5.7 Integrating Data
The integration of safety data provides a large number of benefits, including:
- effective management of safety (e.g. linking intermediate and final outcome data; see Identifying Data requirements);
- identifying crash risk types and locations in a more comprehensive manner (e.g. through combined use of both reactive and proactive approaches, benchmarking, better information on injury outcomes, greater knowledge of crash contributors; see Assessing Potential Risks & Identifying Issues);
- better knowledge in order to identify cost-effective solutions (e.g. through better knowledge of the infrastructure that already exists; see Intervention Selection and Prioritisation).
In addition, linked data can be used to validate other sources of information. As an example, crash database systems can either draw information directly from asset data to provide additional information on road elements, or this linkage can be used to reduce the likelihood that data entry errors will occur by validating the presence of different road features or assets. It can also be used for research on specific topics.
Key linkages include combining crash data with:
- traffic data;
- road inventory data;
- vehicle registrations;
- vehicle inspection data;
- population statistics.
The linkage process involves several stages, and can be temporary (e.g. for a specific project or policy need) or permanent (e.g. for ongoing analysis and monitoring). Data needs to be collected in a format to facilitate linkage. This typically involves provision of a common data element, most usefully the spatial coordinates for road based elements (including crashes), while for non-spatial data, another unique identifier will be required for the datasets to be linked. A comprehensive safety information system may have a large number of component files.
Once the data has been linked, it can be analysed through merging of data files. For spatial data, a GIS software package is able to assist greatly in this task, and is particularly useful for mapping information from different sources.
Once the initial investment in collecting data has been made, it may be a relatively low cost task to join different sources of information together to meet a variety of needs, especially if a unique identifier has been used in each dataset. In other circumstances, especially where data is not in a compatible format, the task might be quite substantial involving considerable investment.
One of the more commonly used linkages is the calculation of crash rates to allow either benchmarking or identification of high risk locations. For example, crash data can be combined with population figures, traffic volumes, or vehicle registrations to provide a useful comparison of risk. Ideally, each of these would be presented as fatal and serious injury crash rates. Each of these measures is useful for different purposes, as outlined below:
- Crashes per 100,000 population: This measure reflects the direct impact of road crashes on a country, region or community. It provides a useful basis to compare road safety outcomes with other types of risk (e.g. risk of death from heart disease). It can also be used to identify risks for different subsets of the community (e.g. risk by age, gender, location). This is one of the most commonly used comparisons in road safety, particularly because it is easy to collect relevant data to make this calculation.
- Crashes per vehicle kilometres travelled (VKT): This measure reflects the level of safety based on the amount of travel undertaken. It can be used to compare different travel modes (such as between car, bus or train), different road types (e.g. undivided compared with divided roads) or infrastructure (e.g. roundabouts compared with traffic signals). This requires good knowledge of traffic volumes and distances travelled, and this information is often difficult to collect.
- Crashes per 10,000 registered vehicles: This measure is often used as a proxy for the amount of travel, as it is much easier to collect than kilometres travelled. It can be useful for analysis of performance at a national level, but has limited applicability for more detailed analysis.
Crash data can also be combined with road inventory data. At a simple level, this can provide information about current road features that may be present, providing information about possible infrastructure safety improvements. For example, crash data of run-off-road crashes could be presented alongside information on current roadside barrier locations on a map to allow for a quick visual analysis of locations that might benefit from further barrier installation.
Combining data on crashes with roadway, asset, environment, and traffic volume data can lead to some important outcomes relating to safety performance of infrastructure. It is possible to compare the safety performance of different types of infrastructure for different traffic volumes. For example, the performance of divided and undivided roads can be compared for different traffic volumes. In addition, crash performance of different infrastructure can be compared at different levels of traffic volume, for different road user types, or for different environment types (e.g. low versus high speed environments). Box 5.18 provides information on the US Highway Safety Information System.
Box 5.18: US highway safety information system
The US Highway Safety Information System (HSIS) is managed under contract to the Federal Highway Administration (FHWA). HSIS contains crash information (e.g. collision type, severity, vehicle information, sex and age of occupants, objects struck, and weather conditions), inventory (e.g. road type and function, cross-section, number of lanes, lane and median width, shoulder width and type), and traffic volume data for several States. Information on curve/grade and intersection variables is also available from some States. The combination of these different sources of data allows powerful analysis to be conducted on specific road safety issues.
A large number of studies have been conducted using this rich source of information. This has led to the production of various research reports, summaries and tools. Recent examples include a study that examined the safety effects of horizontal curves and grades on rural two-lane highways; a safety evaluation of lane and shoulder width combinations; an evaluation of the safety benefits of transverse rumble strips on approaches to stop-controlled intersections in rural areas; and a review of the safety benefits of ‘road diets’ (converting four lane arterial roads to two lanes, plus a central two-way turn lane).
Further information on the HSIS can be found at the following website: www.hsisinfo.org
Recent initiatives in integration have involved the combination of crash data and road risk assessment data. This provides a very powerful tool for identifying risk locations and possible solutions. Further information on the combination of this data can be found in Combining Crash Data and Road Data and Intervention Selection and Prioritisation.
Pathway to EfFective Management and Use of Safety Data
- An assessment of data requirements should be made.
- For countries with no comprehensive crash data, information on final and intermediate outcomes should be collected for high risk routes (e.g. high volume roads) to allow measurement of safety problems and identification of measures. This collection could be undertaken as part of a corridor demonstration project.
- A crash data system should be put in place. The steps required for this include the need to: assess current data sources; engage with key stakeholders (the road agency, police and the health sector are especially important); develop a crash report form; develop a data system; and put in place a process to ensure data quality.
- The focus should be on the collection of the range of data needed to address fatal and serious injury crash outcomes which will include exposure data, final outcome data as well as intermediate outcome data.
- Road infrastructure/asset data collection should be considered to inform safety decisions, for instance through a road assessment programme. This can provide information on likely high risk crash locations as well as affordable treatments in the absence of comprehensive crash data.
- A data collection strategy should be developed to ensure that essential information is collected.
- The crash data system should be routinely checked for accuracy and completeness (e.g. by comparing police and hospital data).
- The database should include basic features to allow comprehensive analysis of crash problem types, and be fit for use by the required stakeholders.
- Information on road assets relating to safety outcomes should be collected.
- Countries should be encouraged to aggregate data at national level, matching the IRTAD structure and progressively contributing to the IRTAD database.
- Other data relevant to the setting and monitoring of road safety targets and trends should be collected, and the accuracy of this data assessed.
- All outputs (such as reports) should be assessed to ensure that they are fit-for-purpose and address the needs of key stakeholders.
- A comprehensive data collection strategy should be put in place and regularly monitored to ensure that it is fit-for-purpose, accurate and complete.
- A crash data base should be fully implemented that contains all crash data. Data should be spatially coded, and appropriate quality control checks should be put in place.
- Information on road assets relating to safety outcomes should be contained within a comprehensive roadway inventory database. This may require the development of a database, or linkage to an existing database.
- Linkages should be made between key sources of data, particularly between data collected by police and hospitals, and between crash and asset data.