalvageable patients from PI endeavors.
This paper is reprinted from:
Byrne JP, Wei X, Gomez D, Mason S, Karanicolas P, Rizoli S, Tien H, Nathens AB.
Redefining “Dead on Arrival”: Identifying the Unsalvageable Patient for the Purpose of
Performance Improvement. J Trauma Acute Care Surg. 2015 Nov; 79(5): 850-7.
The following co-authors are acknowledged for their permission in reprinting the paper:
Wei Xiong, David Gomez, Stephanie Mason, Paul Karanicolas, Sandro Rizoli, Homer Tien,
and Avery B. Nathens.
49
INTRODUCTION
Performance improvement (PI) is an integral process to ensuring high quality care at trauma
centers. External benchmarking of center performance has become the standard and is now a
requirement for center verification by the ACS [61]. ACS TQIP provides a foundation for
benchmarking by allowing direct comparison of trauma center risk-adjusted outcomes [175,
177, 178].
Since the objective of PI efforts is to identify opportunities for improvement, analyses
should ideally exclude patients who have no chance of survival. Inclusion of such patients
introduces a potential source of confounding variability and distracts from targeting quality
improvement where it is most needed. However, significant variation in criteria used to
identify unsalvageable patients exists between registries [189, 190] and we have estimated
that as many as 47% of deaths may be missed due to variable case ascertainment [230].
Since up to 25% of in-hospital trauma deaths occur within the first 15 minutes of arrival [19],
differences in inclusion/exclusion criteria could profoundly impact risk-adjusted mortality.
Previous studies based definitions for “dead on arrival” (DOA) on mechanism of injury, pre-
hospital cardiopulmonary resuscitation (CPR) or interventions performed in the ED [191-
193]. These definitions were problematic, with many patients defined as DOA having times
to death more than 30 minutes after arrival [192].
Therefore, the need remains for a single reliable case definition for the unsalvageable
patient. The objective of this study was to identify the optimal case definition for the
unsalvageable patient, for the purpose of exclusion from PI endeavours.
50
METHODS
Study Design
This study used a retrospective observational design to achieve three goals: (1) to identify
the optimal case definition for patients who, based on presenting vital signs, have no chance
of survival; (2) to demonstrate the construct validity of this case definition; and (3) to
determine the impact of including unsalvageable patients on trauma center risk-adjusted
performance. The project was approved by the Sunnybrook Health Sciences Center research
ethics board (Toronto, Ontario, Canada).
Data Source & Study Subjects
Data were derived from trauma centers participating in ACS TQIP from January 2012 to
September 2013. During the study period there were 192 participating ACS or state verified
level I and II centers across North America. We identified all adult patients (age ≥ 16 years)
recorded at these centers with known discharge status. Only patients with blunt or
penetrating mechanisms of injury, an AIS ≥ 3 in at least one body region, and Injury Severity
Score (ISS) ≥ 9 were included. Patients with isolated hip fractures, non-mechanical
mechanisms of injury, or pre-existing advanced directives to withhold life-sustaining
measures were excluded.
Proposed Case Definitions
We proposed three potential case definitions for the unsalvageable patient: (1) no signs of
life as determined by local providers (NSOL); (2) pre-hospital cardiac arrest (PHCA); and
(3) a proxy definition (PROXY) based on presenting vital signs, defined as ED heart rate
51
(HR) = 0, ED systolic blood pressure (SBP) = 0, and GCS motor component = 1. First
implemented in 2011, “Signs of Life” is a mandatory dichotomous field collected in TQIP.
NSOL is indicated for this variable when, as defined by the National Trauma Data Standard
(NTDS), a patient is found to have “no organized EKG activity, pupillary responses,
spontaneous respiratory effort, or unassisted blood pressure” [231]. PHCA is defined as
“pre-hospital cardiac arrest with resuscitative effort by healthcare provider”. Vital signs and
GCS components of PROXY are collected in TQIP as initial values documented in the ED.
Identifying the Optimal Case Definition
Case definitions were each applied to the study population in turn and compared using
standard predictive tests to determine specificity and positive predictive value (PPV) for in-
hospital mortality, inclusive for deaths in the ED. We prioritized specificity and PPV over
sensitivity and negative predictive value (NPV) due to the desire to include all patients with a
chance of survival. Patient characteristics were explored for each case definition to
understand limitations in predictive utility. NSOL, PHCA or PROXY variables were
missing in less than 5% of patients; patients missing these data were excluded from analysis.
Construct Validity
After identifying the case definition with the best predictive utility, we assessed the construct
validity of this criterion. A definition with good construct validity should be one that
accurately identifies patients with trajectories and patterns of injury associated with
unpreventable early death. Patient characteristics were stratified by mechanism of injury
(blunt vs. penetrating). Unexpected survivors (patients who fit the optimal case definition
52
but went on to live), as well as patients who died but did not meet the optimal definition,
were characterized to better understand the limitations of this criterion.
Impact of the Unsalvageable Patient on Trauma Center Performance
Risk-adjusted odds-ratios (ORs) for mortality with associated 95% confidence intervals (CIs)
were calculated to determine outlier status and performance decile for each hospital. A
trauma center where the upper limit of the 95% CI was less than 1 was a low outlier – an
above-average performer. Conversely, if the lower limit of the 95% CI was greater than 1
the center was a high outlier – a below-average performer. To assess the impact of including
unsalvageable patients on performance benchmarking, change in trauma center outlier status
and performance decile were calculated after excluding patients who met the optimal case
definition. This analysis was repeated for patients with blunt multisystem injury (defined as
blunt injury with AIS ≥ 3 in two or more body regions), penetrating injury and shock in order
to understand impact on performance with respect to these cohorts.
Statistical Analysis
Means and standard deviations, or medians and interquartile ranges (IQRs), were calculated
for continuous variables. Absolute and relative frequencies were determined for discrete
variables. The Student’s t-test was used to compare continuous variables, while χ2 or
Fisher’s exact tests were used to compare proportions. A hierarchical logistic regression
model, accounting for the multilevel nature of patients clustered within trauma centers, was
used to calculate hospital-specific risk-adjusted ORs for mortality with 95% CIs for all
patients, as well as for blunt multisystem, penetrating injury and shock patient cohorts [9].
53
This model was identical to those used by TQIP to calculate trauma risk-adjusted mortality.
Risk-adjustment accounted for known confounders such as patient baseline characteristics,
initial ED vital signs and injury characteristics [9]. Injury severity was modeled using the
single worst injury (SWI), defined as the injury with lowest survival risk ratio (SRR) for each
patient [232], as well as the AIS for each body region. Model discrimination was estimated
using the c-statistic, and calibration was assessed using observed-versus-predicted outcome
plots. Values of P < 0.05 were considered statistically significant. Data were analyzed using
SAS (version 9.4, Cary, NC).
RESULTS
The Optimal Case Definition
During the study period 223,643 patients met inclusion criteria across 192 trauma centers.
Overall in-hospital mortality was 7.2%. Table 2.1 shows predictive measures of each case
definition for in-hospital mortality. For all criteria specificity was high, reflecting an overall
excellent ability for each definition to identify survivors. Other measures of predictive utility
varied by mechanism. The PROXY definition had excellent PPV for death, with less than
1% of patients meeting PROXY criteria surviving. By contrast, the criteria of NSOL and
PHCA had PPVs low enough such that many of these patients went on to live (33% and 10%
respectively). NPVs did not differ significantly across definitions, reflecting the existence of
trajectories to death unrelated to criteria based on presenting characteristics or vital signs.
There was poor agreement between case definitions, with less than 60% of patients with
NSOL or PHCA also meeting the PROXY definition (Figure 2.1). Overall, 40% of NSOL
54
or PHCA patients had initial HR or SBP > 0, and more than 40% of these patients survived
beyond the ED (Table 2.2).
To understand the limitations of NSOL and PHCA to predict in-hospital death we
compared characteristics of survivors to those who died for each case definition (Table 2.3).
Over one-third of patients assigned NSOL survived. These survivors differed significantly
from those who died, with less penetrating trauma (15% vs. 46%, P<0.001) and fewer
multisystem injuries (22% vs. 43%, P<0.001). More than 98% of these survivors had an
initial HR or SBP > 0, with median GCS 15 (IQR 14-15). However, NSOL patients who
died shared many characteristics with PROXY patients, with similar rates of penetrating
trauma, head and chest injury, absent vital signs and time to death (median 8 minutes, IQR 3-
22).
One in 10 patients with PHCA survived to discharge. Compared to those who died,
survivors were older with lower rates of penetrating injury (13% vs. 37%, P<0.001). The
proportion of patients with multisystem injury and rates of severe injury by body region were
similar among survivors and decedents. More than 96% of survivors with PHCA had an
initial HR or SBP > 0. Many patients who died also had initial HR or SBP > 0 (34%), and
38% of deaths following PHCA occurred beyond the ED.
PROXY Definition - Construct Validity
Having identified PROXY as the case definition with best predictive utility, we set out to
determine the construct validity of this criterion to identify unsalvageable patients. Across
the study population, 2,424 patients met the PROXY definition (1.1%), representing 15% of
all deaths. Twenty trauma centers (10%) did not document any PROXY patients during the
55
study period. Across all centers, the median number of PROXY patients recorded was 7
(IQR 3-16) per center.
Table 2.4 compares characteristics of patients who met the PROXY definition to
those who did not, stratified by mechanism of injury. Blunt trauma PROXY patients were
most likely to be injured by MVC, with multisystem injuries (57%) involving the head
(57%), chest (68%), abdomen (20%) and lower extremities (33%). Penetrating trauma
PROXY patients were predominantly injured by firearm (88%) to the chest (63%) or head
(33%). Inter-facility transfer was rare for patients who met the PROXY definition,
regardless of mechanism. The majority of PROXY patients died in the ED (87%), with
median time to death 8 minutes (IQR 3-23). While the incidence of blunt trauma was nearly
10 times higher than that of penetrating trauma in our cohort, patients suffering penetrating
injury were 10 times more likely to meet the PROXY criterion (5.7% vs. 0.6%, P<0.001).
Twenty-two patients who met the PROXY definition (0.9%) went on to live
(unexpected survivors). These patients were treated at 18 different hospitals. Compared to
those who died, unexpected survivors were significantly more likely to suffer penetrating
trauma (77% vs. 48%, P<0.001), with more chest (86% vs. 65%, P=0.042) and fewer head
injuries (23% vs. 45%, P=0.033) (Table 2.3). Those with penetrating trauma (12 firearm, 5
stabbing) sustained mostly isolated injuries (median 1 body region, IQR 1-2) involving lung
(59%), large blood vessels (53%) or heart (35%). Thoracotomy was performed in 71% of
these patients, and 35% received open cardiac massage.
During the study period, 13,659 patients died yet did not meet the PROXY definition
(85% of all deaths). In contrast to deaths predicted by PROXY, these were older patients
(mean age 60 years), most of whom sustained severe head injury (72%) caused by fall (46%)
56
or MVC (26%). Many of these patients underwent inter-facility transfer (30%), with median
time to death of 52 hours (IQR 12-166).
Impact of the Inclusion of Unsalvageable Patients on Trauma Center Performance
Using the results of our hierarchical logistic regression model, outlier status and performance
decile were determined for each trauma center. With all patients considered in analysis 36
centers were below-average performers (high outliers), and 29 were above-average
performers (low outliers). Overall, change in outlier status occurred at only two trauma
centers (1%) after excluding PROXY patients from analysis; one above-average performer,
and one below-average performer, each showed average performance after exclusion of
patients meeting the PROXY definition. Change in performance decile occurred at 14
centers (7%). When we explored the impact of including unsalvageable patients on trauma
center performance stratified by patient type, greatest impact was observed for performance
in patients with penetrating injury and shock. Change in performance decile occurred at 23%
of trauma centers for patients with penetrating injury (n=43), and at 33% of centers for
patients with shock (n=62) (Figure 2.2).
For each patient cohort, the logistic regression model showed good discrimination (c-
statistic, 0.90 – 0.97) and calibration (based on observed-versus-predicted plots).
DISCUSSION
We demonstrated that the PROXY case definition had excellent specificity and PPV for
identification of patients who would go on to die. This case definition accurately excluded
survivors, with less than 1% of PROXY patients going on to live. These predictive measures
57
were superior to those for NSOL or PHCA, which selected survivors 33% and 10% of the
time, respectively.
Only 59% of NSOL patients met the PROXY definition. This finding is
counterintuitive, considering NSOL is defined by lack of EKG activity, pupillary responses,
respiratory attempt, and blood pressure [231]. NSOL and PROXY should therefore identify
similar patients. Patients with NSOL who died were similar to PROXY. However, survivors
were significantly different, with initial vital signs and GCS that should exclude most from
being considered NSOL. Therefore, the limitations of NSOL to identify unsalvageable
patients appear to be due to reporting error. The “Signs of Life” field was implemented in
the NTDS and integrated into TQIP in 2011 with the goal of identifying patients with “no
signs of life”, a term commonly assigned to patients deemed DOA. Although defined within
the NTDS, the need to interpret multiple clinical findings makes NSOL more prone to error
than PROXY. Furthermore, where patient “Signs of Life” status is not clearly documented,
there is possibility for error if abstractors make inferences from the ED trauma flowsheet or
progress notes. Our findings reflect the challenges of operationalizing new variables across
trauma centers, and highlight the importance of using data that are easily captured and
require limited synthesis.
PHCA also showed poor predictive utility for in-hospital death, with 10% of these
patients surviving. Since PHCA denotes “pre-hospital cardiac arrest with resuscitative
efforts”, documentation in these circumstances should be clear, and therefore this field
should be reliable. Nonetheless, many PHCA patients did not meet the PROXY definition.
This may be explained by the fact that cardiac arrest following trauma might be reversible
and accurate assessment of absent vital signs leading to CPR in the field might be imperfect.
58
Shimazu et al. reported that recordable blood pressure was restored in 33% of patients who
presented without vital signs [233], and as many as 79% of patients with cardiac arrest in the
pre-hospital setting have return of spontaneous circulation [234]. Kleber et al. found that
while 48% of traumatic cardiac arrest was due to hypovolemia, 37% were due to
immediately life threatening thoracic injuries that can be treated (tension pneumothorax,
hypoxia, cardiac tamponade or contusion) [234]. This could explain our finding that many
patients with PHCA had initial HR and SBP > 0, and 38% of deaths occurred beyond the first
hour of admission. Although PHCA following trauma carries poor prognosis, it does not
identify those unsalvageable patients that should be excluded from PI efforts.
The characteristics of patients identified by PROXY correspond with prior literature
on lethal injury. These patients died early (median time to death 8 minutes) and
predominantly fell into two groups: young males with firearm injuries to the head or chest,
and patients with severe multisystem injuries from MVC. Furthermore, patients suffering
penetrating trauma were significantly more likely to meet the PROXY definition compared
to blunt. Acosta et al. found that 25% of deaths occurred within 15 minutes of arrival [19],
most often caused by penetrating thoracic vascular and central nervous system (CNS)
injuries, or blunt multisystem trauma. Head and chest are the most common sites of lethal
penetrating injury [235], and previous studies have demonstrated that patients with
penetrating trauma are significantly more likely to die early compared to other mechanisms
[19, 191, 235]. Patient characteristics described for early trauma deaths in these reports are
consistent with our findings for PROXY. Since the optimal case definition for the
unsalvageable patient should identify early deaths, these similarities lend construct validity to
our findings.
59
We identified 22 unexpected survivors who met the PROXY definition. These were
young males, with mainly isolated penetrating thoracic injuries, many of whom underwent
thoracotomy and open cardiac massage. Prior studies have emphasized the role of emergent
thoracotomy for patients in extremis following penetrating trauma [233, 236, 237], and we
feel that these survivors fall into a category that have greatest potential for salvage from such
intervention.
The majority of deaths in our study cohort (85%) occurred in patients who did not
meet PROXY criteria. These were predominantly older patients with severe blunt head
trauma who died days after injury. While thoracic vascular, CNS and multisystem injuries
are leading causes of death in the first hour, CNS injury and acute inflammatory sequelae
predominate beyond the first day [19]. Therefore, these patients represent trajectories to
death not predicted by presenting characteristics, and so lend further support to our findings.
By identifying patients without vital signs, PROXY successfully identified early deaths and
correctly excluded patients who succumbed later from injuries where outcomes might be
modifiable.
In contrast to case definitions previously reported [191-193], we demonstrated that
the PROXY definition had excellent predictive utility and construct validity to identify
unsalvageable patients. Furthermore, initial vital signs are objective metrics with low
probability of misinterpretation, and less than 3% of initial vital signs or GCS data were
missing in our study population. Therefore the PROXY definition is a feasible case
definition generalizable to all trauma centers.
Exclusion of PROXY patients resulted in change of outlier status at only 1% of
trauma centers. Previous studies suggested that inclusion of unsalvageable patients may have
60
minimal impact on trauma center risk-adjusted OR for mortality, and our findings support
this conclusion [192, 230]. This result can be explained by the fact that PROXY patients
represent a small proportion of all patients, and risk-adjustment controlled for presenting
vital signs and injury characteristics. However, exclusion of PROXY patients resulted in
significant change in trauma center performance for patients with penetrating injury and
shock, with change in performance decile at 23% and 33% of centers respectively for these
groups. Therefore, exclusion of patients who meet PROXY criteria for the purpose of
performance benchmarking remains important, with greatest relevance to centers with high
rates of penetrating injury.
There are important limitations of this study to consider. We are unable to adjust for
differences between trauma systems in pre-hospital care or policies for pronouncing death in
the field. These could result in systematic differences between centers with respect to rates
of patients received in extremis, introducing bias into the results of our risk-adjustment
model. Furthermore, there is potential for information bias if a greater proportion of data is
incomplete in patients who die early, compared to those who survive to undergo more
thorough assessment. This would preferentially underestimate characteristics such as injury
severity in patients who die within minutes, influencing our ability to control for such
factors. Despite these limitations, we feel that our results accurately reflect the utility of
PROXY to identify unsalvageable patients, and the importance of excluding them from
benchmarking analyses.
61
CONCLUSION
The PROXY case definition has excellent predictive utility and construct validity to identify
patients with minimal chance of survival. PROXY should therefore be used to exclude
unsalvageable patients from PI endeavours. Future research will aim to identify optimal
structures or processes of care in patients at high risk of death.
62
CHAPTER 3:
THE IMPACT OF SHORT PREHOSPITAL TIMES ON TRAUMA CENTER
PERFORMANCE BENCHMARKING: AN ECOLOGIC STUDY
ABSTRACT
BACKGROUND: EMS prehospital times vary between regions, yet the impact of local
prehospital times on trauma center (TC) performance is unknown. To inform external
benchmarking efforts, we explored the impact of EMS prehospital times on the risk-adjusted
rate of ED death and overall hospital mortality at urban TCs across the US.
METHODS: We used a novel ecologic study design, linking EMS data from the National
EMS Information System (NEMSIS) to TCs participating in ACS TQIP by destination zip
code. This approach provided EMS times for populations of injured patients transported to
TQIP centers. We defined the exposure of interest as the 90th percentile total prehospital
time (PHT) for each TC. TCs were then stratified by PHT quartile. Analyses were limited to
adult patients with severe blunt or penetrating trauma, transported directly by land to urban
TQIP centers. Random-intercept multilevel modeling was used to evaluate the risk-adjusted
relationship between PHT quartile and the outcomes of ED death and overall hospital
mortality.
RESULTS: During the study period, 119,740 patients met inclusion criteria at 113 TCs. ED
death occurred in 1% of patients, and overall mortality was 7.2%. Across all centers, the
median PHT was 61 minutes (interquartile range, 53–71 minutes). After risk adjustment, TCs
63
in regions with the shortest quartile of PHTs (<53 minutes) had significantly greater odds of
ED death compared with those with the longest PHTs (odds ratio, 2.00; 95% confidence
interval, 1.43–2.78). However, there was no association between PHT and overall TC
mortality.
CONCLUSION: At urban TCs, local EMS prehospital times are a significant predictor of
ED death. However, no relationship exists between prehospital time and overall TC risk-
adjusted mortality. Therefore, there is no evidence for the inclusion of EMS prehospital time
in external benchmarking analyses.
This paper is reprinted from:
Byrne JP, Mann NC, Hoeft CJ, Buick J, Karanicolas P, Rizoli S, Hunt JP, Nathens AB. The
Impact of Short Prehospital Times on Trauma Center Performance Benchmarking: An
Ecologic Study. J Trauma Acute Care Surg. 2016 Apr; 80(4): 586-596.
The following co-authors are acknowledged for their permission in reprinting the paper:
N. Clay Mann, Christopher J. Hoeft, Jason Buick, Paul Karanicolas, Sandro Rizoli, John P.
Hunt, and Avery B. Nathens.
64
INTRODUCTION
External benchmarking has become an integral means for improving quality of care at trauma
centers, and has become a requirement for verification by the ACS [61]. At present,
benchmarking analyses use multivariable models accounting for factors at the patient and
hospital level to estimate trauma center risk-adjusted mortality with excellent discrimination
and calibration [9, 238].
Trauma centers function within a broader system, yet the impact of system-level factors
on trauma center performance is not well known. EMS are key components of high performing
trauma systems. EMS prehospital times, an important system factor, might vary between
regions due to differences in prehospital systems of care, density of trauma resources, and
patterns of injury. Taken together, local prehospital times have the potential to significantly
affect the risk profile of patients arriving at trauma centers, and therefore impact hospital-level
outcomes. For example, short prehospital times might favorably affect outcomes by providing
the greatest opportunity for early hemorrhage control and other lifesaving interventions.
Conversely, short prehospital times might also result in a greater proportion of critically-
injured patients at high risk for early death arriving to the ED alive. Therefore, EMS
prehospital times could impact trauma center performance through two different mechanisms
with opposing effects: in the first scenario trauma center mortality rates would be lower, while
in the second mortality rates would be adversely affected.
To determine if EMS prehospital times, as a system-level factor, should be considered
in external benchmarking analyses we set out to determine the impact of EMS prehospital
times on the risk-adjusted rate of ED death and overall hospital mortality at urban trauma
centers across the US.
65
METHODS
Study Design
This study used a retrospective ecologic design. Since EMS run sheet information is often
unavailable in trauma center registries, we used a novel linkage of data from NEMSIS to data
from destination trauma centers participating in ACS TQIP. Using EMS times for populations
of injured patients transported to urban TQIP centers to derive our exposure, we first assessed
the relationship between local prehospital time and rates of ED death. To determine relevance
to trauma center benchmarking, we then estimated the effect of prehospital time on overall
trauma center mortality. This work was approved by the Sunnybrook Health Sciences Center
research ethics board (Toronto, Ontario, Canada).
Data Sources
Hospital and Patient-level Data
Data were derived from trauma centers participating in ACS TQIP from January 1, 2012
through September 30, 2014. TQIP collects more than 100 patient and institutional variables,
including patient baseline information, injury mechanism and severity, presenting ED
physiologic data and in-hospital outcomes [8, 239]. Reliability of data is ensured through
abstractor training and inter-rater reliability audits at participating sites. At the time of this
study, there were more than 200 participating ACS and state-verified level I and II trauma
centers across North America.
66
Emergency Medical Service Data
EMS data were derived from the NEMSIS National Public-Release Research Dataset and
included EMS activations from January 1, 2012 through December 31, 2013. NEMSIS is a
federally funded project designed to standardize EMS patient care reporting and facilitate a
national data repository for the assessment of EMS systems of care [240]. In 2014, NEMSIS
collected data for 25,835,729 EMS activations submitted by 9,693 EMS agencies serving 48
US states and territories. The volume of EMS records submitted account for approximately
90% of the estimated activations nation-wide, and data quality concerns are actively fed back
to state EMS officials to support reliability efforts. Variables collected include patient
demographics, diagnoses, procedures performed, EMS time intervals (response time, scene
time and transport time) and destination hospital information.
