Using data linkage to generate 30-day crash-fatality adjustment factors for Taiwan

https://doi.org/10.1016/j.aap.2005.12.015Get rights and content

Abstract

Different countries have their own police reporting time standards for counting the number of fatalities in reported crashes. A rapid estimation method (such as adjustment factor) for the comparison is important. The data-linkage technique was used to combine police-reported crash data and vital registration data, in order to generate 30-day fatality adjustment factors for various reporting time standards, which could also shed light on the fatal injury trend over time. The major findings were as follows. Firstly, a conservative 30-day fatality adjustment factor for the first day (or 24 h) would be 1.54 (or 1.35) in an area with a large motorcycle population, like Taiwan. This produced 20–40% higher 30-day fatalities than UK Transport Research Laboratory predicted, and 15–25% higher fatalities than those in Europe/Japan. Secondly, after excluding motorcycle impacts, the Taiwanese factors suggested 8–14% higher fatalities within 30 days than those in Europe/Japan. Third, motorcycle fatalities influenced the overall 30-day fatality trend within 3 days. In the future, both the police under-reporting problem and the motorcycle/overall fatal injury pattern within 3 days after crashing in developing countries like Taiwan merit further investigation.

Introduction

Different countries have their own police reporting time standards for counting the number of fatalities in reported crashes. The reporting standards used around the world include 24 h (Japan, Spain, Switzerland and Taiwan), 7 days (Italy) and 30 days (Australia, United Kingdom, United States and New Zealand) (Jacobs et al., 2000, Pozuelo and Izarzugaza, 1996). Only those crash victims who die within this period are reported as crash fatalities by the police. Thus, the same number of fatalities reported by different countries might be based on different reporting time standards. Establishing a rapid and accurate estimation method for the comparison of crash fatalities among different countries is therefore important. Using such a method, information originating from various time standards could shed light on fatal injury trends over time.

The estimation method can be based on the fatality adjustment factor. In order to aggregate fatal data, Jacobs et al. (2000) used an adjustment factor table published by the European Conference of Ministers of Transport (ECMT) to bring those countries that were not using the 30-day reporting standard into line. The fatality adjustment factor is the ratio of the number of fatalities within an acceptable common time period, such as 30 days, divided by the number of fatalities according to the relevant police reporting time standard. For example, if 2401 fatalities are reported by police using the 1-day reporting standard and 3702 fatalities are estimated within 30 days, then the 30-day fatality adjustment factor for the 1-day standard is as follows: 3702/2401 = 1.54. The police-reported fatalities (1-day standard) can therefore be multiplied by a factor of 1.54 to estimate the number of fatalities within 30 days, and then compared with other reported or estimated 30-day values. A higher value of this factor indicates a higher 30-day fatality.

Although there is no agreed international common time definition of a crash fatality, a special report by the International Road Traffic and Accident Database (IRTAD) concluded that a consistent inter-country definition of a crash fatality would be “an injured person who dies within 30 days of the crash” (Pozuelo and Izarzugaza, 1996). Jacobs et al. (2000) also used the 30-day definition. Moreover, the Japanese Government published a 30-day fatality adjustment factor for 24 h police-reported fatalities (Japan National Police Agency, 2002). Based on the abovementioned studies, the 30-day definition (according to which the injured person must die within 30 days of the crash) seems to be most widely adopted among major countries. Accordingly, 30 days was adopted as a common fatality definition in the present study.

In addition to fatalities comparisons, the reciprocal of the fatality adjustment factor (the fatality proportion) expresses the fatality within the reporting standard as a proportion of that within the common time definition. Continuing the example used above, the fatality proportion of the 1-day reporting standard under the 30-day definition is as follows: 1/1.54 = 0.65. Thus, within 30 days of the crash, the deaths on the first day account for around 65%. This represents the fatality proportion over time. If fatality adjustment factors can be obtained for various reporting standards, then many fatality proportions can be calculated. These fatality proportions can then be used to observe the trend in fatalities over time.

The ECMT 30-day fatality adjustment factor table cited by Jacobs et al. (2000) is the only available complete dataset that shows the time profile in days, beginning from the scene/1 day. This could be used to calculate many European fatality proportions over time. Additionally, Japan reported a 30-day factor for the 24 h standard, which standard was not in the ECMT factor table. Although the factor data from the ECMT and Japan were generated using different methods, they demonstrate the variation of fatality data with different police reporting standards, which was of major interest in the current study. We therefore included all of these data in our comparisons.

In order to produce a 30-day time profile of fatality, it is essential to know the time of death of victims after involvement in a crash. It is then possible to count how many of the fatalities would be included by the various possible reporting standards. Although a follow-up survey can determine the exact time of death, this approach has several limitations. First, the cost of this method precludes its use for large numbers of crash victims. Second, an injured person or their family might be lost from contact because they have moved house, are unwilling to respond, or the injured person has died and has no living relatives. A clerical review of vital registration documents is necessary to obtain information about a fatal case where contact has been lost with the individual's family. All such cases require great human effort to access the full information. Rather than a surveyor contacting an injured person and carrying out a clerical review, data linkage provides an effective alternative and was therefore used in this study.

