Data for crash prevention

How did we get to the point of agitating for more and better data for injury prevention? Indeed, how did we even move from the pre-1960 mindset of ‘Accidents happen’ to ‘Injuries are caused and therefore can be prevented’, hence necessitating the demand for data to determine what are the contributory factors? This revolutionary thinking was pioneered by people like William Haddon Jr and Susan Baker.1 2 The latter was a medical ecologist, who, combining engineering technology with medical and specifically public health expertise, to create science-based tools for the analysis of motor vehicle collisions (and their injurious outcomes, usually death) and develop data-driven intervention strategies. One such tool, the Haddon Matrix, has been used for nearly five decades to identify risks and, using epidemiological principles, determine interventions that are most likely to be effective, initially for crashes. The data required to populate the Matrix needed to be extracted from existing sources and these were multiple. Data on damage to the vehicle and the environment would normally reside with transport agencies and insurance companies and on human outcomes with health facilities while police maintain investigative data depending on the circumstances of the collision. Herein lie the first two challenges when dealing with injury data—ownership and non-uniformity of data across the different sectors. Various pieces of information rest with different agencies who each collect for their primary purpose, using definitions and classifications that are peculiar but appropriate to each. Fortunately for Dr Haddon, he was the Director of the National Highway Transport Safety Administration in the USA and later of the Insurance Institute of Highway Safety, thus facilitating the congruence of these two sources, the transport and insurance agencies.

One of the advantages of the Haddon’s matrix was its ability to accommodate contributory factors that preceded and succeeded the injury event, namely the crash. This allowed analysis of data along the injury spectrum, from exposure through the event, to its outcome, be it a minor injury, a grave one perhaps requiring hospitalisation or resulting in disability or death. The tool likely furthered injury prevention beyond the original goal of preventing the injury death to prevention of serious injury and ultimately to the prevention of the injury-causing event. This too, had implications for data capture for, whereas fatal injury data is recorded on the death certificate in a standard manner, and is collected by one sector, this is not the case for non-fatal injuries.

Data for injury prevention – specialised surveillance systems

Supported by Dr Haddon, Mrs Baker initially investigated the causes of crashes and their fatal injury outcomes, but later expanded that research to all injuries. At the same time, Dr Patricia Waller, a clinical psychologist whose work on behaviours in highway safety led to the realisation that alcohol was a factor in injury outcomes not only of motor vehicle crashes complemented Mrs Baker’s research, so that the principles that could be applied to the prevention of motor vehicle injuries, could be used for the prevention of all injuries.3 Through Waller, Baker et al, from the later 1970s, the scope of injury prevention in the USA broadened beyond transport safety until by the late 1980s it included consumer product safety, fire safety, occupational health and safety, poison, sports safety and violence. Each of these areas had specialised data sources and needed data variables that were unique to the type of injury/event. Surveillance systems that collect on one type of injury and/or from one source, were established. In addition to the Fatal Accident Reporting System for highway crashes, there were/are the National Electronic Injury Surveillance System (consumer products), the National Fire Incident Reporting System, the National Crime Victimization Survey, the National Incident-Based Reporting System (which includes non-injury crimes), the Toxic Exposure System, the National Occupational Mortality Surveillance System and the National Collegiate Athletic Association Sports-Related Injury Surveillance System, inter alia. In the 2000s were added the National High School Sports-Related Injury Surveillance System and the National Self-Harm Register.4 Such vertical surveillance systems were periodically supplemented by surveys.

This trend was not restricted to the USA as the search for the most complete and comprehensive injury data could hardly be satisfied by one source, given the variety and complex nature of injuries. Canada too, in addition to provincial information systems, had several vertical surveillance systems including the Canadian Agricultural Injury Surveillance Programme, the Canadian Surveillance System for Water-Related Fatalities, the Traffic Accident Information Database, the National Work Injuries Statistics Programme inter alia.5 The Europeans had registers for traffic, occupational safety breaches and product-related/home and leisure incidents with the Netherlands also having surveillance systems for burns and scalds.6 7 Australia too, has what has been described as a ‘patchwork quilt’ approach to injury surveillance with surveillance residing in the sector with responsibility for the particular type of injury.8