Linkage of Regional Emergency Medical Service Data to Trauma Centers
Each recorded EMS activation within the 2012 NEMSIS National Dataset includes destination
hospital information, allowing for linkage of EMS time data to TQIP trauma centers by
destination hospital zip code. Because it was not possible to link EMS activations to individual
patients within TQIP, this approach provided estimates of EMS time intervals for patients
transported to TQIP centers.
Definition of the Study Cohort
The study cohort was derived following TQIP inclusion criteria. Adult trauma patients (≥16
years of age) with blunt or penetrating injury, Injury Severity Score (ISS) ≥ 9, and known
discharge status were included. Patients with isolated hip fractures, pre-existing advanced
67
directives to withhold life-sustaining care, and those with absent ED vital signs (ED heart rate
[HR] = 0, systolic blood pressure [SBP] = 0 and GCS motor component = 1) were excluded
[241]. We further limited the cohort to those transported directly from the scene of injury to
the receiving hospital by ground ambulance, thereby excluding patients who underwent inter-
facility transfer or transportation by air services. In order to isolate the urban trauma
environment, only trauma centers within metropolitan counties as defined by US Department
of Agriculture Urban Influence Codes [242, 243] were included.
Derivation of Prehospital Times
The total prehospital time, defined as the sum of EMS time intervals from 911 dispatch to
arrival at the destination hospital, was derived from NEMSIS for each trauma patient
transported to an urban TQIP trauma center. Only EMS activations for adult patients (≥16
years of age) with penetrating or blunt mechanisms of injury, transported directly to hospital
by ground ambulance were used.
We defined the exposure of interest as the 90th percentile total prehospital time (PHT)
for each trauma center. Each hospital was associated with a single PHT, reflecting the local
EMS response environment. PHT was then treated as a hospital-level variable in subsequent
analyses.
Patient-level Covariates
Patient-level covariates considered for inclusion in risk-adjustment models included patient
demographics, injury characteristics and ED physiologic variables. Patient demographic
variables included were age, gender, race, comorbid illnesses and health insurance status.
68
Injury characteristics included mechanism and severity. Injury severity was modeled in
adjusted analyses using the survival risk ratio (SRR) of the single worst injury (SWI) [244],
anatomic AIS scores, and ED physiological variables. SRRs are calculated for specific injury
diagnoses as the proportion of all patients with that diagnosis who survive. The SWI is selected
as the lowest SRR for a given patient from their unique set of injury diagnoses, representing
the risk of death associated with their worst injury. SRR-based methods have been shown to
provide superior calibration and fit when modeling trauma patient mortality, compared to
injury severity derived from AIS scores alone (e.g. ISS) [232, 244]. Anatomic injury severity
was included in the models as the AIS score for each body region (head, face, neck, chest,
abdomen, spine, upper and lower extremities). ED physiologic variables included initial ED
HR, SBP, GCS motor component, respiratory rate (RR) and need for assisted respiration. ED
vital signs data (HR, SBP, RR and GCS motor) were missing in less than 3% of patients, and
so were imputed using a multiple imputation technique [186].
Mortality Outcomes
We first evaluated risk-adjusted ED mortality, then overall hospital mortality (inclusive of
death in the ED), as a function of PHT.
Statistical Analysis
Trauma centers were stratified by quartiles of PHT to provide a meaningful basis for
comparing patient and hospital-level characteristics across local EMS systems. Means and
standard deviations, or medians and interquartile ranges (IQRs), were calculated for
continuous variables. Absolute and relative frequencies were determined for discrete
69
variables. The case mix of patients transported to each trauma center was calculated as the
center-specific proportion of all patients with certain demographic, injury, and ED physiologic
characteristics. One-way analysis of variance (ANOVA), or the Kruskal-Wallis test where
appropriate, was then used to compare mean proportions or median values across PHT
quartiles. Differences in hospital-level characteristics, and characteristics of patients who died
in the ED, were also explored across PHT quartiles using the χ2 and Fisher’s exact tests where
appropriate.
Two random-intercept multilevel logistic regression models were used to examine the
adjusted association between PHT quartile and death. The outcome of interest in the first
model was ED death. For the second model, the outcome was overall hospital death inclusive
of deaths in the ED. Both models used a mixed multilevel structure [245] to account for
clustering of patients within trauma centers, while including a random effects term to model
the relative odds of mortality at each center. Covariate selection was performed using the
change-in-estimate approach [246]. The final models therefore included patient-level
covariates, as well as clinically-meaningful interaction terms, that changed the estimate of the
exposure of interest by >10%. For both models, discrimination was estimated using the c-
statistic. Due to sensitivity of the Hosmer-Lemeshow test to large sample size [247],
calibration was assessed using observed-versus-predicted outcome plots.
Multiple sensitivity analyses were performed. First, multilevel modelling was repeated
stratifying the patient population by type of injury (blunt or penetrating). Next, because it is
possible that the impact of PHT on overall mortality might be more pronounced at centers
where more deaths occur in the ED, we stratified trauma centers by the median proportion ED
deaths and created two distinct multilevel models for overall hospital mortality. Finally, to
70
determine the sensitivity of our results to the definition of PHT, we repeated our multivariable
analyses using the median and 25th percentile total prehospital times as the exposure of interest.
All data were analyzed using SAS software (version 9.4, Cary, NC), and statistical
significance was defined by a two-tailed P <0.05.
RESULTS
We identified 119,740 patients transported to 113 urban trauma centers between January 2012
and September 2014 who met inclusion criteria. Approximately 1% of patients (n = 1,240)
died in the ED and overall hospital mortality was 7.2% (n = 8,663). ED deaths accounted for
a median of 13% of all deaths across trauma centers (IQR 9.8-18%). There was significant
variation in local EMS times across hospitals, with a median PHT of 61 minutes (IQR 53 –
71). EMS time intervals and total prehospital times were longest for patients injured by motor
vehicle collision (MVC), and shortest for those injured by firearms (Table 3.1).
There was a relationship between hospital characteristics and PHT (Table 3.2).
Hospitals found in regions with the longest PHTs (quartile 4) were significantly more likely to
be level I trauma centers compared to those with the shortest PHTs (84% vs. 52%, P<0.01).
Further, trauma centers with long PHTs were more commonly university-affiliated or non-
profit hospitals compared to those with short PHTs, although this did not reach statistical
significance.
To explore the relationship between local EMS time and the risk profile of patients
arriving at trauma centers we compared hospital case mix across PHT quartiles (Table 3.3).
Patients were comparable with respect to baseline demographic characteristics across PHT
quartiles. Patients injured as a result of MVCs and those with blunt multisystem injuries were
71
over-represented among hospitals with the longest PHTs, while there was a trend for a greater
proportion of patients with penetrating truncal injuries to be treated at trauma centers with short
PHTs. Although there was no significant difference in anatomic injury severity or ED
physiologic characteristics across PHT quartiles, trauma centers with the shortest EMS times
had a higher proportion of patients die in the ED compared to trauma centers with the longest
PHTs (1.3% vs. 0.77%, P=0.01). Similarly, trauma centers with the shortest PHTs had a
significantly greater proportion of all deaths occur in the ED (18% vs. 11%, P=0.01).
To better understand the relationship between local prehospital times and early deaths,
we explored differences in patients who died in the ED across PHT quartiles (Table 3.4).
While there was no significant difference across PHT quartiles with respect to anatomic injury
severity, patients who died in the ED at trauma centers with the shortest PHTs were
significantly less likely to arrive in shock (28% vs. 37%, P=0.02), pulseless (7.5% vs. 11%,
P=0.01), or in need of assisted respiration (42% vs. 59%, P<0.01). Median time to death was
significantly longer for those patients who died at trauma centers with the shortest PHTs
compared to the longest (51 vs. 27 minutes, P<0.001), a trend which was evident across PHT
quartiles.
We evaluated the association between PHT quartile and risk-adjusted ED death using
our multilevel logistic regression model to adjust for baseline patient and injury characteristics,
and injury severity (Table 3.5). After adjusting for potential confounders, the odds of ED
death at trauma centers with the shortest PHTs (quartile 1) were significantly greater than those
at centers with the longest PHTs (OR 2.00; 95% CI 1.43 – 2.78). This finding was consistent
for patients with penetrating (OR 2.35; 95% CI 1.36 – 4.06) and blunt (OR 2.02; 95% CI 1.34
– 3.03) mechanisms of injury. The regression model for ED death in all patients showed
72
excellent discrimination (c-statistic = 0.95) and calibration (as determined by observed-versus-
predicted plots).
To determine the association between PHT and overall trauma center risk-adjusted
mortality, we used a second multilevel logistic regression model to evaluate the association
between PHT quartile and overall hospital death. After adjusting for covariates, there was no
association between PHT quartile and overall hospital mortality. The lack of any association
was further evident when the analysis was limited to hospitals with a high (≥13%) or low
(<13%) proportion of all deaths occurring in the ED (Figure 3.1). The regression model for
overall hospital death showed excellent discrimination (c-statistic = 0.93) and calibration.
The results of our multilevel logistic regression models for ED death and overall
hospital mortality were consistent within our sensitivity analysis, where PHT was defined as
the median or 25th percentile total prehospital times. Specifically, short PHT was a significant
predictor of ED death, but was not significantly associated with overall hospital mortality,
using the median and 25th percentile times as the exposure of interest.
DISCUSSION
This work evaluated the relationship between local prehospital times and risk-adjusted
mortality at urban trauma centers participating in ACS TQIP. We found that patients
transported to centers in EMS environments where prehospital times are the shortest had twice
the odds of dying in the ED compared to those transported to centers with the longest
prehospital times. This association demonstrated an apparent dose-response relationship
across PHT quartiles, and was consistent for patients with both blunt and penetrating injury.
By contrast, no such relationship was observed when overall hospital mortality was considered.
73
External benchmarking of trauma center performance has become an integral means
for improving quality of care, and is now a requirement for verification by the ACS [61]. The
inception of ACS TQIP was based on the premise that variability in the processes and quality
of care within hospitals is a key contributor to variations in outcomes across trauma centers
[9]. Trauma center outcomes are compared using multivariable models that account for factors
at the patient and hospital-level, estimating risk-adjusted hospital mortality with excellent
discrimination and calibration [238]. However, the impact of system-level factors on trauma
center performance is currently unknown and not considered in TQIP analyses. EMS
prehospital times, an important system factor, vary widely between trauma centers and could
plausibly affect the risk profile of patients arriving in the ED. While the objective of this work
was to determine if local prehospital times should be considered in TQIP risk-adjustment
models, it also provides insight into the influence that EMS times might have on severely
injured patients arriving alive to urban trauma centers.
The finding that short PHTs are significantly associated with increased odds of ED
death can be attributed to several factors acting in concert. First, local prehospital times are a
function of both urban infrastructure and regional patterns of injury, which inherently affect
the risk profile of patients arriving at hospital. For example, patients with penetrating trauma
typically experience short EMS times due to proximity to urban hospitals, while patients
injured in MVCs experience longer prehospital times due to greater distance from trauma
centers and time required for safe extrication. Patients with penetrating injuries are also at
higher risk for early in-hospital death [19, 235], whereas deaths in patients with blunt injuries
who survive EMS transport are more likely to occur later. Therefore, trauma centers with the
shortest prehospital times are often those situated to receive a greater proportion of patients
74
with critical penetrating injuries, and fewer with blunt multisystem trauma, translating to a
patient case mix at higher risk for death in the ED. This relationship is reflected in our results
for the association between PHT quartile and trauma center case mix (Table 3.3).
Trauma centers with short prehospital times might also receive a greater proportion of
severely injured patients at high risk for early death arriving alive to the ED. To overcome the
challenge of variable ascertainment between centers of patients considered “dead on arrival”,
TQIP excludes those with absent presenting vital signs (HR = 0, SBP = 0 and GCS motor =
1), who are shown to have <1% chance of survival [241]. However, it is plausible that such
critically-injured patients will more often arrive with signs of life present when prehospital
times are short, resulting in a higher rate of early in-hospital deaths. This effect has been
described in the military trauma literature, where it has been noted that advances in prehospital
systems in recent conflicts has led to more severely injured patients surviving to medical
treatment facilities, with resultant increase of in-hospital deaths (‘died of wounds’) [121, 248].
Further evidence is provided by our own findings, in which patients who died in the ED at
hospitals with short PHTs were less likely to arrive pulseless or in shock, and in fact survived
longer, compared to those who died at hospitals with longer PHT (Table 3.4). These findings
suggest that shortened lead-times between injury and arrival at trauma centers result in a larger
proportion of fatally-injured patients arriving with greater physiologic reserve.
While PHT was found to be an independent predictor of ED death, we found no
association between EMS prehospital time and overall trauma center risk-adjusted mortality.
There are several potential reasons for this observation. First, while short prehospital times at
the hospital-level may significantly affect the risk-adjusted odds of ED death through
mechanisms outlined above, rates of ED death are relatively low and might have little influence
75
on overall hospital mortality. Even among centers with a large proportion of all deaths
occurring in the ED, we saw no particular influence of prehospital times on overall hospital
mortality. Second, while early deaths account for up to one quarter of all in-hospital trauma
deaths [19], previous studies have demonstrated that the inclusion or exclusion of
unsalvageable patients or ED deaths in benchmarking analyses has minimal impact on the risk-
adjusted mortality of individual trauma centers [192, 230]. Therefore, differences between
trauma centers in early deaths due to PHT would not be expected to affect the measurement of
overall trauma center performance. Most importantly, it is plausible that short prehospital
times do reduce the duration of shock, hypoxemia, and secondary brain injury, and in so doing
reduce the potential for in hospital deaths through lower rates of multiple organ failure or brain
death.
This is the first study to explore the impact of EMS prehospital times on trauma center
performance benchmarking. While previous studies sought to identify a relationship between
EMS times and mortality at the patient level [143, 149, 150, 249], these studies did not explore
the relationship between local prehospital times and mortality at the hospital level. By linking
EMS data contained in the NEMSIS National dataset to patient and hospital-level data in TQIP,
we were able to overcome the prior limitation of sparse prehospital time data collected in
trauma registries. Furthermore, by utilizing an ecologic study design, defining PHT as a
hospital-level exposure, we were able to isolate the net effect of regional EMS environments
on trauma center case mix and mortality. This study design also overcomes the challenge of
confounding by indication observed when considering prehospital times as a predictor of
mortality, occurring due to the more rapid transport by EMS of patients with higher-risk
injuries. Finally, with the exception of limiting the patient cohort to those transported directly
76
to hospital by ground ambulance, our inclusion criteria and multilevel modeling techniques
paralleled those used by TQIP, ensuring the relevance of our study to trauma centers
participating in TQIP.
There are important limitations of this study to consider. First, there is potential for
residual confounding due to unmeasured factors in the prehospital and hospital setting. For
example, differences in EMS practices could impact prehospital times, as well as the
physiological status of patients arriving to hospital, and therefore our outcomes. Second,
because this was an ecologic study, conclusions cannot be made about the causal relationship
between prehospital time and mortality at the patient level (i.e., ecologic fallacy) [250]. Third,
our patient cohort included only patients transported by ground ambulance to urban TQIP
centers. Therefore, the generalizability of our results to all trauma centers may be limited, and
does not consider inter-facility transport. Finally, there is potential for information bias related
to variable ascertainment of injury diagnoses between patients who die early and survivors.
However, by including multiple covariates to reflect injury severity such as SRR, anatomic
AIS, and ED physiologic characteristics we were able to achieve excellent discrimination and
calibration in our risk-adjustment models. Therefore, we are satisfied that our results reflect
the true relationship between EMS prehospital time and trauma center mortality.
CONCLUSION
At urban trauma centers, local EMS prehospital times are a significant predictor of risk-
adjusted ED death. However, no relationship exists between prehospital time and overall
trauma center risk-adjusted mortality. Therefore, there is no evidence for the inclusion of EMS
prehospital time in performance benchmarking analyses.
77
CHAPTER 4:
THE RELATIONSHIP BETWEEN EMERGENCY MEDICAL SERVICE
RESPONSE TIME AND PREHOSPITAL DEATH FROM MOTOR VEHICLE
CRASHES: RURAL-URBAN DISPARITIES AND IMPLICATIONS FOR TRAUMA
SYSTEM IMPROVEMENT
ABSTRACT
BACKGROUND: Motor vehicle crash mortality is greatest in rural regions with lesser access
to prehospital trauma care. We estimated the association between EMS response time (RT)
and the risk of prehospital death in a population-based cohort of MVC occupants with fatal
injuries, and evaluated the relative contribution of RT to the known higher rates of prehospital
mortality in rural environments.
METHODS: This was a retrospective cohort study of MVC occupants who died following
crashes on public roads in the US. Data were derived from the Fatality Analysis Reporting
System (FARS) of NHTSA (2001-2015). Hierarchical logistic regression estimated the
association between RT and prehospital death, adjusted for occupant, vehicle, and crash
variables. Effect modification between RT and rurality (county population density) was
evaluated. The proportional change in variance (PCV) was used to estimate the relative
contribution of EMS response time to prehospital mortality rates across the spectrum of US
county rurality.
78
RESULTS: We identified 117,267 occupants who died following 105,924 crashes in 1,781
counties. Median RT was 9 minutes (IQR 6–14) and 58% died in the prehospital environment.
Longer RT was significantly associated with increased risk of prehospital death (RT>14 vs.
<6 minutes; OR 1.41; 95%CI 1.35-1.46). This relationship was modified by rurality of the
crash location: Long RT (>14 minutes) was associated with significantly higher odds of
prehospital death in rural (<50 ppl/mile2: OR 1.58; 95%CI 1.48-1.70) vs. urban (≥300
ppl/mile2: OR 1.27; 95%CI 1.17-1.39) counties. The PCV due to RT was greatest in the most
rural (>10%) compared to most urban (<2%) counties, indicating that RT explains a
meaningful proportion of known higher rates of prehospital mortality in rural vs. urban regions.
CONCLUSION: In MVC occupants with fatal injury, RT is associated with the risk of
prehospital death. Short RTs are associated with greater likelihood of survival to receive in-
hospital treatment. This relationship is strongest in rural counties, where RTs explain the
greatest proportion of the variance in prehospital mortality.
This work is not yet published.
The following co-authors are acknowledged for their permission to include this manuscript:
N. Clay Mann, Stephanie A. Mason, Jason Buick, Paul Karanicolas, Sandro Rizoli, and Avery
B. Nathens.
79
INTRODUCTION
In 2016, the National Academies of Science, Engineering and Medicine (NASEM) published
a report calling for steps to be taken toward achieving zero preventable deaths from trauma in
the US [251]. This report emphasized the need to minimize variability in access to trauma
system care between urban and rural regions. Given that trauma patient survival is
contingent on a rapid prehospital response and timely in-hospital care, the relative paucity of
such resources in rural regions is an important target for system-level quality improvement
[166].
MVCs are the leading cause of injury-related death in North America [6]. Rural-
urban disparities in MVC-related mortality are well described [30, 214, 221, 223, 224].
While overall MVC mortality has declined [215, 252], annual MVC-related fatalities exceed
30 per 100,000 people in the most rural counties, a rate comparable to some low and middle
income countries [253]. The proportion of crash fatalities that occur in the prehospital
environment is also highest in rural counties, emphasizing the need to identify system-level
factors that can be modified to improve prompt access to in-hospital care for occupants with
critical injuries in these regions [221, 252].
EMS response time, the interval between EMS notification and arrival at the crash
scene, is one such system factor. Response times are highly variable and prolonged in rural
environments [221]. While many prehospital deaths are likely not modifiable by optimal
prehospital care, there is a subgroup of patients with time-critical injuries in whom early
EMS arrival might mean the difference between death at the scene and surviving to receive
in-hospital treatment. These high-risk patients could account for a meaningful number of
prehospital deaths in rural regions where response times are long. At present, the
80
contribution of prolonged EMS response times to known higher rates of prehospital mortality
in rural regions are not known.
To inform trauma system resource allocation, we measured the association between
EMS response time and prehospital mortality in a population-based cohort of MVC
occupants with fatal injuries and estimated the relative contribution of EMS response times
to the known higher rates of prehospital mortality in rural environments.
METHODS
Study Design and Objectives
We used a retrospective cohort study design to achieve three objectives: 1) to measure the
association between EMS response time and prehospital mortality at the vehicle occupant-
level, 2) to estimate the potential number of prehospital fatalities that might have survived to
hospital if EMS response times were shortened, and 3) to determine whether the observed
effect of EMS response time on the risk of prehospital death is dependent on the rurality of
the crash environment. This project was approved by the Sunnybrook Health Sciences
Center research ethics board (Toronto, Ontario, Canada).
Study Population and Data Source
Data for all occupants of passenger vehicles who died within 30 days of an MVC in the US
between January 1, 2001 and December 31, 2015 were derived from the Fatality Analysis
Reporting System (FARS) of NHTSA [254]. Crashes involving heavy trucks, motorcycles,
off-road vehicles, cyclists or pedestrians, were excluded. To ensure reliability of the
81
exposure data, crashes that occurred in counties with >25% missing EMS notification or
arrival times were also excluded.
The FARS database is a population-based registry that collects data for all MVCs
occurring on public roads in the US that result in at least one fatality within 30 days [254].
Data collected includes variables at the occupant, vehicle, and crash levels. These include
granular time information such as the times of EMS notification and arrival on scene, as well
as the time and location (scene, en route, or in-hospital) of death.
Exposure
The exposure was EMS response time, defined as the time elapsed between EMS notification
and EMS arrival on scene. We excluded 9% of records due to missing EMS notification or
arrival time data.
Outcome Measures
The primary outcome of the study was prehospital death. Prehospital death was defined as
death at the scene or en route to hospital, and therefore represented death prior to receiving
in-hospital treatment.
Covariates
We considered occupant, vehicle, and crash characteristics that might confound the
relationship between EMS response time and the risk of prehospital death. Occupant
characteristics included age, gender, seat position, restraint use, personal airbag deployment,
ejection from the vehicle, and need for extrication. Vehicle characteristics included body
82
type (sedan-sized vs. van or pickup truck), model year, and the number of vehicle occupants.
Crash characteristics included variables that reflect the dynamics, severity, and circumstances
of the crash. These included multiple vehicle collision, vehicle roll over, the direction of
principal impact, roadway classification, maximum posted speed limit, pavement and surface
conditions, the time and year of the crash. Fewer than 2% of records were excluded due to
missing occupant data.
The rurality of the crash location was considered using county population density
[218, 221]. County population densities were calculated for each year using US census
population estimates and grouped into quartiles derived based on the study population (<50,
50-99, 100-299, ³300 people/mile2) [255].
Statistical Analysis
Standardized differences were used to compare occupant, vehicle, and crash characteristics
across quartiles of EMS response time [256]. Standardized differences were used because
standard statistical tests are sensitive to large sample sizes and may yield p-values <0.05
where no true difference exists. Standardized differences ³10% represented meaningful
differences between quartiles [257]. We then used multiple analytic techniques to achieve
the study objectives.
First, a hierarchical logistic regression model was used to estimate the risk-adjusted
relationship between EMS response time and the odds of prehospital death [245]. Random-
effects were included in the model to account for clustering of crashes within counties.
Elapsed EMS response time was modelled both in quartiles and as a continuous variable.
The latter approach made no underlying assumptions about the relationship being measured
83
and helped to identify meaningful thresholds in response time. Receiver operating
characteristic (ROC) curves were analyzed to identify the cut-off in EMS response time that
provided greatest discrimination in the odds of prehospital death [258, 259]. This threshold
was used to define ‘long’ (LRT) vs. ‘short’ (SRT) response times.
Second, we estimated the number of vehicle occupants who died in the prehospital
environment that might have survived to hospital if EMS response times were reduced.
Using the cut-off identified by the ROC curve analysis, the attributable risk of prehospital
death due to LRT vs. SRT was calculated. The number of prehospital deaths that might have
survived to hospital if all occupants had received SRT was then estimated.
