FARS is a comprehensive census of all traffic related fatalities in the United States. FARS data is available from 1975 to the present and includes records of approximately 30,000–40,000 fatalities which occur on U.S. highways each year. FARS contains records of traffic fatalities in all vehicle types and crash modes, i.e. cars, light trucks, heavy trucks, bicyclists, motorcyclists, and pedestrians. In injury biomechanics, FARS is useful for determining the characteristics of fatal crashes.

FARS has been maintained and operated by the National Highway Traffic Safety Administration (NHTSA) since 1975 [13]. For a case to be included in this database, it must involve a motor vehicle traveling on a primarily public roadway and death of an individual involved within 30 days of the incident. Each incident is characterized by the collection of approximately 175 data elements split among an accident table, a vehicle table, and a person table. In 2010, FARS was expanded and restructured to include additional crash data in tables on crash events, contributing factors, driver’s vision, violations, pre-crash maneuvers, distraction, impairment, safety equipment, pedestrian and bicyclist crash circumstances.

In the U.S., all states must collect and provide NHTSA with records of all traffic related fatalities on their highways. FARS data can be obtained by downloading any of the published files from the NHTSA web site. The files are available in SAS, DBF, and sequential ASCII file formats. FARS data is assembled by FARS analysts located in each state who supplement each jurisdiction’s police accident report with data from death certificates, coroner/medical examiner reports, driver history data from the state department of motor vehicles, and toxicology from police or medical facilities. The injury severity suffered by each person is coded using the KABCO scale. Note that FARS includes fatalities within 30 days of the MVC and the KABCO scale is an assessment of occupant injury severity on scene, so the KABCO assessment can vary (“Killed”, “Incapacitated”, etc.) for FARS occupants.

NASS/CDS provides a detailed record of a national sample of approximately 5,000 tow-away crashes investigated each year by NHTSA at 24 locations throughout the United States [14]. This database includes a probability sample of thousands of minor, serious, and fatal crashes involving cars, light trucks, vans and sport utility vehicles. Compared with the FARS database, the data collected in NASS/CDS is much more detailed and includes approximately 400 data elements. Each case provides detailed information describing the collision including vehicle characteristics, severity of each injury, crash configuration, occupant contact sources inside the vehicle, and occupant restraint performance. Trained crash investigators obtain data from crash sites, studying evidence such as skid marks, fluid spills, broken glass, and bent guard rails. They locate the vehicles involved, photograph them, measure the crash damage, and identify interior locations that were struck by the occupants. These researchers follow up on their on-site investigations by interviewing crash victims and reviewing medical records to determine the nature and severity of injuries. By applying weighting factors to NASS/CDS data, a US representative population can be analyzed [15]. Injury severity is coded using the AIS scale.

NHTSA requirements for NASS/CDS crash investigations changed in 2009 and many variables (including all injury data) are not collected for model year vehicles greater than 10 years old. Model year vehicles greater than 10 years old accounted for approximately one-third of the unweighted NASS/CDS 2009–2011 cases. Excluding these NASS/CDS 2009–2011 cases with model year vehicles greater than 10 years old, the NASS/CDS 2000–2011 dataset contains 54,703 cases, 94,283 vehicles, 115,159 occupants, and 303,230 injuries (Table 2.3).

GES is a comprehensive database containing information on approximately 60,000 randomly sampled police reported accidents each year. GES can be analyzed to determine the number of occupants who were involved – both injured and uninjured. Because GES contains all crashes, GES analyses are frequently conducted to provide a measure of exposure when computing injury or fatality risk in particular crash modes. Cases from GES are assigned weights that can be used to estimate the number of similar accidents that may have taken place that year that were not sampled. GES data was first made available by NHTSA in 1988. Beginning in 2009, FARS and GES were restructured to standardize the definitions of common variables [16].

The function of NASS/GES is to present a representative sample of all police-reported motor vehicle accidents in the United States [17]. Criteria for selection include the involvement of a motor vehicle traveling on a public road and a crash that results in property damage, occupant injury, a fatality, or any combination of these outcomes. The accident reports are sampled from approximately 400 police jurisdictions in 60 primary sampling units across the United States. Each area is selected with the intent of providing a spectrum of regions that is indicative of geographical, roadway, population, and traffic characteristics in the entire country. Complexity of collected crash information is analogous to that of the FARS database and includes approximately 130 data elements. Injury severity is coded for each person using the KABCO scale. In 2012, NHTSA began an effort to modernize NASS which will include improving the IT infrastructure of NASS and FARS, updating NASS data collected, and reexamining the NASS sample design [18].