Responses to issues created by multiple data sources

The multiplicity of data sources exacerbated the problem of non-comparable data with non-standardised variables and categories, so that data could not be merged to yield a complete injury profile, neither within a country nor across countries. Indeed, even definitions of key terms varied by source or location. It was not until the mid-1980s that a standard definition for ‘injury’ was proposed and widely accepted, a definition that encompassed mechanical, chemical and thermal injuries that were the physical outcomes of inadequate (too little or too much) energy transfers to the human body.9 Much effort was therefore invested in providing guidelines to make at least the capture of data from one sector, uniform within that sector. Manuals for capture of data on neurotrauma, intimate partner violence, injuries due to land mines and unexploded ordnance are just a few examples.10–12

Not only was data from the different sources not comparable but having the responsible agencies release or share their data was also a problem. The challenge of sectoral ownership of data on different aspects of injury has been addressed through different approaches. In Australia, injury data from the various sources is coordinated federally by the National Surveillance Unit of the Australian Institute of Health and Welfare. Colombia and a few Central American countries as well as South Africa, have established violence and injury (and optionally trauma) observatories. These have been described as ‘specialized informational repositories and knowledge-building centres, housing cross-referenced databases with advanced analytical and research capacities’.13 An observatory is really a tool for the mining of secondary databases, whether manually as in the original one in Cali where representatives of the several agencies with data (including newspapers) met regularly to share their data, or virtually as in the Injury Observatory of Britain and Ireland.14 Experiences in South and Central America and South Africa have shown that in localised and high-risk settings, observatories are most likely to produce effective, timely, broad-based responses, especially policy and behaviour changes.15

All-injury surveillance systems

The strategy of multiple specialised vertical surveillance systems is one that is difficult to sustain as it is very demanding of resources. In environments with limited resources, the choices are to capitalise on existing systems which already capture injury-related data, but used for other purposes, or build and maintain a single all-injury surveillance system. As mentioned earlier, the former choice comes with challenges to harmonise the data. Some countries, either by choice or through need, opted for a universal/national injury surveillance system. Norway was the first country to have a national injury surveillance system to capture all types of injuries. They were soon followed by Canada, the Caribbean, Central and South America, Asia and Africa until in the new millennium, there were national injury surveillance systems on every continent except Australia.5 14 There is only one sector which routinely sees and collects some information on all injuries, the health sector. As health events, all fatal injuries are certified and enumerated, and most serious injuries are seen at a health facility. Here, then is the ideal location for the establishment of a national injury surveillance system. There are the Norwegian Injury Surveillance System, the Canadian Hospital Injury Reporting and Prevention Programme and the Jamaican Injury Surveillance System inter alia, which are hospital-based.16 17 Because most hospital admissions (especially injuries) are via the emergency room (ER), the surveillance is initiated on contact with the ER. Although capture by this method is more comprehensive, the injuries recorded are a mixed bag from very trivial (the ER is the only free 24 hours facility available) to the most severe, depending on service utilisation opportunities and practices of the specific environment. This has implications for data quality and ultimately policy decisions. Does the abundance of trivial injuries carry as much weight as fewer but graver injuries in the determination of interventions?

Injury data coding sets – ICD10 versus ICECI

There are limitations too, to the data captured by the health sector which tends to be primarily clinical. While there is standardisation in the classification of health outcomes via the International Classification of Diseases (ICD), with codes from every health outcome, earlier versions considered mainly the nature of the injury and the body site with little on the associated event combined into the code. However, a concussion could be the result of an accidental fall, a motor vehicle collision, an occupational safety breach or an assault, each of which would be the target of a different type of intervention. The public health approach to the epidemiology and prevention of injuries informed the eighth version of the ICD (ICD-8) with an expansion of associated injury events and the addition of a Supplementary Classification of External Causes of Injury and Poisoning in the Ninth Revision (ICD-9) in the late 1970s—the E-codes.18 In the Tenth Revision, this supplementary classification was incorporated into the core classification with its own chapter and instructions. These codes are used to identify the underlying cause of death and combine information on some injury risk factors namely, the mechanism of energy transfer, the type of energy transferred, the location of the injury event and in a limited way, the activity of the injured person. It should be noted that the E-codes were used only for deaths and that many hospitals code only to the nature of the injury, so that data on non-fatal injuries, data critical for prevention was missing from hospital records. Thus, there has been a major thrust for hospital injuries to be coded to both nature and external cause of injury in a manner similar to the dagger and asterisk code used for other health conditions.