Finally, to determine if effect of EMS response time on the risk of prehospital death
is dependent on the rurality of the crash environment, we evaluated effect modification
between EMS response time and county population density using interaction terms. As this
interaction was significant (p < 0.05), we report odds of prehospital death as a function of
response time stratified by rurality. The proportion of county-level variance in prehospital
mortality explained by EMS response times was estimated using the proportional change in
variance (PCV) [260]. The PCV was calculated using the following formula:
PCV = [(V1 – V2) / V1] x 100
where V1 and V2 were the county-level variance estimated from the hierarchical model
before and after inclusion of EMS response time respectively. To determine if the
contribution of EMS response times to prehospital mortality differs across the spectrum of
rurality, the PCV was calculated within deciles of population density representative of all US
counties.
All data were analyzed using SAS software (version 9.4, Cary, NC).
84
RESULTS
Study Population
Over the 15-year study period we identified 117,267 vehicle occupants who suffered fatal
injuries resulting from 105,924 crashes in 1,781 counties (57% of all US counties)
representative of 41 US states. Compared to those excluded, counties included in the final
cohort showed a significant over-representation of rural and wilderness counties with low
population density (Table 4.1).
Vehicle occupants were predominantly male (64%) front seat occupants (89%) of
sedan-sized vehicles (73%). The majority of crashes occurred on interstates or highways
(61%) in counties with a median population density of 95 people/mile2 (interquartile range
[IQR] 37 – 307 people/mile2. The median EMS response time was 9 minutes (IQR 6–14
minutes) and 58% of occupants died in the prehospital environment (57% died at the scene,
1% en route to hospital).
Table 4.2 compares occupant, vehicle, and crash characteristics across quartiles of
EMS response time. Longer response times were associated with younger, unrestrained
occupants, more often ejected from the vehicle. Crashes with longer response times were
more likely to be single vehicle crashes on high-speed motorways in counties with low
population density. Prehospital death was significantly more common among crashes in the
longest (Quartile 4: >14 minutes) vs. shortest quartile (Quartile 1: <6 minutes) of EMS
response time (65% vs. 50%).
The crude relationship between EMS response time and prehospital mortality is
shown in Figure 4.1A. The proportion of deaths that occurred in the prehospital
85
environment increased from 45% to 65% during the first 15 minutes following the crash and
then plateaued at times longer than 20 minutes.
EMS Response Time and Prehospital Mortality
Table 4.3 shows the results of the hierarchical logistic regression model for prehospital
death. Front seat position, ejection from the vehicle, and need for extrication were associated
with prehospital death, as were county roads, higher maximum posted speed limits, and
crashes after midnight. Occupant restraint use and newer vehicle models were predictive of
survival to hospital. Crashes in rural counties were strongly associated with death in the
prehospital environment.
After risk-adjustment, longer EMS response times were independently associated
with prehospital mortality (>14 minutes vs. <6 minutes; OR 1.41; 95%CI 1.35–1.46). When
response time was modelled as a continuous predictor (Figure 4.1B) the odds of prehospital
death increased during the first 10 minutes before reaching a plateau.
Analysis of ROC curves identified a threshold of 8 minutes as providing the best
discrimination in the odds of prehospital death between ‘long’ (LRT, ³8 minutes) and ‘short’
(SRT, <8 minutes) EMS response times. LRT was associated with a 27% increase in the
odds of prehospital death compared to SRT (OR 1.27; 95%CI 1.24–1.30).
Estimated Effect of Reducing EMS Response Times on Survival to Hospital
We calculated the attributable risk of prehospital death associated with LRT vs. SRT to be
6% (Table 4.4). Based on this value, we estimated that prehospital mortality among fatally-
injured occupants with LRT might have been reduced from 63% to 57% if all occupants had
86
received SRT. Therefore, during the study period 6.2% of occupants who died in the
prehospital environment (n=4,233) might have survived to hospital if all EMS response times
were shortened to <8 minutes.
Effect Modification Between EMS Response Time and Rurality
Finally, we sought to determine whether the observed effect of EMS response time on the
odds of prehospital mortality was dependent on the rurality of the crash location. There was
significant interaction between EMS response time and county population density in the
hierarchical model (p < 0.05). Specifically, long EMS response times were associated with a
significantly greater increase in the odds of prehospital death in counties with low population
density (Table 4.5). The proportion of prehospital fatalities that might have survived to
hospital if all EMS response times were shortened to <8 minutes was also greatest in rural (<
50 people/mile2: 8%) vs. urban (≥ 300 people/mile2: 3%) counties (Table 4.6).
Using the PCV, we estimated the proportion of county-level variance in prehospital
mortality attributable to EMS response times. When the PCV was examined across deciles
of population density reflective of all US counties, the proportion of variance in prehospital
mortality rates explained by EMS response times was strongly related to rurality, ranging
from >10% in the most rural counties, to <2% in the most urban (Figure 4.2).
DISCUSSION
In this retrospective study of MVC occupants with fatal injuries, we demonstrated a strong
relationship between EMS response time and the odds of death in the field. Using the
estimated cut-off of 8 minutes, we estimated 6.2% of prehospital deaths could be attributed
87
to ‘long’ response times. Finally, this association was significantly dependent upon the
rurality of the crash environment, with prolonged response times contributing to a greater
proportion of deaths in a dose-response fashion across the spectrum of increasingly rural
counties.
The results of this study shed light on an important observation in the epidemiology
of traffic-related deaths. MVCs are the most common cause of injury-related death in North
America [6]. While overall MVC mortality has declined [215, 252], the proportion of deaths
that occur in the prehospital environment has risen, approaching 60% [252]. Furthermore,
rural-urban disparities in crash mortality are well documented and prehospital mortality is
greatest in rural regions [30, 219, 221]. Our findings show that the risk of death in the field
is independently associated with both EMS response time and rurality, but that longer
response times also contribute significantly to observed higher rates of prehospital death in
rural counties.
There is biologic rationale for expecting that EMS response times will affect the risk
of prehospital death for crash occupants with time-dependent injuries. While many deaths at
the scene occur immediately following impact due to severe neurotrauma, asphyxia, or
hemorrhage [261], and are therefore not modifiable by post-crash interventions, there is a
category of patient at high risk for early death who might benefit from early contact with
EMS in the field. Early arrival of EMS at the scene allows for delivery of potentially-life
saving interventions, timely triage, and expedient transport to definitive care within an
integrated trauma system. By examining a cohort of occupants who ultimately died, our
analysis confirms that shorter EMS response times provide a greater opportunity for
88
occupants with the most critical injuries to reach hospital. At the trauma system-level, this
effect would likely translate into overall improved outcomes.
The results of this study have important implications trauma system resource
allocation. EMS are integral components of high-performing trauma systems, and EMS
response time has been identified as a quality indicator for prehospital trauma care [158].
However, there has been a paucity of data demonstrating a relationship between response
time at the trauma patient, let alone trauma system-level [262]. Due to effect modification of
the relationship between EMS response time and prehospital mortality by county rurality, we
found that a significantly greater proportion of prehospital deaths could be attributed to long
EMS response times in rural (< 50 people/mile2: 8.4%) vs. urban (≥ 300 people/mile2: 3.4%)
jurisdictions. Furthermore, after adjusting for occupant, vehicle, and crash variables in our
hierarchical logistic regression model, EMS response times accounted for more than 10% of
the variance in prehospital mortality between the most sparsely-populated counties,
compared <2% among the most densely-populated. Taken together, these data show that
response times not only influence the odds of prehospital death but contribute in a
meaningful way to rural-urban disparities in prehospital mortality. Therefore, trauma system
performance improvement efforts should consider strategies that incorporate the allocation of
resources toward shortening EMS response times, particularly in remote areas.
The current study offers several methodological strengths. Previous studies
examining the influence of EMS response time on trauma patient outcomes have been
challenged by the heterogeneity of the patient population [147, 149] or have been limited to
making crude unadjusted associations [147, 151]. In contrast, this study utilized a large
population-based sample of MVC fatalities drawing upon 57% of all US counties. The
89
multilevel analytic techniques used provided thorough risk-adjustment for crash-level factors
that might affect occupant risk, while allowing for modelling of county-level variability. The
latter approach acknowledges the organization of EMS that often exists at the county level.
This study also has several important limitations. First, our conclusions are
predicated on the assumption that crash occupants found dead at the scene are rarely
transported to hospital. While there is no published data to confirm this practice, many
regional EMS systems have protocols to facilitate declaration of death in the field. Our
analyses of time to death support the interpretation that occupants transported to hospital
were alive and thus, received active medical treatment. Second, 43% of all US counties were
not included in our analyses. However, the sample of 1,781 counties from 41 US states
represented the full spectrum of rurality, and therefore our findings are likely generalizable to
regional systems across the country. Third, we were not able to account for other system-
level factors related to the nature of, or access to, prehospital or in-hospital medical care near
the crash location. Therefore, it is plausible that longer EMS response times in rural regions
are correlated with differences in the resources and capabilities of first responders, overall
prolonged prehospital times, and more complex patterns of inter-facility transfer required to
reach definitive trauma care. Finally, while we cannot draw conclusions based on these data
related to overall mortality, our findings suggest that short EMS response times provide
MVC occupants with critical injuries an improved likelihood of survival to hospital, and
therefore likely confer an improved probability of survival to discharge.
Despite the stated limitations, the results of this study provide strong evidence that
EMS response times influence the risk of prehospital death among MVC occupants with
critical injuries. Prolonged EMS response times are an important contributor to known
90
disparities in prehospital mortality between rural and urban counties in the US. Therefore,
efforts to reduce EMS response times could result in greater access to in-hospital trauma care
for occupants at high risk of dying in the field. At the trauma system-level, such a strategy
could save lives.
CONCLUSIONS
Among MVC occupants with fatal injuries, EMS response time is strongly associated with
the risk of prehospital death. An important proportion of prehospital deaths are attributable
to prolonged response times. This effect is greatest in rural counties, where long response
times contribute to a significantly greater proportion of deaths in the field than in urban
counties. Trauma systems performance improvement efforts should consider strategies for
achieving shorter response times, particularly in rural regions.
91
CHAPTER 5:
THE RELATIONSHIP BETWEEN EMERGENCY MEDICAL SERVICE RESPONSE
TIME AND MOTOR VEHICLE CRASH MORTALITY: AN ANALYSIS OF 2,268
UNITED STATES COUNTIES
ABSTRACT
BACKGROUND: Deaths from MVCs are a leading public health problem. Despite the
evolution of organized systems of trauma care, significant regional variations in crash fatality
rates persist. We measured the association between EMS response time and MVC mortality
at the population-level in counties across the US.
METHODS: EMS activations to MVCs derived from NEMSIS were linked to rates of
MVC-related deaths derived from the FARS at the US county level (2013-2015). The
exposure was the median county EMS response time. Hierarchical negative binomial
regression modelling was used to examine the relationship between the median EMS
response time and county mortality rate, adjusted for measures of access to trauma care, state
traffic safety laws, and rurality. The population attributable fraction was calculated to
estimate the proportion of all crash fatalities that might be prevented if EMS response were
shortened. Analyses were stratified by county rurality.
RESULTS: We identified 2,214,480 ambulance responses to MVCs in 2,268 counties,
representative of 49 states. Longer EMS response time quartiles were significantly
92
associated with higher rates of MVC-related death (³12 vs. <7 minutes; MRR 1.45; 95%
confidence interval [95%CI] 1.32 – 1.61). Prolonged EMS response times were significantly
associated with crash mortality rates in both urban/suburban and rural/wilderness counties.
Taken together, we estimated that 2,042 passenger vehicle deaths per year (12% of all crash
fatalities) might be prevented if EMS response times were shortened to achieve the current
median values in urban/suburban (7 minutes) and rural/wilderness regions (10 minutes).
CONCLUSION: In this population-based analysis of 2,268 US counties, EMS response
times were significantly associated with crash mortality. A meaningful proportion of crash
fatalities are attributable to prolonged response times in both rural/wilderness and
urban/suburban counties. EMS response times should be evaluated in trauma system quality
improvement efforts.
This work is not yet published.
The following co-authors are acknowledged for their permission to include this manuscript:
N. Clay Mann, Stephanie A. Mason, Jason Buick, Paul Karanicolas, Sandro Rizoli, and Avery
B. Nathens.
93
INTRODUCTION
Motor vehicle crashes (MVCs) are a leading public health concern in the United States (US)[6].
Improvements to road infrastructure, vehicle design, and traffic safety legislation have led to
a decrease in crash mortality from 15.9 per 100,000 people in 1995 to 10.9 per 100,000 people
in 2015 [215]. The implementation of organized trauma systems has reduced deaths by
ensuring that patients with severe injuries have timely access to trauma care [45, 46].
However, significant disparities in access to high-quality trauma care remain [165, 166], with
meaningful differences in outcome [164]. Rates of MVC-related death vary by an order of
magnitude between states, from 3 to 25 deaths per 100,000 people [266]. While regional
differences in crash risk and severity likely contribute to this variation, the role of modifiable
trauma system-level factors is unclear.
Emergency medical service (EMS) response time, defined as time elapsed between
EMS notification and arrival on-scene, is a system-level factor with potential to impact
survival. EMS provide the critical link between injury and definitive care [63]. Early arrival
of EMS at the crash scene allows for stabilization of occupants with life-threatening injuries,
timely triage, and transport to hospital [41, 63], while delays could lead to greater risk of death.
We postulate that a relationship exists between EMS response time and crash mortality,
and that reducing the time to first medical contact in the field might decrease deaths due to
road traffic injuries at the system-level. Therefore, we measured the association between EMS
response times and MVC mortality across the US and estimated the number of fatalities that
might be prevented at the population-level if response times were shortened.
94
METHODS
Study Design
This study was a cross-sectional population-based analysis of MVC-related deaths within US
counties during 2013-2015. The specific objectives of the study were: (1) to determine the
association between EMS response time and MVC mortality at the US county-level, and (2)
to estimate the proportion of MVC-related deaths attributable to prolonged EMS response
times. This project was approved by the Sunnybrook Health Sciences Center research ethics
board (Toronto, Ontario, Canada).
Data Sources
EMS Response Time (Exposure)
Data related to ground EMS activations during 2013-2015 were provided by the National EMS
Information System [267]. NEMSIS is a federally funded project designed to standardize EMS
patient care reporting and facilitate collection of data for assessment of EMS systems of care
[240]. As of 2015, NEMSIS collected data related to EMS activations performed by EMS
agencies in 2,497 of 3,144 US counties and county-equivalents. We included all ground
ambulance responses for possible injury due to MVCs (Figure 5.1). EMS response times were
aggregated at the county-level. The exposure was defined as the county median EMS response
time, a measure of local EMS system responsiveness to traffic crashes. EMS activations with
missing response time were excluded (<0.2%). Counties with fewer than 5 EMS responses to
MVCs over the three-year study period were also excluded (4%).
95
MVC Mortality Rate (Outcome)
MVC-related deaths were derived from the Fatality Analysis Reporting System (FARS) of the
National Highway Traffic Safety Administration [268]. FARS is a population-based registry
that collects data related to all MVCs on public roads in the US that result in at least one fatality
within 30 days [269]. To reduce the potential for confounding, fatalities were limited to
occupants of passenger vehicles. Crashes involving heavy trucks, motorcycles, off-road
vehicles, cyclists, or pedestrians were excluded. The US Census Bureau provided intercensal
population estimates for each county and year [270]. The primary outcome was the county
rate of MVC mortality reported as deaths/100,000 person-years.
Linkage of EMS and MVC Mortality Data
The NEMSIS Technical Assistance Center performed linkage of crash mortality and EMS data
at the county-level using Federal Information Processing Standards (FIPS) codes. FIPS codes
were then replaced using a random identifier to preserve the anonymity of EMS agencies.
Counties where identifiable combinations of data might pose a risk to anonymity (for example,
those with unique MVC mortality rates or median EMS response times) were excluded from
the dataset (<3% of counties).
Potential Confounders
We considered several factors that might confound the relationship between EMS response
time and MVC mortality rate at the county-level. To account for vehicle occupant differences
96
known to be associated with risk of MVC death [271, 272], county mortality rates were
stratified by age (<15, 15-34, 35-64, ≥65 years) and sex.
Rurality, a known potential confounder [30, 214, 218, 219], was estimated using county
population density and rural-urban continuum codes (RUCCs) [273]. County population
densities were calculated using US census population estimates [270] and grouped into
quartiles (<16, 17-42, 43-108, ≥109 people/mile2). RUCCS were used to group counties into
four categories of rurality (urban, suburban, rural, wilderness) based on population and
proximity to metropolitan areas (Table 5.1).
EMS on-scene and transport times to hospital might also influence crash mortality.
These time intervals were derived from NEMSIS for patient transports from the scene of
MVCs where the final destination was a hospital. The median on-scene and transport times
for each county were then calculated. On-scene or transport times were missing in 1% of cases
and were imputed using a multiple imputation technique [186].
Because regional differences in EMS response time are likely to be correlated with
access to trauma resources [165], we accounted for the proximity of trauma centers [67] and
helicopter EMS (HEMS) [125, 138] to county populations. American College of Surgeons
(ACS) verified and state-designated level I and II trauma centers were identified from ACS
[274] and American Trauma Society [275] databases and geocoded by address using
geographic information system (GIS) software. Counties were then categorized by trauma
center proximity (within county, adjacent county, or no proximate trauma center).
Availability of HEMS was estimated for each county as the proportion of the
population within 25-mile flight circles of a HEMS base (corresponding to areas within 15-20
minute HEMS response times [137, 140]). HEMS bases were identified in the Atlas &
97
Database of Air Medical Services (ADAMS) [276] and geocoded by zip code. The geographic
distribution of populations within counties were geocoded using the centroids of census blocks
provided by the US Census Bureau [141].
Finally, state traffic safety laws known to influence the risk of fatal crashes were
included. These included the maximum posted speed limits on urban and rural highways and
interstates [199, 200], the primary enforcement of seatbelt laws [213, 277], administrative
licence revocation for alcohol-impaired driving [205, 209], and legislation prohibiting texting
while driving [278]. Traffic safety laws were obtained from the Insurance Institute for
Highway Safety [208] and the Governors Highway Safety Administration [279].
Statistical Analysis
To examine the generalizability of our results to all US counties, the characteristics of counties
included in the study cohort were compared to those of all US counties. Univariable analyses
compared county characteristics across quartiles of median EMS response time. The Wilcoxon
rank-sum and Kruskal-Wallis tests were used to compare continuous variables between groups,
while frequencies were compared using chi-squared tests.
A hierarchical negative binomial regression model was used to estimate the risk-
adjusted association between EMS response time and MVC mortality rate. This model was a
generalized linear mixed model that included a random-effects term to account for clustering
of counties within states.
To estimate the proportion of MVC-related deaths attributable to prolonged EMS
response times we used the population attributable fraction [280-282]. The population
attributable fraction is defined as the fraction of all cases (in this case, crash fatalities) in a
98
population that is attributable to a specific exposure (in this case, prolonged response times).
This was estimated using adjusted mortality rate ratios (MRRs) from the hierarchical model
using the formula [280]:
!"#(%) = !)*(+,,) − 1)!)*(+,,) − 1) + 1
*100%
where Pi and MRRi are the proportion of the total population and the MRR associated with the
ith quantile of response time respectively. In this study, the population attributable fraction
can be interpreted as the estimated fraction of all MVC-related deaths that might be prevented
if all counties achieved EMS response times in the shortest quantile.
Due to geographic and resource constraints, it is unlikely that rural counties would be
able to achieve EMS response times comparable to more densely-populated counties.
Therefore, we stratified our analysis of the population attributable fraction by rurality,
grouping counties into rural/wilderness and urban/suburban counties based on RUCCs.
Median EMS response times were calculated separately within rural/wilderness and
urban/suburban counties. The population attributable fraction was then calculated to estimate
the proportion of crash fatalities that might be prevented if EMS response times were shortened
to the median value within rural/wilderness and urban/suburban county groups, using the
benchmark appropriate to each context.
Geospatial analyses deriving variables for proximity of trauma centers and HEMS to
county populations were performed using ESRI ArcMap GIS software (version 10.5,
Redlands, CA). Statistical analyses were performed using SAS software (version 9.4, Cary,
NC).
RESULTS
99
During 2013-2015 there were 77,941,796 EMS activations in 2,497 counties (Figure 5.1). We
identified 2,214,480 ambulance responses to MVCs that met inclusion criteria in 2,268
counties (72% of all US counties). The median number of ambulance responses to MVCs for
each county was 229 (interquartile range [IQR], 73 – 697 responses).
Counties included in the study cohort were representative of 49 US states (including
the District of Columbia) and accounted for 75% of the total US population in 2015
(239,464,121 people). Included counties were characteristically similar to all US counties
(Table 5.2).
Table 5.3 compares the characteristics of counties across quartiles of EMS response
time. The median county EMS response time was 9 minutes (IQR 7 – 11 minutes). There was
a strong association between county EMS response times and rurality, with counties in the
longest quartile of EMS response times (³12 minutes) significantly more likely to be in rural
or wilderness regions with low population densities, compared to counties in the shortest
quartile of EMS response times (<7 minutes). Counties with longer EMS response times also
had significantly longer on-scene and transport times, lesser access to level I or II trauma
centers, and lower HEMS availability. Conversely, counties with short EMS response times
were associated with greater presence of protective state traffic safety laws.
The unadjusted MRR of crash fatalities between counties in the longest versus shortest
EMS response time quartile was 1.95 (95% confidence interval [CI] 1.72 – 2.22). There was a
near-linear relationship between EMS response time and MVC mortality rate at the county-
level (Figure 5.2).
Hierarchical Negative Binomial Model
100
We used a hierarchical negative binomial regression model to estimate the risk-adjusted
relationship between county median EMS response time and MVC mortality rate. The result
of this analysis is shown in Table 5.4. Longer EMS response times were significantly
associated with higher rates of MVC-related death (³12 vs. <7 minutes; MRR 1.45; 95%
confidence interval [95%CI] 1.32 – 1.61). Crash mortality was highest among males, ages
15 – 34 years, in rural counties with low population density. The presence of a level I or II
trauma center within a county was associated with lower MVC mortality (MRR 0.65; 95%CI
0.59 – 0.73), while higher maximum speed limits on highways and interstates were
associated with higher mortality. On-scene times, transport times, availability of HEMS
resources, and other traffic safety laws were not associated with MVC mortality.
Population Attributable Fraction: Rural/Wilderness and Urban/Suburban Counties
As outlined in statistical analysis, we estimated the population attributable fraction of crash
fatalities due to prolonged EMS response times in rural/wilderness and urban/suburban
counties (Tables 5.5A and 5.5B). The median EMS response time in rural/wilderness
counties was 10 minutes (IQR 8 – 12 minutes), compared to 7 minutes (6 – 9 minutes) in
urban/suburban counties. Within rural/wilderness counties, the proportion of crash fatalities
attributable to EMS response times ≥10 minutes was 10.4% (349 of 3,363 passenger vehicle
deaths). Within urban/suburban counties, the proportion of crash fatalities attributable to
EMS response times ≥7 minutes was 13.3% (1,693 of 12,735 passenger vehicle deaths).
Taken together, we estimated that 2,042 passenger vehicle deaths per year (12% of all crash
fatalities within the 2,268 counties evaluated) might be prevented if EMS response times
101
were shortened to achieve the current median values in rural/wilderness (10 minutes) and
urban/suburban (7 minutes) regions.
DISCUSSION
In this population-based study of 2,268 counties in 49 US states, increasing EMS response
times were strongly associated with higher rates of MVC mortality. This relationship was
independent of important potential confounders, including rurality, on-scene or transport
times, access to trauma centers or HEMS, and state traffic safety laws. Due to the strength of
association we estimated that a meaningful proportion of MVC-related deaths could be
attributed to prolonged EMS response times in both rural/wilderness and urban/suburban
environments. Therefore, efforts to achieve shorter EMS response times could represent a
viable means for reducing MVC mortality at the population-level in the US.
That shorter EMS response times could save lives is biologically inherent. Severe
trauma is a time-dependent condition. Prompt arrival of first responders at the crash scene
may allow for early stabilization of occupants with life-threatening injuries, timely triage,
and mobilization of the broader trauma system to achieve final disposition to definitive
trauma care [41]. Conversely, delays may confer a greater risk of death to those in need of
urgent medical attention. It follows that regional systems in which delays are more common
would exhibit higher rates of crash mortality after accounting for other regional differences.
