Estimating non-fatal road casualties in a large French county, using the capture–recapture method
Introduction
Non-fatal road casualties are known to be under-reported, with official figures usually coming from the police forces. Hospital discharges are sometimes used as a second source of reporting. When two or more sources exist, it is possible, under certain conditions, to use the capture–recapture method to estimate the number of cases missed by all sources and hence obtain an incidence rate that is corrected for under-count. The capture–recapture method has been used in epidemiology (IWGDMF, 1995a, IWGDMF, 1995b, Gallay et al., 2002, Hook and Regal, 1995) including road injury (Aptel et al., 1999, Dhillon et al., 2001, Meuleners et al., 2006, Razzak and Luby, 1998, Roberts and Scragg, 1994, Tercero and Andersson, 2004) with two papers showing more sceptical views (Jarvis et al., 2000, Morrison and Stone, 2000).
In France, in the Rhône county, which is a large area of 1.6 million inhabitants, a road trauma registry has been in operation since 1996, providing an additional source of reporting to the police. We apply the capture–recapture method to evaluate both the completeness of the registry and to obtain an estimate of the incidence rate of non-fatal road casualties correcting as much as possible for under-count. A capture–recapture analysis stratified on injury severity measured by the New Injury Severity Score, road user type and third party involvement (yes/no) is conducted. One of the conditions of the capture–recapture method is the perfect identification of subjects common to both sources. This is always difficult to reach and to assess. In this study, common casualties are identified by a semi-automated probabilistic record-linkage. Its validity is assessed by two methods: an additional record-linkage using surnames and first names, and a method inspired by Brenner (1994) and Brenner and Schmidtmann (1996) based on the probability of agreement of the linking variables. Based on this, we elaborate three possible scenarios depending on the number of linked casualties, and we apply the capture–recapture method on each of these.
Section snippets
Material and methods
The study is based on the 2001 annual data of non-fatal road casualties, injured in crashes that occurred in the Rhône county (irrespective of the place of residence of the casualties), from the road trauma registry and from the police file. The 2001 annual data are studied as 2001 is the last year before the registry recording procedure has been modified. The Rhône county (“département”) is an area of 1.6 million inhabitants, consisting of a large city (Lyon), its suburbs and a rural area in
Results
In 2001 in the Rhône county, the police record 4135 non-fatal casualties and the registry records 10636 non-fatal casualties that fulfil the police criteria of a road crash casualty. The first scenario, corresponding to the “standard” record-linkage based on date of crash, detailed place of crash, year and month of birth, gender and road user type of the casualty, identifies 2813 casualties common to both sources.
The additional survey on 1322 “police only” records is based on only 868 available
Discussion
The capture–recapture method is based on four assumptions. The first assumption is that the population is closed, i.e. that there are no entries or losses between the two recordings. This is largely met since there is very little delay between the police attending the crash scene and the casualty's visit to the hospital: at most a few days for the most slightly injured casualties.
The second assumption is that there is independence between the two sources. We do believe that there is some
Acknowledgments
We wish to thank the following people for having participated in the data collection and data entry in the framework of the Association for the Registry of road traffic casualties in the Rhône (ARVAC: president Y.N. Martin) or in the framework of INRETS-UMRESTTE (B. Laumon, scientific consultant for the registry, and A. Ndiaye, physician, director of the registry): Ait Idir T., Ait Si Selmi T., Alloatti D., Amoros E., Andrillat M., Artru F., Asencio Y., Assossou I., Auzaneau F., Bagès-Limoges
References (21)
- et al.
Under-reporting of road crash casualties in France
Accid. Anal. Prev.
(2006) - et al.
Road accident statistics: discrepancies between police and hospital data in a French island
Accid. Anal. Prev.
(1999) - et al.
Assessment of hospital and police ascertainment of automobile versus childhood pedestrian and bicyclist collisions
Accid. Anal. Prev.
(2001) - et al.
Estimating crashes involving heavy vehicles in Western Australia, 1999–2000: a capture–recapture method
Accid. Anal. Prev.
(2006) - et al.
Measuring transport injuries in a developing country: an application of the capture–recapture method
Accid. Anal. Prev.
(2004) Application of capture–recapture methods for disease monitoring: potential effects of imperfect record linkage
Methods Inf. Med.
(1994)- et al.
Determinants of homonym and synonym rates of record linkage in disease registration
Methods Inf. Med.
(1996) Practical introduction to record linkage for injury research
Inj. Prev.
(2004)- et al.
The capture–recapture applied to epidemiology: principles, limits and application
Revue d’Epidémiologie et de Santé Publique
(2002) - et al.
Capture–recapture methods in epidemiology: methods and limitations
Epidemiol. Rev.
(1995)
Cited by (39)
Sampling bias and weight factors for in-depth motorcycle crash data in Thailand
2022, IATSS ResearchErrors in accident data, its types, causes and methods of rectification-analysis of the literature
2019, Accident Analysis and PreventionAbdominopelvic injuries due to road traffic accidents: Characteristics in a registry of 162,695 victims
2018, Traffic Injury PreventionInvestigating the reasons behind the intention to report cycling crashes to the police and hospitals in Denmark
2017, Transportation Research Part F: Traffic Psychology and BehaviourCitation Excerpt :While the under-reporting phenomenon is well-documented, very little is known regarding its reasons. It is known that under-reporting is not randomly distributed but suffers from selection bias according to crash severity, crash location, involved road users, and police control area (Amoros et al., 2007). However, it is not known what the causes are other than perceived reporting (non-)usefulness and police distrust (Amoros et al., 2007).
Record linkage for road traffic injuries in Ireland using police hospital and injury claims data
2016, Journal of Safety ResearchPregnant women in vehicles: Driving habits, position and risk of injury
2016, Accident Analysis and Prevention