Consumer information concerning the predicted 'safeness' of a new car
model is based on the results of crash tests. Unfortunately, because i
t allows comparisons only within size/weight groups, the information i
s somewhat incompatible with the normal car-purchase decision process
since consumers often consider cars within different groups. In additi
on, based on past research, the association of the crash-test informat
ion with real-world crash outcomes is, at best, somewhat limited. The
goal of this study was to explore a methodology for improving this inf
ormation, a methodology which incorporates not only the crash-test inf
ormation, but also information concerning real-world occupant injury e
xperience in prior crashes involving similar vehicles ('clones'). The
clone information included both driver injury severity in past clone c
rashes from the North Carolina accident file and various indicators of
relative driver injury in clones extracted from published insurance-r
elated data from the Highway Loss Data Institute (HLDI). Final models
developed included both measures of the Head Index Criteria (HIC) from
the crash test and some measure of clone performances as significant
predictors. While the North Carolina clone data is intuitively 'cleane
r' in that it describes injury level per crash rather than per insured
year, the medical claims indices from the HLDI data consistently were
shown to be the stronger predictors. Future research will need to loo
k at ways of better combining the crash-test variables and of possible
modifications to the HLDI indices. In general, the analyses generated
encouraging results that appear to point to possible improvements in
the crashworthiness information. Copyright (C) 1997 Elsevier Science L
td.