It is important to place the results of this study in context of the current debate
surrounding the importance of response times as a measure of EMS performance. To date,
there is scarce evidence to show that response times influence trauma-related mortality [262].
Studies that have previously examined response time intervals were either hindered by a lack
102
of risk-adjustment [264, 283] or were not designed to make system-level recommendations
[149]. Furthermore, a greater emphasis on speed has potential to put the safety of first-
responding units at risk [41]. For these reasons, there has been a tendency away from using
response time standards as performance indicators.
The present study adds to this discussion by showing, for the first time, that EMS
response times are strongly associated with outcomes at the trauma system-level. Our
findings do not imply that EMS should simply drive faster, but rather that trauma systems
should be organized to ensure a quick response to MVCs. Approaches to shortening
response time intervals include achieving adequate numbers of first-responding units [159]
and optimizing the distribution of first-responding units in a dynamic and predictive fashion
[155]. In regions where geographic and resource constraints are especially limiting,
improvements to on-board telematics systems that predict the risk of severe injury may serve
to better triage deployment [284]. While these approaches are costly, our results demonstrate
that EMS response times should be considered in the evaluation of trauma system
effectiveness.
On-scene and transport times were not associated with crash mortality. This
observation is notable because counties with longer EMS response times also had longer on-
scene and transport times, reflecting differences in crash characteristics, EMS practices, and
greater travel distances in rural areas. The finding that only EMS response times were
significant after accounting for other regional differences suggests that early medical contact
with prehospital personnel is uniquely important to crash occupant survival at the trauma
system-level.
103
Greater access to trauma center care was associated with lower MVC mortality.
Counties with level I or II trauma centers had crash mortality rates 35% lower than those
with no nearby trauma center, providing further evidence that trauma centers reduce the risk
of injury-related death [64, 67]. Conversely, state laws allowing higher speed limits were
associated with greater mortality, emphasizing the dominant role of speed in the
epidemiology of fatal crashes [199, 200]. Contrary to previous reports [46, 205, 209, 213,
277], primary enforcement of seatbelt and administrative licence revocation laws were not
associated with crash mortality. These findings likely reflect the updated era in which the
data were derived. During 2013-2015 seatbelt use in the US surpassed 86% [285] and all
states had enforced the blood alcohol content limit of 0.08% [208]. Therefore, while seatbelt
and alcohol-impaired driving laws certainly prevent traffic deaths, the measurable effect is
likely diminished compared to previous time periods.
This study has several important limitations. First, there is potential for residual
confounding due to regional differences in crash severity and trauma processes of care.
Important processes of care include EMS notification, the quality of medical care delivered,
and patterns of inter-facility transfer. However, given that these factors are correlated with
rurality, the finding that our results were consistent between rural/wilderness and
urban/suburban counties is reassuring that our analyses are robust to these potential
confounders. Second, we were limited to include only 2,268 of 3,144 counties, with a slight
overrepresentation of more-densely populated areas with greater access to trauma center care.
However, the counties analyzed did represent the full spectrum of rurality and resources, and
therefore it is likely our results can be generalized to the majority of systems across the US.
Third, it is difficult to confirm the completeness of the EMS activations collected within
104
NEMSIS. However, NEMSIS captured an average of 26 million EMS activations annually
during the study period (74% of estimated 35 million nationally [286]). Therefore, it is
reasonable to infer that the data from which our exposure variable was derived was
comprehensive for the 72% of US counties included. Finally, the exposure variable itself
(median EMS response time) is an ecologic measure of EMS response capabilities.
Therefore, caution should be taken in drawing patient-level conclusions from our results
(ecological fallacy).
Taken in the appropriate context, our data provide strong evidence that EMS response
times influence crash mortality at the population-level. Therefore, EMS response times
should be evaluated in trauma system quality improvement efforts. In trauma systems where
MVC mortality is high, directing resources to achieve shorter EMS response times could
save lives.
CONCLUSIONS
In this population-based analysis of 2,268 US counties, EMS response times were
significantly associated with crash mortality. A meaningful proportion of crash fatalities are
attributable to prolonged response times in both rural/wilderness and urban/suburban
counties. EMS response times should be evaluated in trauma system quality improvement
efforts.
105
CHAPTER 6: SUMMARY AND FUTURE DIRECTIONS
6.1 SUMMARY OF WORK AND RECOMMENDATIONS
The work in this thesis has provided several actionable findings that should positively impact
the care of the injured patient, at both the hospital and trauma systems level. The clinical
epidemiological and statistical methodologies utilized will provide a framework for future
studies seeking to examine relationships between structures, processes, and outcomes of
trauma care.
First, we showed that among severely-injured patients presenting to level I and II
trauma centers participating in TQIP, the PROXY case definition (ED HR=0, ED SBP=0,
and GCS motor score=1) identifies patients who will die with extremely predictive accuracy
and construct validity. This definition performed in a superior fashion compared to other
definitions previously used for “dead on arrival” and is based on objective measures. This
finding is important because there has been significant heterogeneity between research
studies and trauma registries in how patients “dead on arrival” have been identified and
excluded. Hospitals also vary with respect to the number of patients “dead on arrival” they
receive, with potential to bias the measurement of trauma center performance. Our findings
confirmed that exclusion of patients who met the PROXY definition changed the
performance decile of 7% of trauma centers. This value increased to 33% when considering
risk-adjusted mortality in patients with shock, a cohort in which optimal care is arguably
most important. Therefore, the PROXY definition should be used in both studies of patient-
level outcomes and in external benchmarking analysis by TQIP to minimize bias due to
inclusion of patients who are essentially unsalvageable.
106
Second, utilizing the PROXY definition to minimize case ascertainment bias due to
inclusion of patients “dead on arrival”, we examined the influence of local prehospital times
on the measurement of trauma center performance among hospitals participating in TQIP.
This study utilized a novel linkage between TQIP and the NEMSIS database to provide an
ecologic measure of prehospital times for trauma at each trauma center. We found that while
hospitals with short prehospital times do have significantly higher rates of death in the ED,
prehospital times do not significantly alter overall trauma center risk-adjusted mortality.
This finding is important because participation in TQIP as a mechanism for external
benchmarking is now mandatory for verification by the ACS. Therefore, it is crucial that
participating trauma centers are not unfairly penalized for differences in case mix due to
differences in prehospital times. One particularly acute concern has been whether trauma
centers with short prehospital times receive a greater number of critically-injured patients
that might have otherwise died in the field if prehospital times were longer. The findings of
this study provide evidence that this effect is not borne out in the measurement of trauma
center risk-adjusted mortality. Therefore, this study suggests that local prehospital times
should not be included in external benchmarking analyses.
Third, using the FARS database we showed that EMS response times to fatal
MVCs significantly influence the likelihood of prehospital death for fatally-injured
occupants. This finding was consistent after adjusting for occupant, vehicle, and crash
characteristics, including the rurality of the crash location. Utilizing the attributable risk of
prehospital death due to longer EMS response times, we estimated that more than 6% of
patients who died in the prehospital environment might have survived to hospital if response
times were shortened. Furthermore, the effect of delayed EMS response on the odds of
107
prehospital death was significantly greater in rural locations. The proportional change in
variance in prehospital mortality due to EMS response times was greatest in the most rural
counties. These findings suggest that prolonged EMS response times in rural regions likely
contribute significantly to observed rural-urban disparities in prehospital death due to MVCs,
and may therefore be a significant contributor to regional disparities in overall crash
mortality.
Fourth, utilizing novel linkage of NEMSIS, FARS, and US census data, we showed
that EMS response times are significantly associated with crash mortality rates at the US
county population level. This finding is particularly significant because it is the first
population-based study to show that response times uniquely predict crash mortality, after
other EMS time intervals, measures of access to trauma care, state traffic safety laws, and
rurality are accounted for. Using the population attributable fraction, we showed that a
significant proportion of all crash deaths might be prevented if response times were
shortened. This finding was consistent in both rural and non-rural environments, indicating
that system-level quality improvement efforts should target EMS response times in both
contexts. Taken together, the third and fourth studies included in this thesis provide evidence
that EMS response time, a potentially-modified trauma system factor, should be adopted as a
quality indicator of prehospital trauma care.
6.2 Future Directions
The work provided in this thesis provides the groundwork for future research examining the
relationship between structure, process, and outcomes at the trauma system level.
108
One of the most important tools utilized in future trauma research will be the
aggregation and linkage of data from multiple sources to allow for bigger-picture
understanding of process-outcome relationships that could not previously be examined. Not
long ago, most data existed within silos of hospital-level registries and most trauma research
was conducted using such single-center data. Advances in data storage and coding have
allowed for these data to be aggregated, providing opportunity for initiatives such as TQIP to
perform analyses using multicenter data. While the mission of TQIP is to provide a means
for the reliable measurement and comparison of trauma center outcomes, it also provides
opportunity for higher-power studies of process-outcome relationships at the patient and
center-level that could not previously be achieved. Examples are in studies that have
examined center-level practices for intracranial pressure monitoring in patients with
traumatic brain injury [287] or the optimal agent for pharmacologic thromboprophylaxis
[288].
While the relationship between in-hospital processes of care and patient outcomes are
frequently the focus of new research, there is a relative lack of studies examining the
importance of prehospital system factors to regional variations in trauma mortality. Since a
significant proportion of trauma deaths, particularly MVC deaths, occur in the prehospital
environment this is an important problem. One of the historical reasons for this gap is the
relative lack of robust data sources aggregating data for prehospital trauma care. There is
also a disconnect between data sources collecting information related to prehospital
structures and processes of care, and those that capture patient outcomes. The linkage of data
sources such as NEMSIS, which capture variables related to the nature of EMS care, and
109
dataset such as FARS or TQIP will provide new opportunities to examine how specific
measures of prehospital trauma care are related to trauma mortality.
Another opportunity afforded through such data linkage is external validation of as-
yet unused variables. For example, TQIP captures prehospital time data at the patient level.
FARS also collects EMS response time data the crash level. However, the reliability of these
data is unknown. Linkage with the NEMSIS database at the hospital or county levels will
allow for the assessment of these variables, potentially supporting their use as local trauma
system measures in future studies.
A crucial focus of future trauma research will be to further determine what trauma
system factors contribute to regional differences in trauma mortality. The papers in this
thesis that examined the influence of EMS response time on MVC mortality have contributed
to this effort, however the overall picture is far more complex than these studies have been
able to account for. Future studies should combine multiple data sources to account for
regional differences in weather, geographic features, roadway infrastructure, all levels of
healthcare facility (including trauma centers of all levels, as well as non-trauma centers),
helicopter EMS, and multiple elements of the EMS care available including levels of
training, as well as prehospital times. GIS software will be essential to provide new variables
that describe inter-relationships between these factors. Factor analysis will provide insight
into which variables are important and will distill new composite metrics for comparing
regional trauma systems.
Future studies should also expand upon the hierarchical statistical methods used in
this thesis. The use of multilevel mixed models to account for hierarchical data structures
(such as vehicle occupants, within vehicles, within counties, within states) allows for the
110
cluster-level variance to be modelled. In doing so, so-called measures of variance such as the
intra-class correlation coefficient or proportional change in variance will provide insight into
how significantly specific variables contribute to differences within, and between, clusters.
An example is the role of specific trauma system measures in between-county differences in
trauma mortality. The applications for these measures of variance in studies of trauma
processes of care and outcomes have not been well explored. Finally, the use of hierarchical
modelling in studies of trauma system effectiveness should follow the TQIP model, in which
risk-adjusted trauma center outcomes are compared to provide insight into which practices
should be more broadly adopted. By comprehensively accounting for differences between
regional trauma systems (those both modifiable and not modifiable) in hierarchical models,
risk-adjusted trauma system performance can be measured and compared. These analyses
will yield important information that will serve to provide trauma leaders with the data
required to influence policy makers, to forward the nation-wide goal of minimizing
disparities in trauma care.
111
FIGURES
Figure 1.1 Locations of level I and II trauma centers in the United States (A), compared to population densities by county (B).
A
Traumacenterwithincounty
Traumacenterinadjacentcounty
LevelIorIItraumacenter
Noproximatetraumacenter
113
Figure 1.2 Estimate 25 mile flight circles from helicopter EMS bases, representing 15-20 minute helicopter response areas.
114
Figure 1.3 Forty year change in motor vehicle crash mortality in the United States
EMSResponseTimeandPrehospital Mortality
10.3
Most rural regions: >30 deaths/100,000 person yrs
116
Figure 1.5 Map of proportion of motor vehicle crash fatalities that occur in the prehospital environment (A), compared to the spectrum of population density (B), by US county. A
Higher % Prehospital Death
119
Figure 2.2 Impact of exclusion of PROXY patients on trauma center performance decile, stratified by patient type. The absolute change in trauma center performance decile is represented as the magnitude change (0, 1 or 2 deciles). Greatest impact was observed in performance for patients with penetrating injury and shock. More than 23% and 33% of trauma centers changed performance decile for penetrating injury and shock respectively, compared to <10% of centers for performance in blunt multisystem and all patients.
120
Figure 3.1 Forest plot of adjusted odds ratios (ORs) for overall hospital death by hospital emergency medical service (EMS) prehospital time (PHT) quartile, stratified by center-specific proportion of deaths occurring in the ED. PHT is the 90th percentile regional EMS time calculated for each trauma center. ORs and 95% confidence intervals (CIs) were estimated using the random-intercept multilevel logistic regression model with PHT quartile as the main exposure.
121
Figure 4.1 Crude (A) and risk-adjusted (B) relationship between EMS response time and prehospital mortality.
122
Figure 4.2 The proportional change in variance (PCV) in county prehospital mortality rates due to EMS response times. Deciles of county population density are representative of all United States counties. The PCV was greatest in the most rural counties (>10% in the lowest population density decile), and negligible in the most urban counties.
123
Figure 5.1 Flowchart showing the derivation of EMS activations to MVCs and the county cohort.
NEMSISDatabase2013-2015:77,941,796EMSactivations
2,497counties
11,345,605ambulance responsesto trauma
2,227,699ambulance responsestoMVCs2,436counties
2,222,887 responsestoMVCs2,337counties
EMSactivationstonon-trauma:• 66,596,191ambulanceresponses
MissingEMSresponse times:• 4,585ambulanceresponses(0.2%)
Countieswith<5EMSactivations:• 99counties(227ambulanceresponses)
EMSactivationstotraumaother thanMVCs:• 9,117,906ambulanceresponses
FinalCountyCohort:2,214,480 responsestoMVCs
2,268counties
Countiesexcludedduringdatalinkage:• 69counties(8,407ambulanceresponses)
124
Figure 5.2 Crude relationship between median EMS response time and county rate of MVC-related mortality. Best polynomial fit (solid line) and 95% confidence interval (dashed line) are shown. MVC, motor vehicle crash; EMS, emergency medical services.
125
TABLES
Table 2.1 Predictive Measures of Case Definitions for In-hospital Death Predictive
Measure All Patients (n=223,643)
Blunt (n=202,956)
Penetrating (n=20,687)
Mortality 7.2% 6.4% 14.8%
PROXY Specificity 99.99% 99.99% 99.90% PPV 99.09% 99.60% 98.56% NPV 93.83% 94.17% 90.30%
NSOL Specificity 99.52% 99.55% 99.16%
PPV 66.58% 55.84% 86.09%
NPV 93.63% 94.07% 89.10%
PHCA Specificity 99.94% 99.94% 99.90%
PPV 89.71% 86.38% 95.99%
NPV 93.27% 93.91% 86.93% NSOL, no signs of life; PHCA, pre-hospital cardiac arrest; PPV, positive predictive value; NPV, negative predictive value
126
Table 2.2 Patient Baseline, Injury Characteristics and Outcomes by Case Definition
Parameter PROXY (n=2,424)
NSOL (n=2,998)
PHCA (n=1,224)
Mean age (SD) 39 (18) 42 (20) 42 (20)
Male sex (%) 1,965 (81) 2,322 (77) 973 (79)
Transfer patients (%) 95 (4) 272 (9) 198 (16)
Injury mechanism (%)
Blunt 1,245 (51) 1,934 (65) 800 (65)
Penetrating 1,179 (49) 1,064 (35) 424 (35)
Case definition criteria (%)
PROXY 2,424 (100) 1,754 (59) 706 (58)
NSOL 1,754 (72) 2,998 (100) 621 (51)
PHCA 706 (29) 621 (21) 1,224 (100)
Initial ED vital signs (%)
Pulse Rate = 0 2,424 (100) 1,788 (60) 722 (59)
SBP = 0 2,424 (100) 1,807 (60) 727 (59)
GCS motor = 1 2,424 (100) 2,005 (67) 1,141 (93)
Shock in ED (%) 2,424 (100) 1,914 (64) 854 (70)
ED mortality (%) 2,101 (87) 1,732 (58) 718 (59)
In-hospital mortality (%) 2,402 (>99) 1,996 (67) 1,098 (90) NSOL, no signs of life; PHCA, pre-hospital cardiac arrest; SD, standard deviation; ED, emergency department; SBP, systolic blood pressure; GCS, Glasgow Coma Scale.
