Estimation real number of road accident casualties

Deliverable D1.15 : final report on task 1.5 of the EU FP6 project SafetyNet
Broughton, J.; Amoros, E.; Bos, N.M.; Evgenikos, P.; Hoeglinger, S.; Holló, P.; Perez, C.; Tecl, J.

The objective of Task 1.5 of the SafetyNet IP has been to estimate the actual numbers of road accident casualties in Europe from the CARE database by addressing two issues:

  • the under-reporting in national accident databases and
  • the differences between countries of the definitions used to classify injury severity.

Currently, the only comparable measurement units available in CARE are the numbers of fatal accidents and of people killed, where the degree of under-reporting is acceptably small in most EU Member States and there is a common definition. The same is not true, however, of non-fatal accidents and of casualties who are not killed. As a result, at present the numbers of non-fatal accidents and of people seriously and slightly injured cannot be compared in different Member States. In addition, the definition of injury severity differs among member states, so that a casualty which would be recorded in one country might not be recorded in another. Equally, a casualty which might be recorded as ‘seriously’ injured in one country might be recorded as ‘slightly’ injured in another.

As a result of this lack of comparability, international comparisons of road safety focus entirely on fatal accidents and fatalities, which form only a small minority of the totals. It is highly desirable to extend these comparisons to include the full range of injury severities. The objective of Task 1.5 has been to allow this to happen.

In order to overcome the inconsistencies in the reporting of non-fatal casualties, this Task has:

  1. estimated the under-reporting level for non-fatal casualties by developing a uniform methodology and applying it in several EU countries,
  2. estimated the number of serious casualties per country according to a new common measurement unit.

This report documents the results that have been achieved:

  • The study began by agreeing a common methodology that would be applied by all partners in Task 1.5 for their studies.
  • Studies were carried out in 8 countries according to this methodology, and the report contains detailed descriptions of the individual studies.
  • In each study, files of police and hospital records were assembled for the road accidents that occurred in a common area. These files were compared to identify matching records, i.e. those casualties who were present in both files. For these matching records, certain medical details were added to the police records: length of stay in hospital and injury severity (specifically the Maximum Abbreviated Injury Score MAIS, an internationally accepted summary measure of injury severity).
  • Two matrices were then prepared to summarise the outcome of each study, one based on injury severity and the other on length of stay.
  • These matrices were brought together for analysis, and conversion factors for each study were estimated in a consistent way. These factors allow the actual number of serious casualties in each country to be estimated consistently from police accident statistics.
  • The new common measurement unit for counting serious casualties could be based on either injury severity or length of stay. It is concluded that the most robust definition is of a non-fatal casualty with MAIS>=3 (inclusive).
  • Initial comparisons have been made of casualty data adjusted by the conversion factors estimated by the national studies.

The various national studies encountered a range of problems concerning access to the hospital data and content of the data. In general these were overcome successfully, although there were some implications for the results that could finally be achieved.

The coverage of the studies varied widely, influenced to some extent by whether hospital data had to be collected directly (as in the Czech Republic and Hungary) or were already available from files that had been compiled by national or regional authorities. The size of the datasets varies widely, depending on the size of the study area and the period included.

Ideally, the studies would have covered complete countries and so been truly national. Only 2 studies were truly national, so the question arises in the remaining 6 countries of whether conversion factors estimated from sub-national studies can be generalised to the national data. The answer must vary from country to country, but in general the larger the study area the more likely the conversion factors are to be nationally representative.

The new common measurement unit is a non-fatal casualty with MAIS>=3. Most of these are recorded by the police as seriously injured, but the studies show that the police record some as slightly injured. Consequently, according to this definition the number of casualties C in a particular country is estimated as

C = N1 * police reported serious casualties + N2 * police reported slight casualties

where N1 and N2 vary from country to country. The overall factors from 7 studies are shown below (they could not be estimated in Austria because of data limitations). N2 is considerably smaller than N1.

It was originally envisaged that the conversion factors would be generalised to other countries, in order to increase the utility of the CARE database. However, the results have led to the conclusion that this would not provide reliable results. The only satisfactory approach would be to carry out comparable studies in as many countries as possible.
The results from the Dutch and UK studies have also shown that the conversion factors can change through time as police accident reporting practices evolve. Thus, studies need to be repeated regularly to update the factors.

In summary, the research that has been carried out in the course of SafetyNet Task 1.5 represents a significant step forward and allows for the first time the number of severely injured casualties to be compared meaningfully between countries. The goals of the research were ambitious, but the practical problems that were encountered have meant that some could not be achieved fully. The lessons that have been learnt will allow this type of study to be carried out more effectively in future.

The central problem of this type of study is of obtaining access to anonymised medical records. Access to these records for research purposes is often problematical. Modern linkage techniques such as those used in this study, however, make these data increasingly valuable. Ways need to be found to persuade the custodians of these data to allow them to be used for purposes that support the broader aims and welfare of society.

Gepubliceerd door
European Commission, Brussels


Dit is een publicatie van SWOV, of waar SWOV een bijdrage aan heeft geleverd.