Improving Surveillance of Michigan Motor Vehicle Traffic Crash Deaths by Linking Death Certificate and FARS Data

BACKGROUND: Accurate surveillance of motor vehicle traffic crash (MVTC) deaths is essential to injury epidemiology. Death certificates are often used for this purpose. Due to the structure of ICD10 cause of injury coding, a large proportion of those killed in MVTCs (e.g., occupants, motorcyclists, pedestrians, pedal cyclists) are classified as "Unspecified." For example, in the U.S. between 1999 and 2010, 35% of the 491,758 total MVTC deaths had this non-specific classification.  The National Highway Traffic Safety Administration's Fatality Analysis Reporting System (FARS) is another source of information on fatal traffic crashes. FARS specifies person type for virtually all those killed in MVTCs.  However, FARS does not collect state and county of residence. Thus, these data by themselves cannot be used to calculate population-based rates.

METHODS:  Michigan death certificate data for deaths classified as "MVTC - person type unspecified" were linked with FARS data using year of death and death certificate number for the period 1999-2010. New counts for MVTC occupants, motorcyclists, pedestrians, and pedal cyclists were then generated by combining the original death certificate estimate with results from the linkage.   

RESULTS:  During 1999-2010, “person type” could not be assigned for 5,818 of the 14,123 Michigan resident MVTC deaths based on information from death certificates alone. Of these 5,818 deaths, 4,947 (85.0%) were linked to FARS files. As a result of this linkage, 4,768 deaths were reassigned from “unspecified” to “occupant,” increasing the count by 95% (from 5,041 to 9,809), and 122 deaths were reassigned to “motorcyclist,” increasing the count by 10% (from 1,201 to 1,323). There was minimal impact on pedestrian and pedal cyclist death counts (1.6% and 1.1% increases, respectively). Among occupant deaths, the percentage of cases added each year via the linkage increased dramatically from 38% in 1999 to 154% in 2010 (with a high of 193% in 2009). Using the more robust estimate did not change the age/sex profile of occupant deaths. However, the number of Blacks and residents of two large counties increased disproportionately.

CONCLUSIONS:  Analysts that rely solely on death certificates may underestimate the number of occupants and motorcyclists killed in MVTCs. Such underestimates could result in biased characterizations of these deaths. Simple linkage with FARS data can greatly improve sensitivity and representativeness.