127
Table 2.3 Characteristics of Predicted Deaths and Unexpected Survivors by Case Definition
PROXY NSOL PHCA
Deaths (n=2,402)
Survivors (n=22) P Deaths
(n=1,996) Survivors (n=1,002) P Deaths
(n=1,098) Survivors (n=126) P
Male sex (%) 1,945 (81) 20 (91) 0.409 1,594 (80) 729 (73) <0.001 874 (80) 99 (79) 0.787
Mean age (SD) 39 (17) 41 (23) 0.840 41 (19) 46 (21) <0.001 42 (18) 47 (20) 0.004
Race (%) <0.001 <0.001 <0.001
White 1,088 (45) 6 (27) 942 (47) 539 (54) 560 (51) 98 (78)
Black 843 (35) 14 (64) 682 (34) 78 (8) 313 (29) 15 (12)
Other 359 (15) 2 (9) 286 (14) 379 (38) 152 (14) 7 (6)
Comorbidities (%) <0.001 <0.001 <0.001
0 1,169 (49) 10 (45) 875 (44) 141 (14) 0 (0) 0 (0)
1 966 (51) 3 (55) 860 (43) 547 (55) 756 (69) 37 (29)
³2 267 (11) 9 (41) 261 (13) 314 (31) 342 (31) 89 (71)
Transfer patients (%) 95 (4) 0 (0) 1.000 87 (4) 185 (18) <0.001 154 (14) 44 (35) <0.001
Injury mechanism (%) <0.001 <0.001 <0.001
Blunt 1,240 (52) 5 (23) 1,080 (54) 854 (85) 691 (63) 109 (87)
Penetrating 1,162 (48) 17 (77) 916 (46) 148 (15) 407 (37) 17 (13)
ISS (%) 0.731 <0.001 <0.001
9-15 443 (18) 5 (23) 347 (17) 476 (48) 161 (15) 24 (19)
16-24 477 (20) 5 (23) 399 (20) 366 (37) 202 (18) 38 (30)
25-47 1,049 (44) 10 (45) 902 (45) 155 (15) 545 (50) 57 (45)
48-75 433 (18) 2 (9) 348 (17) 5 (<1) 190 (17) 7 (6)
Multiple system AIS ³3 (%) 1,046 (44) 10 (45) 0.858 849 (43) 218 (22) <0.001 499 (45) 53 (42) 0.470
Severe injury AIS ≥3
Head 1,093 (45) 5 (23) 0.033 972 (49) 359 (36) <0.001 548 (50) 63 (50) 0.985
Chest 1,570 (65) 19 (86) 0.042 1271 (64) 489 (49) <0.001 700 (64) 72 (57) 0.145
Abdomen 513 (21) 4 (18) 1.000 396 (20) 110 (11) <0.001 205 (19) 15 (12) 0.061
Upper extremity 120 (5) 2 (9) 0.305 108 (5) 43 (4) 0.186 51 (5) 8 (6) 0.398
Lower extremity 483 (20) 4 (18) 1.000 396 (20) 239 (24) 0.011 228 (21) 15 (12) 0.018
Spine 188 (8) 2 (9) 0.689 160 (8) 47 (5) <0.001 135 (12) 28 (22) 0.002
Initial ED vital signs (%)
ED Pulse rate > 0 0 (0) 0 (0) 1.000 223 (11) 987 (99) <0.001 381 (35) 121 (96) <0.001 ED SBP > 0 0 (0) 0 (0) 1.000 207 (10) 984 (98) <0.001 375 (34) 122 (97) <0.001 ED GCS motor > 1 0 (0) 0 (0) 1.000 49 (2) 944 (94) <0.001 32 (3) 51 (40) <0.001
Total GCS – median (IQR) 3 (3–3) 3 (3-3) 0.748 3 (3-3) 15 (14-15) <0.001 3 (3-3) 3 (3-10) <0.001
NSOL, no signs of life; PHCA, pre-hospital cardiac arrest; SD, standard deviation; ISS, injury severity score; AIS, abbreviated injury scale; GCS, Glasgow coma scale; IQR, interquartile range
128
Table 2.4 Characteristics of Patients Meeting PROXY Definition Criteria by Mechanism of Injury
Blunt Penetrating
Parameter PROXY (n=1,245)
Non-PROXY (n=201,711) P PROXY
(n=1,179) Non-PROXY (n=19,508) P
Male sex (%) 909 (73) 128,865 (64) <0.001 1,056 (90) 17,271 (89) 0.310
Mean age (SD) 44 (18) 52 (22) <0.001 34 (15) 33 (14) 0.010
Race (%) <0.001 <0.001
White 754 (61) 151,738 (75) 340 (29) 7,211 (37)
Black 224 (18) 19,900 (10) 633 (54) 8,724 (45)
Other 209 (17) 23,129 (11) 152 (13) 2,918 (15)
Comorbidities (%) <0.001 <0.001
0 533 (43) 49,995 (25) 646 (55) 7,246 (37)
1 528 (42) 67,210 (33) 441 (37) 7,014 (36)
³2 184 (15) 84,546 (42) 92 (8) 5,248 (27)
Transfer patients (%) 70 (6) 66,987 (33) <0.001 25 (2) 3,783 (19) <0.001
Injury mechanism (%) <0.001 <0.001
MVC 739 (59) 71,559 (35) - -
Fall 146 (12) 86,560 (43) - -
Other blunt 360 (29) 43,529 (22) - -
Stabbing - - 143 (12) 6,975 (36)
Firearm - - 1,036 (88) 12,526 (64)
ISS (%)
<0.001
<0.001
9-15 243 (19) 98,388 (49) 205 (17) 12,132 (62)
16-24 227 (18) 66,905 (33) 255 (22) 3,713 (19)
25-47 511 (41) 34,383 (17) 548 (46) 3,419 (18)
48-75 264 (21) 2,035 (1) 171 (14) 244 (1)
Multiple system AIS ³3 (%) 711 (57) 41,223 (20) <0.001 345 (29) 3,381 (17) <0.001
Severe injury AIS ≥3
Head 708 (57) 82,421 (41) <0.001 390 (33) 2,886 (15) <0.001
Chest 841 (68) 68,876 (34) <0.001 748 (63) 8,050 (41) <0.001
Abdomen 253 (20) 15,493 (8) <0.001 264 (22) 4,753 (24) 0.125
Upper extremity 72 (6) 12,099 (6) 0.750 50 (4) 2,422 (12) <0.001
Lower extremity 414 (33) 50,210 (25) <0.001 73 (6) 3,943 (20) <0.001
Spine 138 (11) 25,072 (12) 0.151 52 (4) 783 (4) 0.501
Total GCS – median (IQR) 3 (3–3) 15 (14-15) <0.001 3 (3-3) 15 (15-15) <0.001
ED Mortality 1,096 (88) 807 (<1) <0.001 1,005 (85) 349 (2) <0.001
In-hospital Mortality 1,240 (>99) 11,766 (6) <0.001 1,162 (99) 1,893 (10) <0.001
SD, standard deviation; MVC, motor vehicle collision; ISS, injury severity score; AIS, abbreviated injury scale; GCS, Glasgow coma scale; ED, emergency department
129
Table 3.1 EMS Time Intervals by Injury Mechanism
Mechanism Total Prehospital Time (mins) Response Time (mins) Scene Time (mins) Transport Time (mins)
MVC 63 (±40) 14 (±21) 29 (±25) 30 (±21)
Other Blunt 58 (±38) 13 (±18) 25 (±22) 28 (±22)
Firearm 49 (±47) 11 (±26) 21 (±32) 23 (±19)
Stabbing 54 (±35) 12 (±5) 23 (±24) 26 (±24)
All values represent the 90th percentile time (standard deviation)
EMS, emergency medical service; mins, minutes; MVC, motor vehicle collision
130
Table 3.2 Comparison of Trauma Center Characteristics and EMS Time Intervals Across EMS Prehospital Time Quartiles
Parameter Quartile 1 (n=25)
Quartile 2 (n=31)
Quartile 3 (n=26)
Quartile 4 (n=31) P-value
PHT quartile range (minutes) <53 53 – 60 61 – 70 >70 -
Hospital Characteristics
Level of designation (%) <0.01
Level I 13 (52) 12 (39) 19 (73) 26 (84)
Level II 12 (48) 19 (61) 7 (27) 5 (16)
Academic status (%) 0.14
University 9 (36) 10 (32) 16 (62) 17 (55)
Community 14 (56) 14 (45) 8 (31) 11 (35)
Non-teaching 2 (8.0) 7 (23) 2 (7.7) 3 (9.7)
Hospital type (%) 0.06
For profit 6 (24) 8 (26) 1 (3.9) 3 (9.7)
Non-profit 19 (76) 23 (74) 25 (96) 28 (90)
Bed size (%) 0.11
≤ 600 16 (64) 24 (77) 13 (50) 16 (52)
> 600 9 (36) 7 (23) 13 (50) 15 (48)
Median number of patients (IQR) 849 (631 – 1,057) 875 (554 – 1,328) 1,134 (804 – 1,370) 1,079 (675 – 1,476) 0.11
EMS Time Intervals*
Median EMS time interval, minutes (IQR)
Response time 11 (10 – 12) 12 (11 – 14) 14 (13 – 15) 17 (15 – 23) <0.01
Scene time 24 (22 – 27) 27 (24 – 28) 29 (27 – 30) 31 (28 – 33) <0.01
Transport time 20 (16 – 23) 26 (24 – 27) 31 (28 – 34) 42 (38 – 52) <0.01
EMS, emergency medical service; PHT, 90th percentile total prehospital time; IQR, interquartile range *EMS time intervals are calculated as the center-specific 90th percentile time
131
Table 3.3 Comparison of Trauma Center Case Mix Across EMS Prehospital Time Quartiles
Parameter Quartile 1 (n=25)
Quartile 2 (n=31)
Quartile 3 (n=26)
Quartile 4 (n=31) P-value
PHT quartile range (minutes) <53 53 – 60 61 – 70 >70 -
Patient Characteristics Mean age (±SD) 52 (±7.0) 53 (±6.3) 50 (±7.5) 50 (±4.6) 0.13 Male gender, % (±SD) 65 (±8.8) 64 (±7.6) 66 (±8.7) 66 (±4.8) 0.57 Mean proportion of patients by race, % (±SD)
White 61 (±25) 70 (±23) 66 (±20) 74 (±16) 0.12 Black 22 (±21) 17 (±18) 24 (±20) 13 (±13) 0.14 Other 17 (±20) 13 (±17) 11 (±8.7) 12 (±13) 0.47
Comorbid illness, % (±SD) 0 56 (±13) 54 (±14) 58 (±11) 58 (±13) 0.62 1 23 (±6.2) 24 (±5.8) 24 (±4.1) 23 (±5.4) 0.77 ≥2 21 (±8.6) 22 (±9.3) 19 (±7.5) 19 (±8.2) 0.53
Insurance status Commercial 25 (±11) 21 (±10) 25 (±8.8) 26 (±10) 0.22 Non-commercial 65 (±19) 67 (±20) 66 (±15) 66 (±11) 0.99 Unknown or other 10 (±23) 12 (±24) 9.1 (±12) 8.1 (±11) 0.84
Injury Characteristics Mean proportion of injury by mechanism, % (±SD)
MVC 35 (±9.1) 36 (±8.6) 39 (±8.6) 43 (±9.7) 0.01 Fall 40 (±15) 42 (±13) 36 (±14) 35 (±12) 0.12 Other Blunt 11 (±3.5) 11 (±3.5) 12 (±4.2) 13 (±3.6) 0.26 Firearm 9.4 (±7.5) 7.0 (±5.8) 9.6 (±7.8) 6.3 (±3.5) 0.10 Stabbing 4.3 (±2.8) 3.2 (±2.0) 3.4 (±1.8) 3.4 (±1.6) 0.19
ISS > 25 12 (±3.7) 11 (±3.4) 12 (±3.5) 13 (±3.6) 0.07 Mean proportion with severe injury AIS ≥3, % (±SD)
Head 37 (±7.4) 36 (±4.8) 34 (±4.9) 34 (±3.9) 0.20 Chest 33 (±7.0) 33 (±6.4) 35 (±6.4) 36 (±6.8) 0.10 Abdomen 8.8 (±3.7) 7.8 (±2.6) 9.6 (±3.5) 9.0 (±3.0) 0.19 Lower Extremity 30 (±5.6) 29 (±6.8) 30 (±5.8) 29 (±5.2) 0.90
Blunt multisystem injury*, % (±SD) 15 (±4.2) 15 (±3.9) 17 (±3.8) 19 (±4.8) <0.01 Penetrating truncal injury†, % (±SD) 8.6 (±5.9) 5.9 (±4.0) 7.3 (±5.2) 5.7 (±2.7) 0.07
ED Characteristics Mean proportion with ED characteristics, % (±SD)
Shock (SBP < 90 mmHg) 4.0 (±1.6) 3.5 (±1.4) 3.9 (±1.9) 3.5 (±1.1) 0.57 Heart rate < 50 0.95 (±0.42) 0.81 (±0.35) 1.1 (±0.37) 0.88 (±0.54) 0.18 GCS motor ≤ 3 7.3 (±3.0) 6.9 (±2.1) 7.8 (±4.0) 7.1 (±2.1) 0.65 Assisted respiration 3.9 (±4.2) 5.6 (±3.9) 5.9 (±5.0) 6.2 (±5.3) 0.28
ED death, % (±SD) 1.3 (±0.70) 1.1 (±0.66) 1.1 (±0.69) 0.77 (±0.50) 0.01 Overall hospital death, % (±SD) 7.3 (±1.9) 7.4 (±1.4) 7.5 (±1.5) 6.6 (±1.4) 0.08 Proportion of deaths occurring in the ED, % (±SD) 18 (±9.1) 15 (±8.4) 14 (±7.1) 11 (±5.9) 0.01
PHT, 90th percentile total prehospital time; SD, standard deviation; MVC, motor vehicle collision; ISS, Injury Severity Score; AIS, Abbreviate Injury Scale; ED, emergency department; SBP, systolic blood pressure; GCS, Glasgow Coma Scale *Blunt multisystem injury is defined as blunt injury with ≥ 2 body regions with AIS ≥ 3 †Penetrating truncal injury is defined as injury by firearm or stabbing with AIS ≥ 3 to at least one of neck, chest or abdomen
132
Table 3.4 Patient Characteristics and Injury Severity Among ED Deaths
Parameter Quartile 1 (n=294)
Quartile 2 (n=347)
Quartile 3 (n=321)
Quartile 4 (n=278) P-value
PHT quartile range (minutes) <53 53 – 60 61 – 70 >70 -
Patient Characteristics
Mean age, years (±SD) 50 (24) 47 (22) 46 (22) 49 (20) 0.09
Male gender (%) 221 (75) 253 (73) 257 (80) 215 (77) 0.17
Race (%) <0.01
White 141 (48) 180 (52) 162 (50) 195 (70)
Black 97 (33) 101 (29) 123 (38) 42 (15)
Other 56 (19) 66 (19) 36 (11) 41 (15)
Insurance status (%) <0.01
Commercial 40 (14) 46 (13) 46 (14) 41 (15)
Non-commercial 191 (65) 265 (76) 224 (70) 221 (80)
Unknown or other 63 (21) 36 (10) 51 (16) 16 (5.8)
Injury Characteristics
Injury mechanism (%) 0.02
MVC 126 (43) 164 (47) 123 (38) 126 (45)
Fall 47 (16) 41 (12) 31 (9.7) 28 (10)
Other Blunt 3 (1.0) 14 (4.0) 15 (4.7) 14 (5.0)
Firearm 109 (37) 122 (35) 141 (44) 105 (38)
Stabbing 9 (3.1) 6 (1.7) 11 (3.4) 5 (1.8)
Mean ISS (±SD) 27 (11) 26 (13) 26 (12) 25 (12) 0.10
Severe injury AIS ≥3 (%)
Head 173 (59) 207 (60) 182 (57) 169 (61) 0.77
Chest 162 (55) 187 (54) 170 (53) 138 (50) 0.60
Abdomen 55 (19) 55 (16) 47 (15) 36 (13) 0.27
Lower Extremity 65 (22) 81 (23) 72 (22) 56 (20) 0.81
Blunt multisystem injury* (%) 97 (33) 125 (36) 96 (30) 96 (35) 0.39
Penetrating truncal injury† (%) 65 (22) 50 (14) 69 (22) 38 (14) 0.01
ED characteristics
Shock (SBP < 90 mmHg) (%) 81 (28) 121 (35) 127 (40) 102 (37) 0.02
Heart rate = 0 22 (7.5) 29 (8.4) 48 (15) 31 (11) 0.01
GCS motor = 1 224 (76) 273 (79) 230 (72) 217 (78) 0.75
Assisted respiration 122 (42) 185 (53) 174 (54) 163 (59) <0.01
Median time to death, minutes (IQR) 51 (17 – 141) 45 (17 – 118) 30 (11 – 101) 27 (14 – 91) <0.01
PHT, 90th percentile total prehospital time; SD, standard deviation; MVC, motor vehicle collision; ISS, Injury Severity Score; AIS, Abbreviate Injury Scale; ED, emergency department; SBP, systolic blood pressure; GCS, Glasgow Coma Scale; IQR, interquartile range *Blunt multisystem injury is defined as blunt injury with ≥ 2 body regions with AIS ≥ 3 †Penetrating truncal injury is defined as injury by firearm or stabbing with AIS ≥ 3 to at least one of neck, chest or abdomen
133
Table 3.5 Adjusted Model for Predictors of ED Death
Parameter Adjusted OR 95% CI Patient Characteristics Age (years)
≤ 35 Reference NA 36 - 65 1.56 1.34 – 1.82 > 65 4.49 3.66 – 5.51
Gender (male) 0.91 0.78 – 1.06 Race
White Reference NA Black 1.15 0.96 – 1.37 Other 1.07 0.88 – 1.30
Insurance status Commercial Reference NA Non-commercial 1.60 1.33 – 1.94 Other or unknown 1.54 1.18 – 2.03
Injury Characteristics Mechanism of injury
Fall Reference NA MVC 2.58 2.06 – 3.23 Other Blunt 1.27 0.88 – 1.83 Firearm 5.33 4.11 – 6.91 Stabbing 1.07 0.66 – 1.73
Survival risk ratio (single worst injury) 0.14 0.10 – 0.17 Severe injury AIS ≥3
Head 0.79 0.67 – 0.93 Face 0.50 0.33 – 0.78 Neck 1.16 0.77 – 1.75 Chest 1.53 1.31 – 1.78 Abdomen 0.60 0.50 – 0.73 Lower extremity 1.29 1.10 – 1.53
ED Characteristics Shock (SBP < 90 mmHg) 2.88 2.46 – 3.37 Heart rate < 50 (beats per minute) 5.29 4.22 – 6.63 GCS motor
1 – 2 Reference NA 3 – 4 0.31 0.25 – 0.40 5 – 6 0.03 0.02 – 0.04
Assisted respiration 1.34 1.13 – 1.59 Respiratory rate (breaths per minute)
≤12 Reference NA 13 - 20 0.51 0.44 – 0.60 >20 0.64 0.52 – 0.78
EMS Total Prehospital Time (PHT)
Quartile 1 (<53 minutes) 2.00 1.43 – 2.78 Quartile 2 (53 – 60 minutes) 1.54 1.13 – 2.11 Quartile 3 (61 – 70 minutes) 1.23 0.89 – 1.70 Quartile 4 (>70 minutes) Reference NA
EMS, emergency medical service; ED, emergency department; OR, odds ratio; CI, confidence interval; MVC, motor vehicle collision; AIS, Abbreviated Injury Scale; SBP, systolic blood pressure; GCS, Glasgow Coma Scale; PHT, 90th percentile total prehospital time The interaction term Assisted respiration*Respiratory rate was not statistically significant (P=0.13)
134
Table 4.1 Comparison of counties represented in study cohort compared to those excluded
Parameter Cohort (n = 1,781)
Excluded Counties (n = 1,361) P-value
States representeda 41 36
Region/Division (%) West Mountain Pacific Midwest East North Central West North Central North East New England Middle Atlantic South South Atlantic East South Central West South Central
10.3 2.5
13.3 22.4
2.8 3.1
11.3 11.0 23.3
7.1 9.0
14.7 16.1
1.3 6.9
28.4 12.4 4.1
<0.001
Rurality (%) Urban Suburban Rural Wilderness
31.4 9.7 36.2 22.7
44.6 9.9
28.0 17.6
<0.001
Population density, ppl/mile2 (%) <16 16 – 41 42 – 108 ≥109
29.3 26.9 24.7 19.2
19.5 22.0 26.2 32.4
<0.001
a Including the District of Columbia
135
Table 4.2 Baseline characteristics Response Time (minutes)
Parameter Shortest RT Quartile 1 Quartile 2 Quartile 3 Longest RT
Quartile 4 Standardized
Difference (%)a EMS Response Time (minutes) <6 6 – 9 10 – 14 >14 Number of occupants 29,037 32,984 27,050 28,196 Occupant Characteristics Age, median years (IQR) Age ≥ 65 years (%)
38 (22 – 61) 21.6
36 (22 – 56) 17.2
35 (22 – 54) 14.9
34 (21 – 53) 13.8
16.2 20.5
Male gender (%) 62.5 64.0 64.1 64.4 4.1 Driver (%) 69.8 70.5 69.6 65.1 10.1 Front seat position (%) 89.9 90.0 89.0 86.9 9.2 Restraint use (%) 51.1 48.7 44.9 42.6 17.1 Airbag deployed (%) 40.2 40.2 36.5 33.6 13.8 Ejected from vehicle (%) 24.4 27.9 31.6 33.6 20.4 Extrication required (%) 31.6 31.2 28.8 24.5 15.9 Vehicle Characteristics Vehicle body type (%)
Sedan-sized Van or pickup
77.5 22.5
74.9 25.1
72.0 28.0
67.4 32.6
15.9
Year of model (%) 2000 – 2016 1990 – 1999 Before 1990
40.2 48.3 11.5
42.1 47.1 10.7
41.8 47.3 10.9
43.3 46.2 10.5
6.2
Number of occupants in vehicle (%) 1 2 >2
50.6 28.5 20.9
50.4 28.0 21.6
48.4 27.6 24.0
41.9 29.1 29.1
20.9
Crash Characteristics Multiple vehicle crash (%) 52.3 46.3 41.0 37.5 30.1 Vehicle rolled over (%) 29.2 36.1 41.1 45.9 35.0 Most harmful impact to vehicle (%)
Motor vehicle in transit Rollover Tree Utility pole Other
47.7 19.5 12.6 4.9 15.4
42.1 25.5 14.2 3.9 14.3
37.7 30.5 15.2 2.9 13.6
34.6 36.3 14.2 2.0 12.9
42.4
Direction of principal impact (%) Front Right side Left side Rear Other
51.6 15.6 16.8 4.9 11.1
51.6 14.1 15.2 4.7 14.4
52.6 13.2 12.8 4.0 17.4
50.9 11.8 10.4 3.9 23.1
36.3
Roadway classification (%) Interstate or highway County road Local street Other
57.1 13.2 25.8 3.9
60.0 19.1 16.9 3.9
62.6 22.4 10.9 4.1
65.0 20.8 7.9 6.4
51.1
Speed limit, mph (%) < 35 35 – 55 > 55 or no limit
12.2 70.7 17.1
6.5 70.0 23.5
4.1 67.9 28.0
3.7 60.7 35.7
50.3
Unpaved road (%) 5.9 7.6 8.2 10.9 18.3 Time of collision (%)
0600 – 1159 1200 – 1759 1800 – 2359 0000 - 0559
19.9 33.4 28.2 18.5
20.1 30.9 28.1 20.9
20.3 30.2 27.4 22.1
19.9 29.9 28.5 21.7
9.4
136
County population density, ppl/mile2 (%) <50 50 – 99 100 – 299 ≥300
20.7 16.1 24.5 38.7
25.6 18.8 25.3 30.4
34.9 22.2 23.5 19.5
49.9 19.9 18.4 11.8
80.0
Mortality Outcomes Prehospital death (%) 49.5 57.0 62.2 65.3 32.3 a Standardized differences calculated between longest and shortest RT quartiles; values ≥ 10% represent meaningful differences
137
Table 4.3 Hierarchical logistic regression model for prehospital death Parameter Odds Ratio 95% CI Response time (minutes)
< 6 6 – 9 10 – 14 > 14
Reference
1.19 1.33 1.41
NA
1.15 – 1.23 1.28 – 1.38 1.35 – 1.46
Occupant Characteristics Age of occupant (years)
< 15 15 – 34 35 – 64 ≥ 65
Reference
1.46 1.48 0.69
NA
1.37 – 1.56 1.38 – 1.58 0.64 – 0.74
Male gender 1.07 1.04 – 1.10 Driver 1.19 1.14 – 1.23 Front seat (vs. back) 1.15 1.09 – 1.22 Restraint use 0.94 0.91 – 0.96 Airbag deployed 1.02 0.99 – 1.05 Ejected from vehicle 1.20 1.16 – 1.25 Extrication required 1.40 1.36 – 1.45 Vehicle Characteristics Vehicle body type
Sedan-sized Van or pickup
Reference
0.97
NA
0.95 – 1.00
Year of model Before 1990 1990 – 1999 2000 – 2016
Reference
0.95 0.92
NA
0.91 – 0.99 0.88 – 0.97
Number of occupants in vehicle (%) 1 2 >2
Reference
1.24 1.55
NA
1.20 – 1.29 1.48 – 1.62
Crash Characteristics Multiple vehicle crash 1.28 1.20 – 1.38 Vehicle rolled over 1.50 1.44 – 1.57 Most harmful impact to vehicle
Motor vehicle in transit Rollover Tree Utility pole Other
Reference
0.85 1.38 1.04 1.13
NA
0.79 – 0.93 1.28 – 1.50 0.94 – 1.14 1.05 – 1.22
Direction of principal impact Front Right side Left side Rear Other
Reference
1.05 0.99 0.82 0.91
NA
1.01 – 1.09 0.95 – 1.03 0.77 – 0.87 0.87 – 0.95
Roadway classification Interstate or highway County road Local street Other
Reference
1.08 0.87 0.97
NA
1.05 – 1.13 0.84 – 0.91 0.91 – 1.03
Speed limit (mph) < 35 35 – 55 > 55
Reference
1.27 1.60
NA
1.20 – 1.34 1.50 – 1.71
Unpaved road (vs. paved) 0.93 0.88 – 0.97 Time of collision
0600 – 1159 1200 – 1759 1800 – 2359 2400 – 0559
Reference
0.90 1.06 1.57
NA
0.87 – 0.94 1.02 – 1.10 1.50 – 1.63
138
County population density (ppl/mile2) <50 50 – 99 100 – 299 ≥300
1.60 1.28 1.18
Reference
1.49 – 1.73 1.18 – 1.39 1.09 – 1.28
NA Year of collision 1.01 1.00 – 1.01 c-statistic, 0.66
139
Table 4.4 Estimation of additional survivors to hospital if response times were shortened to <8 minutes
Cohort: All crash occupants
Response Time N PHDs PHD
rate Adjusted
OR NNH Attributable Risk
Achievable PHD rate
Achievable PHDs
<8 minutes 46,657 24,087 0.52 1.00 - - - 24,087
≥8 minutes 70,610 44,335 0.63 1.27 17 0.06 0.57 40,102
Observed PHDs = 68,422 Achievable PHDs = 64,189 Prehospital deaths transported to hospital alive if response time minimized = (68,422 – 64,189) = 4,233 (6.2% of prehospital deaths)
140
Table 4.5 Effect modification by county rurality
Odds of Prehospital Death (95%CI)
Population density (people/mile2) <50 50 - 99 100 - 299 ≥300
Response time (mins) < 6 6 – 9 10 – 14 > 14
Reference
1.25 (1.17 – 1.34) 1.44 (1.34 – 1.55) 1.58 (1.48 – 1.70)
Reference
1.23 (1.14 – 1.34) 1.41 (1.30 – 1.53) 1.40 (1.29 – 1.53)
Reference
1.21 (1.13 – 1.29) 1.33 (1.24 – 1.44) 1.33 (1.22 – 1.44)
Reference
1.11 (1.05 – 1.18) 1.19 (1.10 – 1.27) 1.27 (1.17 – 1.39)
CI, confidence interval
141
Table 4.6 Estimation of additional survivors to hospital with shorter response time stratified by county rurality
Cohort: Counties with population density < 50 people/mile2
Response Time N PHDs PHD
rate Adjusted
OR NNH Attributable Risk
Achievable PHD rate
Achievable PHDs
<8 minutes 10,174 5,955 0.59 1.00 - - - 5,955
≥8 minutes 27,787 18,953 0.68 1.36 13 0.08 0.61 16,861
Observed PHDs = 24,908 Achievable PHDs = 22,816 Prehospital deaths transported to hospital alive if response time minimized = (24,908– 22,816) = 2,092 (8.4% of
prehospital deaths)
Cohort: Counties with population density 50 – 99 people/mile2
Response Time N PHDs PHD
rate Adjusted
OR NNH Attributable Risk
Achievable PHD rate
Achievable PHDs
<8 minutes 7,796 4,169 0.53 1.00 - - - 4,169
≥8 minutes 14,685 9,202 0.63 1.31 15 0.07 0.56 8,217
Observed PHDs = 13,371 Achievable PHDs = 12,386 Prehospital deaths transported to hospital alive if response time minimized = (13,371 – 12,386) = 985 (7.4% of prehospital
deaths)
Cohort: Counties with population density 100 – 299 people/mile2
Response Time N PHDs PHD
rate Adjusted
OR NNH Attributable Risk
Achievable PHD rate
Achievable PHDs
<8 minutes 11,571 5,955 0.51 1.00 - - - 5,955
≥8 minutes 15,403 9,304 0.60 1.26 18 0.06 0.55 8,437
Observed PHDs = 15,259 Achievable PHDs = 14,392 Prehospital deaths transported to hospital alive if response time minimized = (15,259 – 14,392) = 867 (5.7% of prehospital
deaths)
Cohort: Counties with population density ≥ 300 people/mile2
Response Time N PHDs PHD
rate Adjusted
OR NNH Attributable Risk
Achievable PHD rate
Achievable PHDs
<8 minutes 17,116 8,008 0.47 1.00 - - - 8,008
≥8 minutes 12,735 6,876 0.54 1.18 25 0.04 0.50 6,363
Observed PHDs = 14,884 Achievable PHDs = 14,371 Prehospital deaths transported to hospital alive if response time minimized = (14,884 – 14,371) = 513 (3.4% of prehospital
deaths)
142
Table 5.1 Definition of County Rurality using Rural-Urban Continuum Codes (2013 edition)
RUCC Category RUCC Description Rurality Category
1 Counties in metro areas of 1 million population or more Urban
2 Counties in metro areas of 250,000 to 1 million population Urban