NMVCCS is a nationally representative dataset that includes a probability sample of over 6,900 U.S. crashes involving 13,300 vehicles which occurred from July 3, 2005 to December 31, 2007 [19]. NMVCCS is an in-depth crash investigation database similar to NASS/CDS. However unlike NASS/CDS, NMVCCS focuses on crash causation rather than crash outcomes. In order to be included in NMVCCS, Emergency Medical Services (EMS) must have been activated and an investigator must have been at the scene of the crash before it was cleared. This allowed investigators to interview occupants, witnesses, and first responders in order to determine crash causation factors not available in traditional crash databases which focus on injury. The NMVCCS database was developed by NHTSA as a means to obtain insight into the primary pre-crash circumstances and behaviors that are associated with driver crash risk in light vehicles.

During data collection, NMVCCS relied on special arrangements between crash investigators, EMS, and police agencies as well as constant monitoring of crash occurrences with the aid of police scanners to allow for immediate crash-site investigations and on-site driver interviews. To further ensure the accuracy of the data and inhibit the loss of critical information, the NMVCCS protocol required that a responding officer was on-scene at the time of the crash investigation and a particular focus was placed on driver interviews. This provided an opportunity to collect evidence and conduct interviews with the involved parties immediately after the crash regarding the pre-crash events and behaviors. Injury severity is coded for each person using the KABCO scale.

In the U.S., many states maintain databases containing digital records of all police accident reports. The datasets are large and are used by many injury researchers to obtain exposure data which is used to compute injury or fatality risk. State data is similar in function to GES. An example is the NJCRASH database of all police reported crashes in New Jersey. NJCRASH contains electronic summaries of the approximately 300,000 police reported crashes that occur each year in New Jersey. NJCRASH can be downloaded from the New Jersey Department of Transportation web site.

State data can be obtained in a number of ways: (1) directly from the states, (2) from NHTSA through their state data system, and (3) through the Highway Safety Information Systems (HSIS). HSIS is a collection of state data from seven states (California, Illinois, Maine, Minnesota, North Carolina, Ohio, and Washington State) which is maintained by Federal Highway Administration (FHWA). Additional historical state data is also available through HSIS for Michigan and Utah.

The advantage of using state data over GES is that state data is a census of crashes and avoids the analytical difficulties of using a probability sample which is the basis for GES. The difficulty with state data is that (1) it can be difficult to obtain from the states, (2) reliance on a single state may not be representative of other states in the U.S., (3) there is lack of standardization, and (4) typically injury is only coded using the KABCO scale. One approach to achieve regional balance is to combine the state data of several states for an analysis. However, the lack of standardization can make this challenging. The lack of standardization may be resolved in the future. Many states are now adopting the Model Minimum Uniform Crash Criteria Guideline (MMUCC) which provides standardized definitions for many critical police-reported crash characteristics [20].

One promising method of combining the benefits of large state accident databases with the detailed medical coding of injuries practiced in CIREN and NASS/CDS is to link databases of police accident reports with associated medical records. NHTSA founded an effort known as Crash Outcome Data Evaluation System (CODES) which includes the efforts of numerous state Departments of Transportation for linking vehicle, medical, and insurance related sources [21, 22]. CODES links police accident reports, EMS records, emergency department records, and hospital discharge records to build a comprehensive set of crash records with associated medically-based accounts of injury. CODES was implemented by states on a voluntary basis with NHTSA sponsorship. The number of states in the system varied over the course of the CODES program. In 1996, there were seven CODES states (Hawaii, Maine, Missouri, New York, Pennsylvania, Utah, and Wisconsin). In 2009, there were 16 CODES states (Connecticut, Delaware, Georgia, Illinois, Indiana, Kentucky, Maryland, Massachusetts, Minnesota, Missouri, Nebraska, New York, Ohio, South Carolina, Utah, and Virginia). Although the CODES program has now ended, many states still maintain these linked crash-medical records data programs.

Since 1972, the Special Crash Investigations (SCI) Program by the National Center for Statistics and Analysis (NCSA) has provided NHTSA with detailed crash data collected for the purpose of studying new and rapidly changing vehicle technologies or special crash circumstances. Approximately 200 SCI cases are collected annually from NHTSA’s Auto Safety Hotline, the Department of Transportation’s National Crash Alert System, NHTSA’s regional offices, automotive manufacturers, other government agencies, law enforcement agencies, engineers, and medical personnel. Case selection is based on the program manager’s discretion and may include investigation of new emerging technologies such as the safety performance of alternative fueled vehicles, child safety restraints, airbag systems, adapted vehicles, safety belts, vehicle-pedestrian interactions, and potential safety defects. SCI crashes have been used widely in the past to study airbag systems and school bus crashes [23, 24]. Data collected on SCI cases may include police reports, insurance crash reports, and reports from professional crash investigation teams. Hundreds of data elements are collected describing the vehicle, scene, occupants, safety systems, injuries, and injury mechanisms. SCI cases can be queried using NHTSA’s electronic case viewer [25].

