Comparison of Distributions of Key Predictor Variables in CIREN and NASS-CDS Cases Meeting CIREN Inclusion Criteria

Abstract

Objective: The Crash Injury Research and Engineering Network (CIREN) database contains data from occupants who are seriously injured in motor vehicle crashes. Because of the number of injured occupants and the level of detail in CIREN, a number of previous studies have attempted to use CIREN data to estimate injury risk or relative risk by comparing the frequencies of CIREN occupants with an injury of interest to those without that injury. However, these comparisons provide little useful information because CIREN case occupants cannot provide the control and exposure information that is needed to appropriately estimate injury risk. One potential source of exposure/control information for CIREN is the database established by the National Automotive Sampling System-Crashworthiness Data System (NASS-CDS). However, using NASS-CDS for this purpose requires that NASS-CDS and CIREN cases are drawn from the same source population. The objective of this article is to assess whether this requirement is met.

Methods: Occupants from NASS-CDS that meet CIREN inclusion criteria, termed CIREN-eligible NASS-CDS cases, were selected to represent the source population for CIREN. These CDS cases were compared to CIREN cases on a number of key variables that are used in many analyses of crash injury data. Comparisons were evaluated using chi-square tests.

Results: The comparisons indicate that CIREN cases are similar to CIREN-eligible NASS-CDS cases on most variables, with some important exceptions. First, CIREN contains more frontal crashes than CIREN-eligible NASS-CDS. Second, there are fewer CIREN cases with two or more AIS 2 injuries and no AIS 3+ injuries than would be expected based on CIREN-eligible NASS-CDS. Finally, on average, occupants in CIREN have a greater number of AIS 3+ injuries, are less often belted in the front seat, and tend to be in higher severity crashes than occupants in CIREN eligible NASS-CDS.

Discussion: These differences suggest that analyses aimed at estimating either relative risk of injury by comparing CIREN cases to base rates from NASS-CDS, or estimating risk of injury using control information from NASS-CDS, should be limited to a specific crash type and injury definition, and should only be performed for AIS 3+ injuries. In addition, analyses should account for the greater overall crash and injury severity of CIREN cases, at least in interpretation of the results.

Keywords:

• CIREN
• Injury
• Automobile crash
• Risk
• Case-cohort analysis

ACKNOWLEDGEMENTS

The authors are grateful to Pat Dischinger, Dennis Durbin, Refaat Hanna, Joe Kufera, Shashi Kuppa, Cathy McCullough, Mark Scarboro, and Joel Stitzel for fruitful discussions of appropriate analysis methods for CIREN. This research was supported by the National Highway Traffic Safety Administration under contract DTNH22-05-01020.

Notes

¹Note that the use of the term case in epidemiology, and in the context of this article, is different from the usual use of the term in crash investigation literature, in which every sampled occupant or every sampled crash is called a case, whether injured or not. In this article, as in the epidemiology literature, case refers only to injured occupants.

∗AIS of 2 in 2 or more body regions with medical significance or AIS of 2 in the lower extremity with significant articular injury (pilon/talus/calcaneus/ Lisfranc/Choparts).