3 Counties in metro areas of fewer than 250,000 population Urban
4 Urban population of 20,000 or more, adjacent to a metro area Suburban
5 Urban population of 20,000 or more, not adjacent to a metro area Suburban
6 Urban population of 2,500 to 19,999, adjacent to a metro area Rural
7 Urban population of 2,500 to 19,999, not adjacent to a metro area Rural
8 Completely rural or less than 2,500 urban population, adjacent to metro area Wilderness
9 Completely rural or less than 2,500 urban population, not adjacent to metro area Wilderness
RUCC, rural-urban continuum code
143
Table 5.2 Comparison of counties represented in study cohort to all US counties
Parameter Cohort (n = 2,268)
All US counties (n = 3,144) P-value
States representeda 49 51
Region/Division (%) West Mountain Pacific Midwest East North Central West North Central North East New England Middle Atlantic South South Atlantic East South Central West South Central
10.2 5.1
13.4 22.7
2.5 6.5
24.2 8.4 7.1
8.9 5.3
13.9 19.7
2.1 4.8
18.7 11.6 15.0
<0.001
Rurality (%) Urban Suburban Rural Wilderness
39.1 10.1 32.2 18.6
37.1 9.7
32.6 20.5
0.252
Population density, people/mile2 (%) <16 17 – 42 43 – 108 ³109
22.9 25.4 24.3 27.5
25.0 25.2 24.9 24.9
0.114
Proximity to level I/II trauma center (%) Within county Adjacent county No proximate trauma center
12.0 36.4 51.6
10.2 35.1 54.7
0.035
Population within 25 miles of HEMS base, median % (IQR) 65 (4 – 99) 65 (3 – 99) 0.737
Total population in 2015 239,464,121 321,391,018
MVC-related deaths in 2015 b (passenger vehicles) 16,098 22,132
Mortality rate (deaths/100,000 person-years) 6.7 6.9
a Including the District of Columbia b Occupants of passenger vehicles as defined in study cohort (national rate for all deaths in 2015 was 10.3) HEMS, helicopter emergency medical services; IQR, interquartile range; MVC, motor vehicle crash
144
Table 5.3 Comparison of County Characteristics Across Quartiles of EMS Response Time
Median EMS Response Time (minutes)
<7 7 – 8 9 – 11 ³12
Counties, n 480 629 656 503 P-value
County Rurality Rural-urban continuum (%)
Urban Suburban Rural Wilderness
60.8 14.2 21.3 3.8
51.0 13.4 28.5 7.2
27.9 8.4
44.7 19.1
17.9 4.6
31.0 46.5
<0.001
Population density, people/mile2 (%) <16 17 – 42 43 – 108 ³109
6.5 14.0 24.4 55.2
8.4
21.6 27.7 42.3
24.1 36.0 29.1 10.8
55.3 27.0 13.5 4.2
<0.001
EMS time intervals, median minutes (IQR)
On-scene time Transport time
16 (14 – 17) 11 (9 – 14)
17 (15 – 19) 13 (11 – 18)
17 (15 – 20) 17 (13 – 22)
19 (16 – 22) 23 (15 – 32)
<0.001 <0.001
Measures of Access to Definitive Care Proximity to level I/II trauma center (%)
Within county Adjacent county No proximate trauma center
27.5 33.3 39.2
15.1 41.2 43.7
2.6
38.4 59.0
5.6
30.6 63.8
<0.001
Population within 25 miles of HEMS base, median % (IQR) 97 (37 – 100) 85 (20 – 100) 52 (1 – 93) 13 (0 – 71) <0.001
State Traffic Safety Laws
Maximum speed limit >65 mph (%) Urban highways/interstates Rural highways/interstates
37.5 85.2
45.8 88.4
44.1 91.3
45.7 91.5
0.024 0.003
Primary enforcement of seatbelt laws (%) 77.7 76.6 70.1 51.7 <0.001
Administrative license suspension (%) 84.8 82.8 84.3 80.1 0.183
Text messaging ban (%) 90.8 89.0 87.8 79.5 <0.001
Total Population 123,029,515 77,014,040 24,786,836 14,633,730
MVC-related deaths (passenger vehicles) 6,031 5,233 3,097 1,737
Mortality rate (deaths/100,000 person-years) 4.9 6.8 12.5 11.9
Crude Mortality Rate Ratio (95% CI) Reference 1.27 (1.12 – 1.44) 1.63 (1.45 – 1.84) 1.95 (1.72 – 2.22) <0.001
EMS, emergency medical services; IQR, interquartile range; HEMS, helicopter EMS; mph, miles per hour; MVC, motor vehicle crash; CI, confidence interval
145
Table 5.4 Association Between County EMS Response Time and MVC Mortality Rate
Parameter Mortality Rate Ratio 95% CI P-value County EMS Response Time Median EMS response time (minutes)
<7 7 – 8 9 – 11 ³12
Reference
1.23 1.33 1.45
NA
1.14 – 1.33 1.23 – 1.44 1.32 – 1.61
<0.001
Population Demographic Groups
Male sex 1.60 1.52 – 1.68 <0.001
Age (years) <15 15 – 34 35 – 64 ≥65
Reference
7.39 4.37 5.71
NA
6.88 – 7.94 4.07 – 4.69 5.32 – 6.14
<0.001
County Rurality
Rural-urban continuum Urban Suburban Rural Wilderness
Reference
1.03 1.17 1.26
NA
0.93 – 1.13 1.08 – 1.26 1.13 – 1.39
<0.001
Population density, people/mile2 <16 17 – 42 43 – 108 ³109
1.92 1.49 1.36
Reference
1.68 – 2.21 1.35 – 1.65 1.25 – 1.48
NA
<0.001
EMS Time Intervals (per 1 minute increase) On-scene time Transport time
1.00 1.00
1.00 – 1.01 1.00 – 1.01
0.364 0.274
Measures of Access to Care
Proximity to level I/II trauma center Within county Adjacent county No proximate trauma center
0.65 0.98
Reference
0.59 – 0.73 0.92 – 1.04
NA
<0.001
Population within 25 miles of HEMS base (per 1% increase) 1.00 1.00 – 1.00 0.278
Traffic Safety Laws
Maximum speed limit >65 mph Urban highways/interstates Rural highways/interstates
1.17 1.42
1.01 – 1.37 1.11 – 1.80
0.044 0.005
Primary enforcement of seatbelt laws 0.87 0.75 – 1.01 0.066
Administrative license suspension 0.91 0.72 – 1.14 0.399
Text messaging ban 0.93 0.79 – 1.09 0.344
MVC, motor vehicle crash; CI, confidence interval; EMS, emergency medical services; NA, not applicable; HEMS, helicopter EMS; mph, miles per hour
146
Table 5.5A Population Attributable Fraction due to Prolonged EMS Response Times in Rural/Wilderness Counties
EMS Response Time (minutes) Population Proportion of
Population, Pi MRR PAF Deaths Mortality Rate (deaths/100,000 person-years)
<10 10,394,760 0.52 Reference Reference 1,706 16.4
≥10 9,658,389 0.48 1.24 0.104 1,657 17.2
Totals: 20,053,149 1.00 10.4% 3,363 16.0
Deaths attributable to prolonged EMS response times among rural/wilderness counties = 0.104 x 3,363 = 349 deaths
Table 5.5B Population Attributable Fraction due to Prolonged EMS Response Times in Urban/Suburban Counties
EMS Response Time (minutes) Population Proportion of
Population, Pi MRR PAF Deaths Mortality Rate (deaths/100,000 person-years)
<7 120,613,977 0.55 Reference Reference 5,649 4.7
≥7 98,796,995 0.45 1.34 0.133 7,086 7.2
Totals: 219,410,972 1.00 13.3% 12,735 5.9
Deaths attributable to prolonged EMS response times among urban/suburban counties = 0.133 x 12,735 = 1,693 deaths
147
REFERENCES
1. Merriam-Webster. Merriam-Webster Dictionary 2017 [Available from: https://www.merriam-webster.com/dictionary/trauma. 2. Krug EG, Sharma GK, Lozano R. The global burden of injuries. American journal of public health. 2000;90(4):523-6. 3. Baker SP ONR, Karpf RS. The Injury Fact Book. Lexington, Mass: Lexington Books; 1984 1984. 4. LS R. Injury Epidemiology. New York: Oxford University Press; 1998 1984. 5. Mattox KL ME, Feliciano DV. Trauma: McGraw-Hill Medical; 2013 1984. 6. Prevention CfDCa. Injury Prevention and Control: Data and Statistics. 2015 [cited 2017. Available from: https://www.cdc.gov/injury/wisqars/facts.html. 7. ICD-9-CM FOS. The International Classification of Diseases, 9th Revision, Clinical Modification. 6th Edition. 2009 [cited 2017. Available from: https://www.cdc.gov/injury/wisqars/facts.html. 8. Hemmila MR, Nathens AB, Shafi S, Calland JF, Clark DE, Cryer HG, et al. The Trauma Quality Improvement Program: pilot study and initial demonstration of feasibility. J Trauma. 2010;68(2):253-62. doi: 10.1097/TA.0b013e3181cfc8e6. 9. Newgard CD, Fildes JJ, Wu L, Hemmila MR, Burd RS, Neal M, et al. Methodology and analytic rationale for the American College of Surgeons Trauma Quality Improvement Program. J Am Coll Surg. 2013;216(1):147-57. doi: 10.1016/j.jamcollsurg.2012.08.017. Epub Oct 11. 10. Medicine AftAoA. The Abbreviated Injury Scale: Overview 2016 [cited 2017. Available from: https://www.aaam.org/abbreviated-injury-scale-ais/. 11. trauma.org. The Abbreviated Injury Scale 2017 [cited 2017. Available from: http://www.trauma.org/archive/scores/ais.html. 12. Copes WS SW, Champion HR, Bain LW. Progress in Characterising Anatomic Injury. Proceedings of the 33rd Annual Meeting of the Association for the Advancement of Automotive Medicine. 1989;Baltimore, MA. 13. Haas B, Xiong W, Brennan-Barnes M, Gomez D, Nathens AB. Overcoming barriers to population-based injury research: development and validation of an ICD10-to-AIS algorithm. Canadian journal of surgery Journal canadien de chirurgie. 2012;55(1):21-6. 14. Baker SP, O'Neill B, Haddon W, Jr., Long WB. The injury severity score: a method for describing patients with multiple injuries and evaluating emergency care. The Journal of trauma. 1974;14(3):187-96. 15. trauma.org. The Injury Severity Score 2017 [cited 2017. Available from: http://www.trauma.org/archive/scores/iss.html. 16. Trunkey DD. Trauma. Accidental and intentional injuries account for more years of life lost in the U.S. than cancer and heart disease. Among the prescribed remedies are improved preventive efforts, speedier surgery and further research. Scientific American. 1983;249(2):28-35. 17. Sauaia A, Moore FA, Moore EE, Moser KS, Brennan R, Read RA, et al. Epidemiology of trauma deaths: a reassessment. The Journal of trauma. 1995;38(2):185-93.
148
18. Demetriades D, Kimbrell B, Salim A, Velmahos G, Rhee P, Preston C, et al. Trauma deaths in a mature urban trauma system: is "trimodal" distribution a valid concept? J Am Coll Surg. 2005;201(3):343-8. 19. Acosta JA, Yang JC, Winchell RJ, Simons RK, Fortlage DA, Hollingsworth-Fridlund P, et al. Lethal injuries and time to death in a level I trauma center. J Am Coll Surg. 1998;186(5):528-33. 20. Department for the Management of Noncommunicable Diseases, Disability VaIP. Injuries and violence: the facts 2014 2014 [Available from: http://www.who.int/violence_injury_prevention/media/news/2015/Injury_violence_facts_2014/en/. 21. Organization WH. Disease Burden: Estimates for 2000-2015 2017 [cited 2017. Available from: http://www.who.int/healthinfo/global_burden_disease/estimates/en/index2.html. 22. Organization WH. Road Traffic Injuries 2017 [cited 2017. Available from: http://www.who.int/mediacentre/factsheets/fs358/en/. 23. Hofman K, Primack A, Keusch G, Hrynkow S. Addressing the growing burden of trauma and injury in low- and middle-income countries. American journal of public health. 2005;95(1):13-7. 24. Canada S. Leading causes of death 2014 [cited 2017. Available from: http://www.statcan.gc.ca/tables-tableaux/sum-som/l01/cst01/hlth36a-eng.htm. 25. Information CIfH. Trauma and Injuries: Ontario Death Dataset 2006-2007 2006 [cited 2017. Available from: https://www.cihi.ca/en/trauma-and-injuries. 26. Peek-Asa C, Zwerling C, Stallones L. Acute traumatic injuries in rural populations. American journal of public health. 2004;94(10):1689-93. 27. Rutledge R, Fakhry SM, Baker CC, Weaver N, Ramenofsky M, Sheldon GF, et al. A population-based study of the association of medical manpower with county trauma death rates in the United States. Ann Surg. 1994;219(5):547-63; discussion 63-7. 28. Rogers FB, Shackford SR, Hoyt DB, Camp L, Osler TM, Mackersie RC, et al. Trauma deaths in a mature urban vs rural trauma system. A comparison. Archives of surgery (Chicago, Ill : 1960). 1997;132(4):376-81; discussion 81-2. 29. Grossman DC, Kim A, Macdonald SC, Klein P, Copass MK, Maier RV. Urban-rural differences in prehospital care of major trauma. J Trauma. 1997;42(4):723-9. 30. Baker SP, Whitfield RA, O'Neill B. Geographic variations in mortality from motor vehicle crashes. N Engl J Med. 1987;316(22):1384-7. 31. Times NY. WILLIAM HADDON JR., 58, DIES 1985 [cited 2017. Available from: http://www.nytimes.com/1985/03/05/us/william-haddon-jr-58-dies-authority-on-highway-safety.html. 32. Haddon W, Jr. The changing approach to the epidemiology, prevention, and amelioration of trauma: the transition to approaches etiologically rather than descriptively based. American journal of public health and the nation's health. 1968;58(8):1431-8. 33. Donabedian A. The quality of care. How can it be assessed? Jama. 1988;260(12):1743-8.
149
34. Moore L, Lavoie A, Bourgeois G, Lapointe J. Donabedian's structure-process-outcome quality of care model: Validation in an integrated trauma system. J Trauma Acute Care Surg. 2015;78(6):1168-75. 35. National Academy of S, National Research Council Committee on T, National Academy of S, National Research Council Committee on S. Accidental Death and Disability: The Neglected Disease of Modern Society. Washington (DC): National Academies Press (US) Copyright (c) National Academy of Sciences.; 1966. 36. Eastman AB, Mackenzie EJ, Nathens AB. Sustaining a coordinated, regional approach to trauma and emergency care is critical to patient health care needs. Health affairs (Project Hope). 2013;32(12):2091-8. 37. Trauma ACoSCo. Regional trauma systems: Optimal elements, integration, and assessment 2008 [cited 2017. Available from: https://www.facs.org/~/media/files/quality%20programs/trauma/tsepc/pdfs/regionaltraumasystems.ashx. 38. Trauma ACoSCo. Trauma Systems Components/Models 2017 [cited 2017. Available from: https://www.facs.org/quality-programs/trauma/tscp/components. 39. Driving MAD. MADD: No More Victims 2017 [cited 2017. Available from: https://www.madd.org/. 40. Violence ICAH. ICHV 2017 [cited 2017. Available from: http://www.ichv.org/. 41. Brown J, Sajankila N, Claridge JA. Prehospital Assessment of Trauma. The Surgical clinics of North America. 2017;97(5):961-83. 42. Mullner R, Goldberg J. An evaluation of the Illinois trauma system. Medical care. 1978;16(2):140-51. 43. Schechtman D, He JC, Zosa BM, Allen D, Claridge JA. Trauma system regionalization improves mortality in patients requiring trauma laparotomy. J Trauma Acute Care Surg. 2017;82(1):58-64. 44. Mann NC, Cahn RM, Mullins RJ, Brand DM, Jurkovich GJ. Survival among injured geriatric patients during construction of a statewide trauma system. The Journal of trauma. 2001;50(6):1111-6. 45. Nathens AB, Jurkovich GJ, Rivara FP, Maier RV. Effectiveness of state trauma systems in reducing injury-related mortality: a national evaluation. The Journal of trauma. 2000;48(1):25-30; discussion -1. 46. Nathens AB, Jurkovich GJ, Cummings P, Rivara FP, Maier RV. The effect of organized systems of trauma care on motor vehicle crash mortality. Jama. 2000;283(15):1990-4. 47. Mullins RJ, Mann NC, Hedges JR, Worrall W, Jurkovich GJ. Preferential benefit of implementation of a statewide trauma system in one of two adjacent states. The Journal of trauma. 1998;44(4):609-16; discussion 17. 48. Mullins RJ, Veum-Stone J, Hedges JR, Zimmer-Gembeck MJ, Mann NC, Southard PA, et al. Influence of a statewide trauma system on location of hospitalization and outcome of injured patients. The Journal of trauma. 1996;40(4):536-45; discussion 45-6. 49. Mullins RJ, Veum-Stone J, Helfand M, Zimmer-Gembeck M, Hedges JR, Southard PA, et al. Outcome of hospitalized injured patients after institution of a trauma system in an urban area. Jama. 1994;271(24):1919-24.
150
50. Guss DA, Meyer FT, Neuman TS, Baxt WG, Dunford JV, Jr., Griffith LD, et al. The impact of a regionalized trauma system on trauma care in San Diego County. Annals of emergency medicine. 1989;18(11):1141-5. 51. Shackford SR, Mackersie RC, Hoyt DB, Baxt WG, Eastman AB, Hammill FN, et al. Impact of a trauma system on outcome of severely injured patients. Archives of surgery (Chicago, Ill : 1960). 1987;122(5):523-7. 52. Shackford SR, Hollingworth-Fridlund P, Cooper GF, Eastman AB. The effect of regionalization upon the quality of trauma care as assessed by concurrent audit before and after institution of a trauma system: a preliminary report. The Journal of trauma. 1986;26(9):812-20. 53. West JG, Cales RH, Gazzaniga AB. Impact of regionalization. The Orange County experience. Archives of surgery (Chicago, Ill : 1960). 1983;118(6):740-4. 54. Tallon JM, Fell DB, Ackroyd-Stolarz S, Petrie D. Influence of a new province-wide trauma system on motor vehicle trauma care and mortality. The Journal of trauma. 2006;60(3):548-52. 55. Vali Y, Rashidian A, Jalili M, Omidvari AH, Jeddian A. Effectiveness of regionalization of trauma care services: a systematic review. Public health. 2017;146:92-107. 56. Moore L, Turgeon AF, Lauzier F, Emond M, Berthelot S, Clement J, et al. Evolution of patient outcomes over 14 years in a mature, inclusive Canadian trauma system. World J Surg. 2015;39(6):1397-405. 57. Gabbe BJ, Lyons RA, Fitzgerald MC, Judson R, Richardson J, Cameron PA. Reduced population burden of road transport-related major trauma after introduction of an inclusive trauma system. Ann Surg. 2015;261(3):565-72. 58. Utter GH, Maier RV, Rivara FP, Mock CN, Jurkovich GJ, Nathens AB. Inclusive trauma systems: do they improve triage or outcomes of the severely injured? The Journal of trauma. 2006;60(3):529-35; discussion 35-37. 59. Mann NC, Hedges JR, Mullins RJ, Helfand M, Worrall W, Zechnich AD, et al. Rural Hospital Transfer Patterns before and after Implementation of a Statewide Trauma System. Academic emergency medicine : official journal of the Society for Academic Emergency Medicine. 1997;4(8):764-71. 60. Olson CJ, Arthur M, Mullins RJ, Rowland D, Hedges JR, Mann NC. Influence of trauma system implementation on process of care delivered to seriously injured patients in rural trauma centers. Surgery. 2001;130(2):273-9. 61. Trauma ACoSCo. Resources for Optimal Care of the Injured Patient: American College of Surgeons Committee on Trauma; 2014 [Available from: https://www.facs.org/quality%20programs/trauma/vrc/resources. 62. Jackson TL, Balasubramaniam S. Trauma centers: an idea whose time has come. Journal of the National Medical Association. 1981;73(7):611-6. 63. National Academies of Sciences E, Medicine. A National Trauma Care System: Integrating Military and Civilian Trauma Systems to Achieve Zero Preventable Deaths After Injury. Berwick D, Downey A, Cornett E, editors. Washington, DC: The National Academies Press; 2016. 530 p.
151
64. MacKenzie EJ, Rivara FP, Jurkovich GJ, Nathens AB, Frey KP, Egleston BL, et al. A national evaluation of the effect of trauma-center care on mortality. N Engl J Med. 2006;354(4):366-78. 65. Mackenzie EJ, Rivara FP, Jurkovich GJ, Nathens AB, Egleston BL, Salkever DS, et al. The impact of trauma-center care on functional outcomes following major lower-limb trauma. The Journal of bone and joint surgery American volume. 2008;90(1):101-9. 66. MacKenzie EJ, Weir S, Rivara FP, Jurkovich GJ, Nathens AB, Wang W, et al. The value of trauma center care. The Journal of trauma. 2010;69(1):1-10. 67. Rutledge R, Fakhry SM, Meyer A, Sheldon GF, Baker CC. An analysis of the association of trauma centers with per capita hospitalizations and death rates from injury. Ann Surg. 1993;218(4):512-21; discussion 21-4. 68. Rutledge R, Messick J, Baker CC, Rhyne S, Butts J, Meyer A, et al. Multivariate population-based analysis of the association of county trauma centers with per capita county trauma death rates. The Journal of trauma. 1992;33(1):29-37; discussion -8. 69. Brown JB, Rosengart MR, Billiar TR, Peitzman AB, Sperry JL. Geographic distribution of trauma centers and injury-related mortality in the United States. J Trauma Acute Care Surg. 2016;80(1):42-9; discussion 9-50. 70. Administration NHTS. The Association Between Crash Proximity to Level 1 and 2 Trauma Centers and Crash Scene Mortality of Drivers Injured in Fatal Crashes: NHTSA; 2012 [Available from: https://crashstats.nhtsa.dot.gov/Api/Public/ViewPublication/811599. 71. Society AT. Trauma Center Levels Explained 2017 [Available from: http://www.amtrauma.org/?page=traumalevels. 72. Trauma ACoSCo. Verification, Review, and Consultation Program for Hospitals 2017 [Available from: https://www.facs.org/quality-programs/trauma/vrc. 73. Nathens AB, Jurkovich GJ, MacKenzie EJ, Rivara FP. A resource-based assessment of trauma care in the United States. The Journal of trauma. 2004;56(1):173-8; discussion 8. 74. Prevention CfDCa. Guidelines for Field Triage of Injured Patients: Recommendations of the National Expert Panel on Field Triage, 2011 2012 [Available from: https://www.cdc.gov/mmwr/preview/mmwrhtml/rr6101a1.htm. 75. van Rein EAJ, Houwert RM, Gunning AC, Lichtveld RA, Leenen LPH, van Heijl M. Accuracy of prehospital triage protocols in selecting severely injured patients: A systematic review. J Trauma Acute Care Surg. 2017;83(2):328-39. 76. Voskens FJ, van Rein EAJ, van der Sluijs R, Houwert RM, Lichtveld RA, Verleisdonk EJ, et al. Accuracy of Prehospital Triage in Selecting Severely Injured Trauma Patients. JAMA Surg. 2017. 77. Newgard CD, Fu R, Zive D, Rea T, Malveau S, Daya M, et al. Prospective Validation of the National Field Triage Guidelines for Identifying Seriously Injured Persons. J Am Coll Surg. 2016;222(2):146-58.e2. 78. Haas B, Gomez D, Zagorski B, Stukel TA, Rubenfeld GD, Nathens AB. Survival of the fittest: the hidden cost of undertriage of major trauma. J Am Coll Surg. 2010;211(6):804-11. 79. Kappel DA, Rossi DC, Polack EP, Avtgis TA, Martin MM. Does the rural trauma team development course shorten the interval from trauma patient arrival to decision to transfer? The Journal of trauma. 2011;70(2):315-9.