As injury research using the E-codes grew, the limitations to the classification became even more apparent. Additional risk factors and more detail on existing factors were required but could not be accommodated by the current coding system. For example, the E-codes do not capture the victim/perpetrator relationship in violent events, or the proximal factors in a self-harm incident. Nor was the list of products/agents that could inflict injury exhaustive. To address the deficiencies, the International Classification of External Cause of Injuries (ICECI) was developed as a complement to the External Cause Chapter of the ICD-10. The ICECI, which is a member of the Family of International Classifications and related to the ICD, is multiaxial, hierarchical and modular in structure with options to pick, choose and combine so that it can be adapted to any environment, resource-limited or resource-rich.19 The core module which is common to all injuries, contains six mandatory data elements and a text summary—age and sex of the victim and with respect to the circumstances of the event, intent, mechanism, location and activity with these optional elements, namely date, time, object, alcohol use, drug use, severity and disposition. Additionally, there are five modules which capture additional data on special types of injuries, namely assault, self-harm, transport, occupation and sports (figure 1). Each module has a minimum set of mandatory data elements which must be collected and an optional expanded data set (figure 2). Whereas the ICD-10 has one code for the external cause, the ICECI has a code for every data element. Like the ICD-10, every data element is coded such that more characters in the code reflect greater specificity. Hence, the code for spouse or partner in perpetrator/victim relationship is 1, the codes for legal spouse and ex-partner are 1.1 and 1.5, respectively. Therefore, an epidemiologist or researcher wanting to collect injury data has a choice of collection systems, from the simplest consisting only of the minimum data elements of the core module with single digit values to the most complicated with all modules containing minimum and optional data sets whose elements carry two-digit and three-digit codes. With its established correspondence to the ICD-10, the ICECI ‘provides guidance to injury researchers and practitioners on how to classify and code data on injuries according to agreed international standards’.20

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Modular structure of the International Classification of External Cause of Injuries.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Core module – minimum and optional data sets.

Surveillance versus surveys

The introduction of the ICECI with its flexibility, facilitated the development of injury surveillance systems in diverse environments. In an effort to promote standardisation, the WHO in collaboration with the Centers for Disease Control and Prevention and drawing on the practical experience of injury researchers globally, developed the Injury Surveillance Guidelines which, in addition to giving the simplest instructions on how to establish an injury surveillance system in any setting, provides templates of data collection instruments using the broadest coding sets of the ICECI.20

The data on non-fatal injuries is complemented by data on injury deaths, extracted from death certificates. It was stated earlier that fatal injuries tend to be certified and enumerated, according to law. However, in places and situations where forensic facilities are overburdened, there may be lengthy delays in autopsies and coroners’ findings (to which injury deaths are subject) and sometimes, some deaths are missed. Even when the autopsies have been conducted, the autopsy or coroner’s report may not include key epidemiological information. To this end, WHO in partnership with Monash University’s Department of Forensic Medicine, produced, ‘Fatal Injury Surveillance in Mortuaries and Hospitals: A Manual for Practitioners’, a publication which ‘focuses on improving the documentation of fatal injuries and violence’.21 In other areas, verbal autopsies may be the only information available on some deaths. The proposed 11^th version of the ICD, ICD-11, will permit capture of data from verbal autopsies and will address the shortcomings of the ICD-10 which ICECI attempted to overcome.22

Finally, having captured data on fatal injuries and on injuries that present to the ER, there are still those injured persons who may not have access to a health facility or choose not to be treated at a facility within the surveillance system. Population-based non-fatal injury data may only be captured through community-based surveys, and to this end, WHO has produced a document which walks any injury researcher through the techniques of conducting a survey and provides sample data collection instruments.23