CIREN is an in-depth crash investigation system sponsored by NHTSA and select automakers. CIREN investigates approximately 300 cases each year which are collected by six level 1 trauma centers. CIREN teams are multi-disciplinary teams which combine the expertise of trauma physicians, crash investigators, biomechanics and automotive engineers, and emergency medical personnel to determine the detailed injury mechanisms associated with a crash.

CIREN cases provide extraordinary detail on the injuries suffered by each crash victim. Injuries are coded using the AIS scale, and members of the CIREN network have access to radiology and other medical data which support enhanced diagnosis of injury mechanisms. CIREN is however a clinical sample of persons who enter a level 1 trauma center, and, unlike NASS/CDS, is not nationally representative of the U.S. population. Methods have been developed by Stitzel et al. (2007), Yu et al. (2008) and Elliott et al. (2010) to assess CIREN and NASS similarity and allow national estimates of injury outcomes to be estimated using CIREN [26–28].

Summaries of CIREN cases can be reviewed using the NHTSA online CIREN case viewer [29]. CIREN is a sample of seriously injured MVC occupants. MVC occupants admitted to a level 1 trauma center at an enrolling CIREN center are eligible for enrollment in the database if they meet certain inclusion criteria. Case occupants 13 years of age and older must sustain an AIS 3 or higher injury, with limited exceptions for AIS 2 injuries in multiple body regions or injuries of special interest to NHTSA. The model year of the vehicle of the case occupant must also be within 6 years of the time of enrollment unless NHTSA approves enrollment of an older vehicle. Cases of catastrophic impacts and occupant ejection are also excluded because the injury causation scenarios are difficult to define and often do not yield information that could be used in countermeasure development [30, 31]. Through the CIREN program, detailed vehicle, crash, and medical information is collected for each patient and presented at a multidisciplinary case review. The goal of each case review is to bring together engineering and medical knowledge to assess the crash mechanics, biomechanics, and clinical aspects of an injury. All cases undergo a case review with medical, engineering, and crash reconstruction specialists to determine injury causation [32].

PCDS is a 5-year compilation of pedestrian crash data collected from six major United States cities from 1994 to 1998 [33, 34]. The database focused on late model year vehicles that struck pedestrians. The PCDS contains over 500 cases with detailed information describing the collision including injury severity, vehicle characteristics, and crash configuration. The U.S. National Highway Traffic Safety Administration conducted the study to better define the problem of pedestrian safety and to compare current data with previously conducted pedestrian reports to determine any modifications in trends over the years. PCDS is invaluable for the analysis of pedestrian crashes, both fatal and non-fatal. Each crash was investigated in detail, and provided information unavailable through FARS, NASS/GES, or NASS/CDS, including detailed descriptions of injuries. Although this is an older dataset, PCDS is still widely used for pedestrian studies in the U.S. At the time of this report, there was no more current in-depth crash investigation data on U.S. pedestrian crashes. PCDS has been used to understand the fatality and injury risk of pedestrian crashes [35, 36].

The National Trauma Data Bank (NTDB) is the largest aggregation of trauma registry data ever assembled [37]. It is supported by the American College of Surgeons (ACS) and provides information about patients, injuries, and treatments. NTDB collects trauma registry data from participating trauma centers on an annual basis. All hospitals with trauma registries are encouraged to participate and all institutions that contribute are either ACS-verified, state-designated, or self-designated as trauma centers. The criteria for participation in the database is based on the reporting institution’s trauma accreditation and hospital or emergency department admission. Data submitted to NTDB are rigorously examined using both the NATIONAL TRACS system (Digital Innovation, Inc., Forest Hill, MD) institutionally and an additional logical checks system created and enforced by NTDB administrators. Data regarding patient demographics, injury severity, and injury origin is collected as well as descriptive accounts of each traumatic incident. Patients who are dead upon arrival are excluded from the NTDB. NTDB contains injuries coded using both the ICD-9 and AIS coding lexicons.

The NTDB Research Data System (RDS) contains all records sent to NTDB for each admission year. NTDB-RDS version 7.1 contains 1,926,245 cases from 2002 to 2006 admission years (Table 2.4). ICD-9 ecodes 810–819 with a post-dot decimal of 0 or 1 which correspond to motor vehicle traffic accidents involving drivers and passengers were used to designate MVC cases, resulting in 529,362 cases. No crash data is available from NTDB aside from ICD-9 ecodes indicating MVC as the mechanism of injury, but NTDB serves as a large dataset of AIS and ICD-9 coded injuries for studying traumatic injury.