The majority of crashes reported in previous research were attributed to car accidents. The motorcycle, however, is the preferred method of transportation among the Taiwanese. The population of Taiwan in 2002, which numbered over 22 million persons, reported ownership of 11,983,757 motorcycles and 5,923,200 cars. The number of motorcycles was almost double that of cars, which is significantly different from the pattern seen in other countries. The high proportion of motorcycles mixing with other road users constitutes a potentially more dangerous environment, which might cause greater numbers of fatalities and consequently affect the estimation of fatality adjustment factors. This was considered in the current paper.

Accordingly, our study initially used data linkage to generate 30-day crash-fatality adjustment factors for Taiwan, using various presumed reporting standards, and compared these factors and the trends over time with those in Europe/Japan. We also examined whether the high proportion of motorcycles in Taiwan had an impact on fatality factors and fatal injury trends.

Section snippets

Data linkage

In most previous studies, databases used were to protect anonymity, so unique identifiers were usually not available. In such cases, linking was dependent upon using several identifying variables (for example, name and street address), each of which was only a partial identifier, but which provided sufficient accuracy when combined for use of linked data (Kim, 1999, Li et al., 1999). To carry out data linkage, one requires a set of rules that determines the extent to which identifying variables

Results

Around 96% of the records were linked (that is, 2734 linked fatalities out of a total of 2861 fatalities reported by the police). The remaining 127 (4%) police-reported fatalities were unlinked and were excluded from this study. There were no false-linked records in the linked dataset.

As the number of linked police-reported fatalities (24 h standard) in Taiwan was 2734 and the number of injured victims who died within 30 days was 968, the total number of fatalities within 30 days after crashes

Percentage of linkage

The unlinked police-reported fatalities (4%; 127 victims) could have been caused by a lack of correct ID data, due to it being unavailable to the police, the victim being a foreigner or coding errors. Of the 2861 reported fatalities in 2002, 97 victims had no ID data in the crash reports. This accounted for 75% of the 127 unlinked fatalities.

The 96% linkage, however, was sufficiently high for the linked records to be used in this study. A 93% rate of police–death links was reported by Ferrante

Conclusions

Different crash fatality reporting standards influence fatality data. International comparisons should be made using a common time standard. Base on a common 30-day time definition, 30-day crash-fatality adjustment factors for various police reporting standards could be combined to demonstrate a fatal injury trend over time. However, a factor adjusting for the police under-reporting problem should be further investigated, in order to improve the crash 30-day fatality adjustment factor.

For a

Acknowledgments

This study was funded by a research grant from the Institute of Transportation, Ministry of Transportation and Communications, Taiwan. The opinions expressed here are those of the authors and are not necessarily shared by the Institute of Transportation. The authors would like to thank the two anonymous reviewers and editor for their suggestions and comments.

References (15)

There are more references available in the full text version of this article.

Cited by (12)

  • Survival risk factors for fatal injured car and motorcycle drivers in single alcohol-related and alcohol-unrelated vehicle crashes

    2011, Journal of Safety Research
    Citation Excerpt :

    The linkage algorithm employed in the present study consisted of four stages: (1) cleaning the records, (2) preparing the data, (3) performing the linkage, and (4) conducting clerical reviews. This method for compiling crash-related data in Taiwan has been used in previous studies (e.g., Lai et al., 2006; Li, Doong, Chang, Lu, & Jeng, 2008). The third stage (performing the linkage), however, was designed for the purpose of the present study.

  • Survival hazards of road environment factors between motor-vehicles and motorcycles

    2009, Accident Analysis and Prevention
    Citation Excerpt :

    In Taiwan, the population is over 22 million, with 12 million motorcycles (all with engine size <250 c.c.) and 5 million passenger cars, constituting a complex traffic environment. In our previous study (Lai et al., 2006), the 30-day crash fatalities were estimated to range from 1.35 to 1.54 × the number of fatalities in the first day. Such values are higher than the 1.15–1.30 reported for Europe and the UK (Jacobs et al., 2000) and Japan (Japan National Police Agency, 2002).

  • Methodological issues in motorcycle injury epidemiology

    2008, Accident Analysis and Prevention
    Citation Excerpt :

    For comparisons of fatality data based on different reporting time intervals, adjustment factors (i.e., dividing the number of fatalities within a common time interval by the number of fatalities reported in an actual time interval) are needed (Jacobs et al., 2000). For example, in Taiwan the number of motorcycle fatalities within 30 days was estimated by multiplying the number of 24-h fatalities by a factor of 1.62 (Lai et al., 2006). In addition, fatality data from different time intervals can lead to different results on the effectiveness of helmet use or other exposures on motorcycle deaths; for instance, the effectiveness of helmet use was 70% in Taiwan (Keng, 2005) compared to 37% in the US (Deutermann, 2004), since helmets more efficiently prevent brain injuries to motorcycle riders who die within 24 h, if nonhelmeted.

View all citing articles on Scopus
View full text