Validation of a Data Resource for Research in Motor Vehicle Fatalities in Children

Background. The Fatality Analysis Reporting System (FARS) and Multiple Causes of Death (MCOD) datasets are two important resources for vehicular injury researchers. Until recently, linkage of these data was not possible, making research into associations of crash and vehicle characteristics with cause of death and validation of shared concepts impossible. Methods. FARS and MCOD were linked by death certificate number to obtain cause of death for children (0 to 15 years) who died in motor vehicle crashes (MVCs). Three taxonomic axes were used for the validation: Role (driver, passenger, pedestrian), Occupant-of (vehicle types, including cars, vans, and trucks), and Collided-with (vehicle types and stationary objects). These axes were based on conceptual intersections between the two contributing datasets: the axes in FARS were represented by specific variables and in MCOD by ICD-10 codes for external causes of injury associated with motor vehicle accidents (V01-V89). Analysis was restricted to death years 1999-2002 for 47 states and the District of Columbia, the underlying cause of death in MCOD and children identified in FARS as passengers in passenger vehicles. Results. A total of 6065 child passengers died in MVCs during the observation period. Agreement between FARS and MCOD was good (>99%) except for Occupant-of/Car (57.1%), Occupant-of/Truck (72.4%), Role/Car passenger (57.7%), Role/Truck passenger (75.1%), Collided-with/motor vehicle (70.5%) and Collided-with/Fixed object (81.2%). Conclusions. The linked FARS-MCOD data provide a valid resource for research into MVC-related child fatalities, but investigators should use caution in using only immediate cause of death or FARS-specific variables that share similar concepts.