152
80. Brown JB, Gestring ML, Forsythe RM, Stassen NA, Billiar TR, Peitzman AB, et al. Systolic blood pressure criteria in the National Trauma Triage Protocol for geriatric trauma: 110 is the new 90. J Trauma Acute Care Surg. 2015;78(2):352-9. 81. Newgard CD, Richardson D, Holmes JF, Rea TD, Hsia RY, Mann NC, et al. Physiologic field triage criteria for identifying seriously injured older adults. Prehospital emergency care : official journal of the National Association of EMS Physicians and the National Association of State EMS Directors. 2014;18(4):461-70. 82. Brown JB, Lerner EB, Sperry JL, Billiar TR, Peitzman AB, Guyette FX. Prehospital lactate improves accuracy of prehospital criteria for designating trauma activation level. J Trauma Acute Care Surg. 2016;81(3):445-52. 83. Guyette F, Suffoletto B, Castillo JL, Quintero J, Callaway C, Puyana JC. Prehospital serum lactate as a predictor of outcomes in trauma patients: a retrospective observational study. The Journal of trauma. 2011;70(4):782-6. 84. Guyette FX, Meier EN, Newgard C, McKnight B, Daya M, Bulger EM, et al. A comparison of prehospital lactate and systolic blood pressure for predicting the need for resuscitative care in trauma transported by ground. J Trauma Acute Care Surg. 2015;78(3):600-6. 85. Davidson GH, Rivara FP, Mack CD, Kaufman R, Jurkovich GJ, Bulger EM. Validation of prehospital trauma triage criteria for motor vehicle collisions. J Trauma Acute Care Surg. 2014;76(3):755-61. 86. Services FICoEM. FICEMS Annual Report to Congress 2013-2015 2013 [Available from: https://www.ems.gov/pdf/2011-2012-FICEMS-RTC-Mikulski.pdf. 87. Administration NHTS. National EMS Scope of Practice Model 2007 [Available from: https://www.ems.gov/education/EMSScope.pdf. 88. Seamon MJ, Doane SM, Gaughan JP, Kulp H, D'Andrea AP, Pathak AS, et al. Prehospital interventions for penetrating trauma victims: a prospective comparison between Advanced Life Support and Basic Life Support. Injury. 2013;44(5):634-8. 89. Rappold JF, Hollenbach KA, Santora TA, Beadle D, Dauer ED, Sjoholm LO, et al. The evil of good is better: Making the case for basic life support transport for penetrating trauma victims in an urban environment. J Trauma Acute Care Surg. 2015;79(3):343-8. 90. Wandling MW, Nathens AB, Shapiro MB, Haut ER. Association of Prehospital Mode of Transport With Mortality in Penetrating Trauma: A Trauma System-Level Assessment of Private Vehicle Transportation vs Ground Emergency Medical Services. JAMA Surg. 2017. 91. Ivatury RR, Nallathambi MN, Roberge RJ, Rohman M, Stahl W. Penetrating thoracic injuries: in-field stabilization vs. prompt transport. J Trauma. 1987;27(9):1066-73. 92. Kondo Y, Fukuda T, Uchimido R, Hifumi T, Hayashida K. Effects of advanced life support versus basic life support on the mortality rates of patients with trauma in prehospital settings: a study protocol for a systematic review and meta-analysis. BMJ open. 2017;7(10):e016912. 93. Bukur M, Kurtovic S, Berry C, Tanios M, Margulies DR, Ley EJ, et al. Pre-hospital intubation is associated with increased mortality after traumatic brain injury. The Journal of surgical research. 2011;170(1):e117-21.
153
94. Davis DP, Peay J, Sise MJ, Vilke GM, Kennedy F, Eastman AB, et al. The impact of prehospital endotracheal intubation on outcome in moderate to severe traumatic brain injury. The Journal of trauma. 2005;58(5):933-9. 95. Irvin CB, Szpunar S, Cindrich LA, Walters J, Sills R. Should trauma patients with a Glasgow Coma Scale score of 3 be intubated prior to hospital arrival? Prehospital and disaster medicine. 2010;25(6):541-6. 96. Wang HE, Peitzman AB, Cassidy LD, Adelson PD, Yealy DM. Out-of-hospital endotracheal intubation and outcome after traumatic brain injury. Annals of emergency medicine. 2004;44(5):439-50. 97. Davis DP, Hoyt DB, Ochs M, Fortlage D, Holbrook T, Marshall LK, et al. The effect of paramedic rapid sequence intubation on outcome in patients with severe traumatic brain injury. The Journal of trauma. 2003;54(3):444-53. 98. Chou D, Harada MY, Barmparas G, Ko A, Ley EJ, Margulies DR, et al. Field intubation in civilian patients with hemorrhagic shock is associated with higher mortality. J Trauma Acute Care Surg. 2016;80(2):278-82. 99. Bernard SA, Nguyen V, Cameron P, Masci K, Fitzgerald M, Cooper DJ, et al. Prehospital rapid sequence intubation improves functional outcome for patients with severe traumatic brain injury: a randomized controlled trial. Ann Surg. 2010;252(6):959-65. 100. Davis DP, Peay J, Sise MJ, Kennedy F, Simon F, Tominaga G, et al. Prehospital airway and ventilation management: a trauma score and injury severity score-based analysis. The Journal of trauma. 2010;69(2):294-301. 101. Inaba K, Karamanos E, Skiada D, Grabo D, Hammer P, Martin M, et al. Cadaveric comparison of the optimal site for needle decompression of tension pneumothorax by prehospital care providers. J Trauma Acute Care Surg. 2015;79(6):1044-8. 102. Chang SJ, Ross SW, Kiefer DJ, Anderson WE, Rogers AT, Sing RF, et al. Evaluation of 8.0-cm needle at the fourth anterior axillary line for needle chest decompression of tension pneumothorax. J Trauma Acute Care Surg. 2014;76(4):1029-34. 103. Ball CG, Wyrzykowski AD, Kirkpatrick AW, Dente CJ, Nicholas JM, Salomone JP, et al. Thoracic needle decompression for tension pneumothorax: clinical correlation with catheter length. Canadian journal of surgery Journal canadien de chirurgie. 2010;53(3):184-8. 104. Schroll R, Smith A, McSwain NE, Jr., Myers J, Rocchi K, Inaba K, et al. A multi-institutional analysis of prehospital tourniquet use. J Trauma Acute Care Surg. 2015;79(1):10-4; discussion 4. 105. Inaba K, Siboni S, Resnick S, Zhu J, Wong MD, Haltmeier T, et al. Tourniquet use for civilian extremity trauma. J Trauma Acute Care Surg. 2015;79(2):232-7;quiz 332-3. 106. Bulger EM, Snyder D, Schoelles K, Gotschall C, Dawson D, Lang E, et al. An evidence-based prehospital guideline for external hemorrhage control: American College of Surgeons Committee on Trauma. Prehospital emergency care : official journal of the National Association of EMS Physicians and the National Association of State EMS Directors. 2014;18(2):163-73. 107. Callaway DW, Smith ER, Cain J, Shapiro G, Burnett WT, McKay SD, et al. Tactical emergency casualty care (TECC): guidelines for the provision of prehospital trauma care in high threat environments. Journal of special operations medicine : a peer reviewed journal for SOF medical professionals. 2011;11(3):104-22.
154
108. Kragh JF, Jr., Walters TJ, Baer DG, Fox CJ, Wade CE, Salinas J, et al. Practical use of emergency tourniquets to stop bleeding in major limb trauma. The Journal of trauma. 2008;64(2 Suppl):S38-49; discussion S-50. 109. Frascone RJ, Jensen JP, Kaye K, Salzman JG. Consecutive field trials using two different intraosseous devices. Prehospital emergency care : official journal of the National Association of EMS Physicians and the National Association of State EMS Directors. 2007;11(2):164-71. 110. Sampalis JS, Tamim H, Denis R, Boukas S, Ruest SA, Nikolis A, et al. Ineffectiveness of on-site intravenous lines: is prehospital time the culprit? The Journal of trauma. 1997;43(4):608-15; discussion 15-7. 111. Bickell WH, Wall MJ, Jr., Pepe PE, Martin RR, Ginger VF, Allen MK, et al. Immediate versus delayed fluid resuscitation for hypotensive patients with penetrating torso injuries. The New England journal of medicine. 1994;331(17):1105-9. 112. Brown JB, Cohen MJ, Minei JP, Maier RV, West MA, Billiar TR, et al. Goal-directed resuscitation in the prehospital setting: a propensity-adjusted analysis. J Trauma Acute Care Surg. 2013;74(5):1207-12; discussion 12-4. 113. Shackford SR. Prehospital fluid resuscitation of known or suspected traumatic brain injury. The Journal of trauma. 2011;70(5 Suppl):S32-3. 114. Cotton BA, Jerome R, Collier BR, Khetarpal S, Holevar M, Tucker B, et al. Guidelines for prehospital fluid resuscitation in the injured patient. The Journal of trauma. 2009;67(2):389-402. 115. Dutton RP, Mackenzie CF, Scalea TM. Hypotensive resuscitation during active hemorrhage: impact on in-hospital mortality. The Journal of trauma. 2002;52(6):1141-6. 116. Dula DJ, Wood GC, Rejmer AR, Starr M, Leicht M. Use of prehospital fluids in hypotensive blunt trauma patients. Prehospital emergency care : official journal of the National Association of EMS Physicians and the National Association of State EMS Directors. 2002;6(4):417-20. 117. Association AaSTN. ASTNA Patient Transport Principles & Practice. 4th ed: Mosby Elsevier; 2010. 118. Weidenburner CW. FIRST RECORDED MILITARY RESCUE BY HELICOPTER: The Hoverfly in CBI 2015 [Available from: http://www.cbi-theater.com/hoverfly/hoverfly.html. 119. Baxt WG, Moody P. The impact of a rotorcraft aeromedical emergency care service on trauma mortality. Jama. 1983;249(22):3047-51. 120. Kotwal RS, Howard JT, Orman JA, Tarpey BW, Bailey JA, Champion HR, et al. The Effect of a Golden Hour Policy on the Morbidity and Mortality of Combat Casualties. JAMA Surg. 2015;30:1-10. 121. Holcomb JB, Stansbury LG, Champion HR, Wade C, Bellamy RF. Understanding combat casualty care statistics. J Trauma. 2006;60(2):397-401. 122. World E. Helicopter EMS: Part 1: A Brief History 2010 [Available from: https://www.emsworld.com/article/10319182/helicopter-ems-part-1-brief-history. 123. Cleveland HC, Bigelow DB, Dracon D, Dusty F. A civilian air emergency service: a report of its development, technical aspects, and experience. The Journal of trauma. 1976;16(6):452-63.
155
124. Moylan JA, Fitzpatrick KT, Beyer AJ, 3rd, Georgiade GS. Factors improving survival in multisystem trauma patients. Ann Surg. 1988;207(6):679-85. 125. Galvagno SM, Jr., Haut ER, Zafar SN, Millin MG, Efron DT, Koenig GJ, Jr., et al. Association between helicopter vs ground emergency medical services and survival for adults with major trauma. Jama. 2012;307(15):1602-10. 126. Brown JB, Gestring ML, Guyette FX, Rosengart MR, Stassen NA, Forsythe RM, et al. Helicopter transport improves survival following injury in the absence of a time-saving advantage. Surgery. 2016;159(3):947-59. 127. Bekelis K, Missios S, Mackenzie TA. Prehospital helicopter transport and survival of patients with traumatic brain injury. Ann Surg. 2015;261(3):579-85. 128. Hannay RS, Wyrzykowski AD, Ball CG, Laupland K, Feliciano DV. Retrospective review of injury severity, interventions and outcomes among helicopter and nonhelicopter transport patients at a Level 1 urban trauma centre. Canadian journal of surgery Journal canadien de chirurgie. 2014;57(1):49-54. 129. Andruszkow H, Lefering R, Frink M, Mommsen P, Zeckey C, Rahe K, et al. Survival benefit of helicopter emergency medical services compared to ground emergency medical services in traumatized patients. Crit Care. 2013;17(3):R124. 130. Stewart KE, Cowan LD, Thompson DM, Sacra JC, Albrecht R. Association of direct helicopter versus ground transport and in-hospital mortality in trauma patients: a propensity score analysis. Academic emergency medicine : official journal of the Society for Academic Emergency Medicine. 2011;18(11):1208-16. 131. Brathwaite CE, Rosko M, McDowell R, Gallagher J, Proenca J, Spott MA. A critical analysis of on-scene helicopter transport on survival in a statewide trauma system. The Journal of trauma. 1998;45(1):140-4; discussion 4-6. 132. Bulger EM, Guffey D, Guyette FX, MacDonald RD, Brasel K, Kerby JD, et al. Impact of prehospital mode of transport after severe injury: a multicenter evaluation from the Resuscitation Outcomes Consortium. J Trauma Acute Care Surg. 2012;72(3):567-73; discussion 73-5; quiz 803. 133. Cunningham P, Rutledge R, Baker CC, Clancy TV. A comparison of the association of helicopter and ground ambulance transport with the outcome of injury in trauma patients transported from the scene. The Journal of trauma. 1997;43(6):940-6. 134. Schiller WR, Knox R, Zinnecker H, Jeevanandam M, Sayre M, Burke J, et al. Effect of helicopter transport of trauma victims on survival in an urban trauma center. The Journal of trauma. 1988;28(8):1127-34. 135. Brown JB, Gestring ML, Guyette FX, Rosengart MR, Stassen NA, Forsythe RM, et al. External validation of the Air Medical Prehospital Triage score for identifying trauma patients likely to benefit from scene helicopter transport. J Trauma Acute Care Surg. 2017;82(2):270-9. 136. Brown JB, Gestring ML, Guyette FX, Rosengart MR, Stassen NA, Forsythe RM, et al. Development and Validation of the Air Medical Prehospital Triage Score for Helicopter Transport of Trauma Patients. Ann Surg. 2016;264(2):378-85. 137. Services ADoAM. National & State Overview of Air Medical Coverage in 2016: ADAMS; 2016 [Available from: http://www.adamsairmed.org/pubs/ADAMS_Intro.pdf.
156
138. Rhinehart ZJ, Guyette FX, Sperry JL, Forsythe RM, Murdock A, Alarcon LH, et al. The association between air ambulance distribution and trauma mortality. Ann Surg. 2013;257(6):1147-53. 139. Administration NHTS. The Association of the Proximity of Fatal Motor Vehicle Crash Locations to the Availability of Helicopter Emergency Medical Service Response: NHTSA; 2012 [Available from: https://crashstats.nhtsa.dot.gov/Api/Public/ViewPublication/811542. 140. Services ADoAM. Air Medical Civilian Fleet by Rotor Wing Make/Model: ADAMS; 2014 [Available from: http://www.adamsairmed.org/pubs/rw_make_model_in_ADAMS_multi_year.pdf. 141. Bureau UC. TIGER/Line® with Selected Demographic and Economic Data: US Census Bureau; 2017 [Available from: https://www.census.gov/geo/maps-data/data/tiger-data.html. 142. Administration FH. The National Highway Planning Network: Version 14.05: Federal Highway Administration; 2017 [Available from: https://www.fhwa.dot.gov/planning/processes/tools/nhpn/. 143. Swaroop M, Straus DC, Agubuzu O, Esposito TJ, Schermer CR, Crandall ML. Pre-hospital transport times and survival for Hypotensive patients with penetrating thoracic trauma. J Emerg Trauma Shock. 2013;6(1):16-20. doi: 10.4103/0974-2700.106320. 144. Crandall M, Sharp D, Unger E, Straus D, Brasel K, Hsia R, et al. Trauma deserts: distance from a trauma center, transport times, and mortality from gunshot wounds in Chicago. Am J Public Health. 2013;103(6):1103-9. doi: 10.2105/AJPH.013.301223. Epub 2013 Apr 18. 145. Tien HC, Jung V, Pinto R, Mainprize T, Scales DC, Rizoli SB. Reducing time-to-treatment decreases mortality of trauma patients with acute subdural hematoma. Ann Surg. 2011;253(6):1178-83. doi: 10.097/SLA.0b013e318217e339. 146. Dinh MM, Bein K, Roncal S, Byrne CM, Petchell J, Brennan J. Redefining the golden hour for severe head injury in an urban setting: the effect of prehospital arrival times on patient outcomes. Injury. 2013;44(5):606-10. doi: 10.1016/j.injury.2012.01.011. Epub Feb 14. 147. Feero S, Hedges JR, Simmons E, Irwin L. Does out-of-hospital EMS time affect trauma survival? Am J Emerg Med. 1995;13(2):133-5. 148. Baez AA, Lane PL, Sorondo B, Giraldez EM. Predictive effect of out-of-hospital time in outcomes of severely injured young adult and elderly patients. Prehospital and disaster medicine. 2006;21(6):427-30. 149. Newgard CD, Schmicker RH, Hedges JR, Trickett JP, Davis DP, Bulger EM, et al. Emergency medical services intervals and survival in trauma: assessment of the "golden hour" in a North American prospective cohort. Ann Emerg Med. 2010;55(3):235-46.e4. doi: 10.1016/j.annemergmed.2009.07.024. Epub Sep 23. 150. McCoy CE, Menchine M, Sampson S, Anderson C, Kahn C. Emergency medical services out-of-hospital scene and transport times and their association with mortality in trauma patients presenting to an urban Level I trauma center. Ann Emerg Med. 2013;61(2):167-74. doi: 10.1016/j.annemergmed.2012.08.026. Epub Nov 9. 151. Gonzalez RP, Cummings GR, Phelan HA, Mulekar MS, Rodning CB. Does increased emergency medical services prehospital time affect patient mortality in rural motor vehicle
157
crashes? A statewide analysis. Am J Surg. 2009;197(1):30-4. doi: 10.1016/j.amjsurg.2007.11.018. Epub 8 Jun 16. 152. Funder KS, Petersen JA, Steinmetz J. On-scene time and outcome after penetrating trauma: an observational study. Emerg Med J. 2011;28(9):797-801. doi: 10.1136/emj.2010.097535. Epub 2010 Oct 9. 153. Kidher E, Krasopoulos G, Coats T, Charitou A, Magee P, Uppal R, et al. The effect of prehospital time related variables on mortality following severe thoracic trauma. Injury. 2012;43(9):1386-92. doi: 10.016/j.injury.2011.04.014. Epub May 12. 154. Stein C, Wallis L, Adetunji O. Meeting national response time targets for priority 1 incidents in an urban emergency medical services system in South Africa: More ambulances won't help. S Afr Med J. 2015;105(10):840-4. doi: 10.7196/SAMJnew.8087. 155. Lam SS, Zhang J, Zhang ZC, Oh HC, Overton J, Ng YY, et al. Dynamic ambulance reallocation for the reduction of ambulance response times using system status management. Am J Emerg Med. 2015;33(2):159-66. doi: 10.1016/j.ajem.2014.10.044. Epub Nov 8. 156. McLay LA, Mayorga ME. Evaluating emergency medical service performance measures. Health Care Manag Sci. 2010;13(2):124-36. 157. Peyravi M, Khodakarim S, Ortenwall P, Khorram-Manesh A. Does temporary location of ambulances ("fluid deployment") affect response times and patient outcome? International journal of emergency medicine. 2015;8(1):37. 158. Santana MJ, Stelfox HT. Development and evaluation of evidence-informed quality indicators for adult injury care. Ann Surg. 2014;259(1):186-92. doi: 10.1097/SLA.0b013e31828df98e. 159. Myers JB, Slovis CM, Eckstein M, Goodloe JM, Isaacs SM, Loflin JR, et al. Evidence-based performance measures for emergency medical services systems: a model for expanded EMS benchmarking. Prehospital emergency care : official journal of the National Association of EMS Physicians and the National Association of State EMS Directors. 2008;12(2):141-51. 160. Honigman B, Rohweder K, Moore EE, Lowenstein SR, Pons PT. Prehospital advanced trauma life support for penetrating cardiac wounds. Annals of emergency medicine. 1990;19(2):145-50. 161. Eachempati SR, Robb T, Ivatury RR, Hydo LJ, Barie PS. Factors associated with mortality in patients with penetrating abdominal vascular trauma. The Journal of surgical research. 2002;108(2):222-6. 162. Brown JB, Rosengart MR, Forsythe RM, Reynolds BR, Gestring ML, Hallinan WM, et al. Not all prehospital time is equal: Influence of scene time on mortality. J Trauma Acute Care Surg. 2016;81(1):93-100. 163. Champion HR, Mabee MS, Meredith JW. The state of US trauma systems: public perceptions versus reality--implications for US response to terrorism and mass casualty events. J Am Coll Surg. 2006;203(6):951-61. 164. Minei JP, Schmicker RH, Kerby JD, Stiell IG, Schreiber MA, Bulger E, et al. Severe traumatic injury: regional variation in incidence and outcome. Ann Surg. 2010;252(1):149-57.
158
165. Branas CC, MacKenzie EJ, Williams JC, Schwab CW, Teter HM, Flanigan MC, et al. Access to trauma centers in the United States. Jama. 2005;293(21):2626-33. 166. Carr BG, Bowman AJ, Wolff CS, Mullen MT, Holena DN, Branas CC, et al. Disparities in access to trauma care in the United States: A population-based analysis. Injury. 2017;48(2):332-8. doi: 10.1016/j.injury.2017.01.008. Epub Jan 3. 167. Gomez D, Berube M, Xiong W, Ahmed N, Haas B, Schuurman N, et al. Identifying targets for potential interventions to reduce rural trauma deaths: a population-based analysis. The Journal of trauma. 2010;69(3):633-9. 168. Ostroff C. BdFC. Millions of Americans live nowhere near a hospital, jeopardize their lives: CNN; 2017 [updated August 3, 2017. Available from: http://www.cnn.com/2017/08/03/health/hospital-deserts/index.html. 169. Promotion OoDPaH. Injury and Violence Prevention: HealthyPeople.gov; 2017 [updated Februrary 1, 2018. Available from: https://www.healthypeople.gov/2020/topics-objectives/topic/injury-and-violence-prevention/objectives. 170. Witiw CD, Byrne JP, Nassiri F, Badhiwala JH, Nathens AB, da Costa LB. Isolated Traumatic Subarachnoid Hemorrhage: An Evaluation of Critical Care Unit Admission Practices and Outcomes From a North American Perspective. Crit Care Med. 2017. 171. Tignanelli CJ, Joseph B, Jakubus JL, Iskander GA, Napolitano LM, Hemmila MR. Variability in Management of Blunt Liver Trauma and Contribution of Level of ACS-COT Verification Status on Mortality. J Trauma Acute Care Surg. 2017. 172. Byrne JP, Nathens AB, Gomez D, Pincus D, Jenkinson RJ. Timing of femoral shaft fracture fixation following major trauma: A retrospective cohort study of United States trauma centers. PLoS medicine. 2017;14(7):e1002336. 173. Nathens AB, McMurray MK, Cuschieri J, Durr EA, Moore EE, Bankey PE, et al. The practice of venous thromboembolism prophylaxis in the major trauma patient. J Trauma. 2007;62(3):557-62; discussion 62-3. 174. Shafi S, Barnes SA, Rayan N, Kudyakov R, Foreman M, Cryer HG, et al. Compliance with recommended care at trauma centers: association with patient outcomes. J Am Coll Surg. 2014;219(2):189-98. 175. Shafi S, Nathens AB, Cryer HG, Hemmila MR, Pasquale MD, Clark DE, et al. The Trauma Quality Improvement Program of the American College of Surgeons Committee on Trauma. J Am Coll Surg. 2009;209(4):521-30.e1. 176. Nathens AB, Cryer HG, Fildes J. The American College of Surgeons Trauma Quality Improvement Program. The Surgical clinics of North America. 2012;92(2):441-54, x-xi. 177. Hemmila MR, Nathens AB, Shafi S, Calland JF, Clark DE, Cryer HG, et al. The Trauma Quality Improvement Program: pilot study and initial demonstration of feasibility. The Journal of trauma. 2010;68(2):253-62. 178. Shafi S, Nathens AB, Parks J, Cryer HM, Fildes JJ, Gentilello LM. Trauma quality improvement using risk-adjusted outcomes. J Trauma. 2008;64(3):599-604; discussion -6. doi: 10.1097/TA.0b013e31816533f9. 179. Surgeons ACo. Trauma Quality Improvement Program 2017 [Available from: https://www.facs.org/quality-programs/trauma/tqip. 180. Khuri SF, Daley J, Henderson W, Hur K, Gibbs JO, Barbour G, et al. Risk adjustment of the postoperative mortality rate for the comparative assessment of the quality of surgical
159
care: results of the National Veterans Affairs Surgical Risk Study. J Am Coll Surg. 1997;185(4):315-27. 181. Khuri SF. The NSQIP: a new frontier in surgery. Surgery. 2005;138(5):837-43. 182. Gomez D, Haas B, Hemmila M, Pasquale M, Goble S, Neal M, et al. Hips can lie: impact of excluding isolated hip fractures on external benchmarking of trauma center performance. The Journal of trauma. 2010;69(5):1037-41. 183. Haukoos JS, Newgard CD. Advanced statistics: missing data in clinical research--part 1: an introduction and conceptual framework. Academic emergency medicine : official journal of the Society for Academic Emergency Medicine. 2007;14(7):662-8. 184. Newgard CD, Haukoos JS. Advanced statistics: missing data in clinical research--part 2: multiple imputation. Academic emergency medicine : official journal of the Society for Academic Emergency Medicine. 2007;14(7):669-78. 185. Sterne JA, White IR, Carlin JB, Spratt M, Royston P, Kenward MG, et al. Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls. BMJ (Clinical research ed). 2009;338:b2393. 186. Yuan Y. Multiple Imputation Using SAS Software. Journal of Statistical Software. 2011;45(6). 187. Merlo J, Chaix B, Ohlsson H, Beckman A, Johnell K, Hjerpe P, et al. A brief conceptual tutorial of multilevel analysis in social epidemiology: using measures of clustering in multilevel logistic regression to investigate contextual phenomena. J Epidemiol Community Health. 2006;60(4):290-7. 188. Shafi S, Rayan N, Barnes S, Fleming N, Gentilello LM, Ballard D. Moving from "optimal resources" to "optimal care" at trauma centers. J Trauma Acute Care Surg. 2012;72(4):870-7. 189. Mann NC, Guice K, Cassidy L, Wright D, Koury J. Are statewide trauma registries comparable? Reaching for a national trauma dataset. Acad Emerg Med. 2006;13(9):946-53. Epub 2006 Aug 10. 190. Bergeron E, Lavoie A, Moore L, Bamvita JM, Ratte S, Clas D. Paying the price of excluding patients from a trauma registry. J Trauma. 2006;60(2):300-4. 191. Van Haren RM, Thorson CM, Curia E, Schulman CI, Namias N, Livingstone AS, et al. Impact of definitions on trauma center mortality rates and performance. J Trauma Acute Care Surg. 2012;73(6):1512-6. doi: 10.097/TA.0b013e318270d40f. 192. Calland JF, Nathens AB, Young JS, Neal ML, Goble S, Abelson J, et al. The effect of dead-on-arrival and emergency department death classification on risk-adjusted performance in the American College of Surgeons Trauma Quality Improvement Program. J Trauma Acute Care Surg. 2012;73(5):1086-91; discussion 91-2. doi: 10.97/TA.0b013e31826fc7a0. 193. Pasquale MD, Rhodes M, Cipolle MD, Hanley T, Wasser T. Defining "dead on arrival": impact on a level I trauma center. J Trauma. 1996;41(4):726-30. 194. Sarkar B, Brunsvold ME, Cherry-Bukoweic JR, Hemmila MR, Park PK, Raghavendran K, et al. American College of Surgeons' Committee on Trauma Performance Improvement and Patient Safety program: maximal impact in a mature trauma center. The Journal of trauma. 2011;71(5):1447-53; discussion 53-4.
160
195. Heaney JB, Guidry C, Simms E, Turney J, Meade P, Hunt JP, et al. To TQIP or not to TQIP? That is the question. The American surgeon. 2014;80(4):386-90. 196. Nathens AB, Jurkovich GJ, Cummings P, Rivara FP, Maier RV. The effect of organized systems of trauma care on motor vehicle crash mortality. JAMA. 2000;283(15):1990-4. 197. Brown JB, Rosengart MR, Billiar TR, Peitzman AB, Sperry JL. Distance matters: Effect of geographic trauma system resource organization on fatal motor vehicle collisions. J Trauma Acute Care Surg. 2017;83(1):111-8. 198. Vernon DD, Cook LJ, Peterson KJ, Michael Dean J. Effect of repeal of the national maximum speed limit law on occurrence of crashes, injury crashes, and fatal crashes on Utah highways. Accident; analysis and prevention. 2004;36(2):223-9. 199. Gallaher MM, Sewell CM, Flint S, Herndon JL, Graff H, Fenner J, et al. Effects of the 65-mph speed limit on rural interstate fatalities in New Mexico. Jama. 1989;262(16):2243-5. 200. Wagenaar AC, Streff FM, Schultz RH. Effects of the 65 mph speed limit on injury morbidity and mortality. Accident; analysis and prevention. 1990;22(6):571-85. 201. Baum HM, Lund AK, Wells JK. The mortality consequences of raising the speed limit to 65 mph on rural interstates. American journal of public health. 1989;79(10):1392-5. 202. Gonzalez RP, Cummings GR, Phelan HA, Harlin S, Mulekar M, Rodning CB. Increased rural vehicular mortality rates: roadways with higher speed limits or excessive vehicular speed? J Trauma. 2007;63(6):1360-3. doi: 10.097/TA.0b013e31815b83b3. 203. Zador PL, Krawchuk SA, Voas RB. Alcohol-related relative risk of driver fatalities and driver involvement in fatal crashes in relation to driver age and gender: an update using 1996 data. Journal of studies on alcohol. 2000;61(3):387-95. 204. Tippetts AS, Voas RB, Fell JC, Nichols JL. A meta-analysis of .08 BAC laws in 19 jurisdictions in the United States. Accident; analysis and prevention. 2005;37(1):149-61. 205. Zador PL, Lund AK, Fields M, Weinberg K. Fatal crash involvement and laws against alcohol-impaired driving. Journal of public health policy. 1989;10(4):467-85. 206. Zwerling C, Jones MP. Evaluation of the effectiveness of low blood alcohol concentration laws for younger drivers. American journal of preventive medicine. 1999;16(1 Suppl):76-80. 207. Villaveces A, Cummings P, Koepsell TD, Rivara FP, Lumley T, Moffat J. Association of alcohol-related laws with deaths due to motor vehicle and motorcycle crashes in the United States, 1980-1997. American journal of epidemiology. 2003;157(2):131-40. 208. Safety IIfH. Highway Safety Topics: IIHS; 2017 [Available from: http://www.iihs.org/iihs/topics#statelaws. 209. Yao J, Johnson MB, Tippetts S. Enforcement uniquely predicts reductions in alcohol-impaired crash fatalities. Addiction (Abingdon, England). 2016;111(3):448-53. 210. Borgialli DA, Hill EM, Maio RF, Compton CP, Gregor MA. Effects of alcohol on the geographic variation of driver fatalities in motor vehicle crashes. Academic emergency medicine : official journal of the Society for Academic Emergency Medicine. 2000;7(1):7-13. 211. Evans L. Safety-belt effectiveness: the influence of crash severity and selective recruitment. Accident; analysis and prevention. 1996;28(4):423-33. 212. Lestina DC, Williams AF, Lund AK, Zador P, Kuhlmann TP. Motor vehicle crash injury patterns and the Virginia seat belt law. Jama. 1991;265(11):1409-13.
161
213. Rivara FP, Thompson DC, Cummings P. Effectiveness of primary and secondary enforced seat belt laws. American journal of preventive medicine. 1999;16(1 Suppl):30-9. 214. Peura C, Kilch JA, Clark DE. Evaluating adverse rural crash outcomes using the NHTSA State Data System. Accid Anal Prev. 2015;82:257-62.(doi):10.1016/j.aap.2015.06.005. Epub Jun 24. 215. Administration NHTS. FARS Annual Crash Statistics 2017 [Available from: https://www.nhtsa.gov/research-data/fatality-analysis-reporting-system-fars. 216. Goldstein GP, Clark DE, Travis LL, Haskins AE. Explaining regional disparities in traffic mortality by decomposing conditional probabilities. Inj Prev. 2011;17(2):84-90. doi: 10.1136/ip.2010.029249. Epub 2011 Jan 6. 217. Zwerling C, Peek-Asa C, Whitten PS, Choi SW, Sprince NL, Jones MP. Fatal motor vehicle crashes in rural and urban areas: decomposing rates into contributing factors. Inj Prev. 2005;11(1):24-8. 218. Clark DE. Effect of population density on mortality after motor vehicle collisions. Accid Anal Prev. 2003;35(6):965-71. 219. Brown LH, Khanna A, Hunt RC. Rural vs urban motor vehicle crash death rates: 20 years of FARS data. Prehosp Emerg Care. 2000;4(1):7-13. 220. Clark DE, Cushing BM. Predicting regional variations in mortality from motor vehicle crashes. Academic emergency medicine : official journal of the Society for Academic Emergency Medicine. 1999;6(2):125-30. 221. Muelleman RL, Mueller K. Fatal motor vehicle crashes: variations of crash characteristics within rural regions of different population densities. J Trauma. 1996;41(2):315-20. 222. Maio RF, Green PE, Becker MP, Burney RE, Compton C. Rural motor vehicle crash mortality: the role of crash severity and medical resources. Accid Anal Prev. 1992;24(6):631-42. 223. Travis LL, Clark DE, Haskins AE, Kilch JA. Mortality in rural locations after severe injuries from motor vehicle crashes. J Safety Res. 2012;43(5-6):375-80. doi: 10.1016/j.jsr.2012.10.004. Epub Oct 24. 224. Muelleman RL, Wadman MC, Tran TP, Ullrich F, Anderson JR. Rural motor vehicle crash risk of death is higher after controlling for injury severity. J Trauma. 2007;62(1):221-5; discussion 5-6. 225. Aftyka A, Rybojad B, Rudnicka-Drozak E. Are there any differences in medical emergency team interventions between rural and urban areas? A single-centre cohort study. The Australian journal of rural health. 2014;22(5):223-8. 226. Newgard CD, Fu R, Bulger E, Hedges JR, Mann NC, Wright DA, et al. Evaluation of Rural vs Urban Trauma Patients Served by 9-1-1 Emergency Medical Services. JAMA Surg. 2017;152(1):11-8. 227. Mell HK, Mumma SN, Hiestand B, Carr BG, Holland T, Stopyra J. Emergency Medical Services Response Times in Rural, Suburban, and Urban Areas. JAMA Surg. 2017. 228. Ray JJ, Meizoso JP, Satahoo SS, Davis JS, Van Haren RM, Dermer H, et al. Potentially preventable prehospital deaths from motor vehicle collisions. Traffic Inj Prev. 2016;17(7):676-80. doi: 10.1080/15389588.2016.1149580. Epub 2016 Feb 18.
162
229. Champion HR, Lombardo LV, Wade CE, Kalin EJ, Lawnick MM, Holcomb JB. Time and place of death from automobile crashes: Research endpoint implications. J Trauma Acute Care Surg. 2016;81(3):420-6. 230. Gomez D, Xiong W, Haas B, Goble S, Ahmed N, Nathens AB. The missing dead: the problem of case ascertainment in the assessment of trauma center performance. J Trauma. 2009;66(4):1218-24; discussion 24-5. doi: 10.097/TA.0b013e31819a04d2. 231. Trauma ACoSCo. National Trauma Data Standard. 2015. 232. Clarke JR, Ragone AV, Greenwald L. Comparisons of survival predictions using survival risk ratios based on International Classification of Diseases, Ninth Revision and Abbreviated Injury Scale trauma diagnosis codes. J Trauma. 2005;59(3):563-7; discussion 7-9. 233. Shimazu S, Shatney CH. Outcomes of trauma patients with no vital signs on hospital admission. J Trauma. 1983;23(3):213-6. 234. Kleber C, Giesecke MT, Lindner T, Haas NP, Buschmann CT. Requirement for a structured algorithm in cardiac arrest following major trauma: epidemiology, management errors, and preventability of traumatic deaths in Berlin. Resuscitation. 2014;85(3):405-10. doi: 10.1016/j.resuscitation.2013.11.009. Epub Nov 25. 235. Efron DT, Haider A, Chang D, Haut ER, Brooke B, Cornwell EE, 3rd. Alarming surge in nonsurvivable urban trauma and the case for violence prevention. Arch Surg. 2006;141(8):800-3; discussion 3-5. 236. Seamon MJ, Shiroff AM, Franco M, Stawicki SP, Molina EJ, Gaughan JP, et al. Emergency department thoracotomy for penetrating injuries of the heart and great vessels: an appraisal of 283 consecutive cases from two urban trauma centers. J Trauma. 2009;67(6):1250-7; discussion 7-8. doi: 10.097/TA.0b013e3181c3fef9. 237. Rhee PM, Acosta J, Bridgeman A, Wang D, Jordan M, Rich N. Survival after emergency department thoracotomy: review of published data from the past 25 years. J Am Coll Surg. 2000;190(3):288-98. 238. Program ACoSTQI. ACS TQIP Aggregate Report. 2014 March 2014. 239. Shafi S, Nathens AB, Cryer HG, Hemmila MR, Pasquale MD, Clark DE, et al. The Trauma Quality Improvement Program of the American College of Surgeons Committee on Trauma. J Am Coll Surg. 2009;209(4):521-30.e1. doi: 10.1016/j.jamcollsurg.2009.07.001. Epub Aug 13. 240. Mann NC, Kane L, Dai M, Jacobson K. Description of the 2012 NEMSIS Public-Release Research Dataset. Prehosp Emerg Care. 2015;19(2):232-40. doi: 10.3109/10903127.2014.959219. Epub 2014 Oct 7. 241. Byrne JP, Xiong W, Gomez D, Mason S, Karanicolas P, Rizoli S, et al. Redefining "dead on arrival": Identifying the unsalvageable patient for the purpose of performance improvement. J Trauma Acute Care Surg. 2015;79(5):850-7. doi: 10.1097/TA.0000000000000843. 242. NEMSIS: Measuring Urbanicity: The National Emergency Medical Service Information System; 2013 [Available from: https://nemsis.org/reportingTools/documents/MeasuringUrbanicity.ppt.
163
243. The United States Department of Agriculture: Urban Influence Codes: The United States Department of Agriculture; 2013 [Available from: http://www.ers.usda.gov/data-products/urban-influence-codes.aspx. 244. Rutledge R, Osler T, Emery S, Kromhout-Schiro S. The end of the Injury Severity Score (ISS) and the Trauma and Injury Severity Score (TRISS): ICISS, an International Classification of Diseases, ninth revision-based prediction tool, outperforms both ISS and TRISS as predictors of trauma patient survival, hospital charges, and hospital length of stay. J Trauma. 1998;44(1):41-9. 245. Hubbard AE, Ahern J, Fleischer NL, Van der Laan M, Lippman SA, Jewell N, et al. To GEE or not to GEE: comparing population average and mixed models for estimating the associations between neighborhood risk factors and health. Epidemiology. 2010;21(4):467-74. doi: 10.1097/EDE.0b013e3181caeb90. 246. Mickey RM, Greenland S. The impact of confounder selection criteria on effect estimation. Am J Epidemiol. 1989;129(1):125-37. 247. Kramer AA, Zimmerman JE. Assessing the calibration of mortality benchmarks in critical care: The Hosmer-Lemeshow test revisited. Crit Care Med. 2007;35(9):2052-6. 248. Martin M, Oh J, Currier H, Tai N, Beekley A, Eckert M, et al. An analysis of in-hospital deaths at a modern combat support hospital. J Trauma. 2009;66(4 Suppl):S51-60; discussion S-1. doi: 10.1097/TA.0b013e31819d86ad. 249. Sampalis JS, Lavoie A, Williams JI, Mulder DS, Kalina M. Impact of on-site care, prehospital time, and level of in-hospital care on survival in severely injured patients. J Trauma. 1993;34(2):252-61. 250. Morgenstern H. Ecologic studies in epidemiology: concepts, principles, and methods. Annu Rev Public Health. 1995;16:61-81. 251. Committee on Military Trauma Care's Learning Health S, Its Translation to the Civilian S, Board on Health Sciences P, Board on the Health of Select P, Health, Medicine D, et al. In: Berwick D, Downey A, Cornett E, editors. A National Trauma Care System: Integrating Military and Civilian Trauma Systems to Achieve Zero Preventable Deaths After Injury. Washington (DC): National Academies Press (US) Copyright 2016 by the National Academy of Sciences. All rights reserved.; 2016. 252. Champion HR, Lombardo LV, Wade CE, Kalin EJ, Lawnick MM, Holcomb JB. Time and place of death from automobile crashes: Research endpoint implications. J Trauma Acute Care Surg. 2016;81(3):420-6. doi: 10.1097/TA.0000000000001124. 253. Organization WH. Global Status Report on Road Safety: Country Profiles 2015 [Available from: http://www.who.int/violence_injury_prevention/road_safety_status/2015/en/. 254. Administration NHTS. Fatality Analysis Reporting System 2017 [Available from: https://www.nhtsa.gov/research-data/fatality-analysis-reporting-system-fars. 255. Bureau USC. US Population Estimates: United States Census Bureau; 2016 [Available from: https://www.census.gov/programs-surveys/popest/data/data-sets.html. 256. Austin PC. Using the Standardized Difference to Compare the Prevalence of a Binary Variable Between Two Groups in Observational Research. Communications in Statistics - Simulation and Computation. 2009;38(6):1228-34.
164
257. Mamdani M, Sykora K, Li P, Normand SL, Streiner DL, Austin PC, et al. Reader's guide to critical appraisal of cohort studies: 2. Assessing potential for confounding. BMJ. 2005;330(7497):960-2. 258. Hart PD. Receiver Operating Characteristic (ROC) Curve Analysis: A Tutorial Using Body Mass Index (BMI) as a Measure of Obesity. Journal of Physical Activity Research. 2016;1(1):5-8. 259. Youden WJ. Index for rating diagnostic tests. Cancer. 1950;3(1):32-5. 260. Merlo J, Yang M, Chaix B, Lynch J, Rastam L. A brief conceptual tutorial on multilevel analysis in social epidemiology: investigating contextual phenomena in different groups of people. J Epidemiol Community Health. 2005;59(9):729-36. 261. Davis JS, Satahoo SS, Butler FK, Dermer H, Naranjo D, Julien K, et al. An analysis of prehospital deaths: Who can we save? J Trauma Acute Care Surg. 2014;77(2):213-8. doi: 10.1097/TA.0000000000000292. 262. Harmsen AM, Giannakopoulos GF, Moerbeek PR, Jansma EP, Bonjer HJ, Bloemers FW. The influence of prehospital time on trauma patients outcome: a systematic review. Injury. 2015;46(4):602-9. doi: 10.1016/j.injury.2015.01.008. Epub Jan 16. 263. Clark DE, Cushing BM. Predicted effect of automatic crash notification on traffic mortality. Accid Anal Prev. 2002;34(4):507-13. 264. Gonzalez RP, Cummings G, Mulekar M, Rodning CB. Increased mortality in rural vehicular trauma: identifying contributing factors through data linkage. J Trauma. 2006;61(2):404-9. 265. Rutledge R, Smith CY, Azizkhank RG. A population-based multivariate analysis of the association of county demographic and medical system factors with per capita pediatric trauma death rates in North Carolina. Ann Surg. 1994;219(2):205-10. 266. Safety IIfH. General Statisics: Fatal Crash Totals: IIHS; 2015 [Available from: http://www.iihs.org/iihs/topics/t/general-statistics/fatalityfacts/state-by-state-overview. 267. NEMSIS. NEMSIS - National EMS Information System: NEMSIS; 2017 [Available from: https://nemsis.org. 268. Administration NHTS. Fatality Analysis Reporting System (FARS) Encyclopedia: NHTSA; 2017 [Available from: https://www-fars.nhtsa.dot.gov/Main/index.aspx. 269. Administration NHTS. FARS Analytical User’s Manual: NHTSA; 2015 [Available from: https://crashstats.nhtsa.dot.gov/Api/Public/Publication/812315. 270. Bureau USC. US Population and Housing Unit Estimates: United States Census Bureau; 2016 [Available from: https://www.census.gov/programs-surveys/popest/data/data-sets.html. 271. Organization WH. WHO: Road Traffic Injuries: WHO; 2017 [Available from: http://www.who.int/mediacentre/factsheets/fs358/en/. 272. Tavris DR, Kuhn EM, Layde PM. Age and gender patterns in motor vehicle crash injuries: importance of type of crash and occupant role. Accident; analysis and prevention. 2001;33(2):167-72. 273. Agriculture USDo. USDA: Rural-Urban Continuum Codes: US Department of Agriculture; 2016 [Available from: https://www.ers.usda.gov/data-products/rural-urban-continuum-codes/.
165
274. Surgeons ACo. ACS: Searching for Verified Trauma Centers: ACS; 2017 [Available from: https://www.facs.org/search/trauma-centers. 275. Society AT. Find Your Local Trauma Center: ATS; 2017 [Available from: http://www.amtrauma.org/?page=findtraumacenter. 276. Services ADoAM. ADAMS Annual Public Atlas: ADAMS; 2016 [Available from: http://www.adamsairmed.org/public_site.html. 277. Robertson LS. Estimates of motor vehicle seat belt effectiveness and use: implications for occupant crash protection. American journal of public health. 1976;66(9):859-64. 278. Klauer SG, Guo F, Simons-Morton BG, Ouimet MC, Lee SE, Dingus TA. Distracted driving and risk of road crashes among novice and experienced drivers. The New England journal of medicine. 2014;370(1):54-9. 279. Association GHS. State Laws by Issue: GHSA; 2017 [Available from: http://www.ghsa.org/state-laws/issues. 280. Rockhill B, Newman B, Weinberg C. Use and misuse of population attributable fractions. American journal of public health. 1998;88(1):15-9. 281. Organization WH. Metrics: Population Attributable Fraction (PAF): WHO; 2017 [Available from: http://www.who.int/healthinfo/global_burden_disease/metrics_paf/en/. 282. Mansournia MA, Altman DG. Population attributable fraction. BMJ (Clinical research ed). 2018;360:k757. 283. Feero S, Hedges JR, Simmons E, Irwin L. Does out-of-hospital EMS time affect trauma survival? The American journal of emergency medicine. 1995;13(2):133-5. 284. Kononen DW, Flannagan CA, Wang SC. Identification and validation of a logistic regression model for predicting serious injuries associated with motor vehicle crashes. Accident; analysis and prevention. 2011;43(1):112-22. 285. Prevention CfDCa. Buckle Up: Restraint Use State Fact Sheets 2015 [cited 2017. Available from: https://www.cdc.gov/motorvehiclesafety/seatbelts/states.html. 286. Services FICoEM. FICEM Annual Report to Congress 2013-2015: FICEMS; 2013 [Available from: https://www.ems.gov/pdf/2011-2012-FICEMS-RTC-Mikulski.pdf. 287. Alali AS, Fowler RA, Mainprize TG, Scales DC, Kiss A, de Mestral C, et al. Intracranial pressure monitoring in severe traumatic brain injury: results from the American College of Surgeons Trauma Quality Improvement Program. J Neurotrauma. 2013;30(20):1737-46. 288. Byrne JP, Geerts W, Mason SA, Gomez D, Hoeft C, Murphy R, et al. Effectiveness of low-molecular-weight heparin versus unfractionated heparin to prevent pulmonary embolism following major trauma: A propensity-matched analysis. J Trauma Acute Care Surg. 2017;82(2):252-62. doi: 10.1097/TA.0000000000001321.