The new car assessment program: does it predict the relative safety of vehicles in actual crashes?

BACKGROUND

Federal motor vehicle safety standards are based on crash test dummy analyses that estimate the relative risk of traumatic brain injury (TBI) and severe thoracic injury (STI) by quantifying head (Head Injury Criterion [HIC]) and chest (Chest Gravity Score [CGS]) acceleration. The New Car Assessment Program (NCAP) combines these probabilities to yield the vehicle's five-star rating. The validity of the NCAP system as it relates to an actual motor vehicle crash (MVC) remains undetermined. We therefore sought to determine whether HIC and CGS accurately predict TBI and STI in actual crashes, and compared the NCAP five-star rating system to the rates of TBI and/or STI in actual MVCs.

METHODS

We analyzed frontal crashes with restrained drivers from the 1994 to 1998 National Automotive Sampling System. The relationship of HIC and CGS to the probabilities of TBI and STI derived from crash tests were respectively compared with the HIC-TBI and CGS-STI risk relationships observed in actual crashes while controlling for covariates. Receiver operating characteristic curves determined the sensitivity and specificity of HIC and CGS as predictors of TBI and STI, respectively. Estimates of the likelihood of TBI and/or STI (in actual MVCs) were compared with the expected probabilities of TBI and STI (determined by crash test analysis), as they relate to NCAP ratings.

RESULTS

The crash tests overestimate TBI likelihood at HIC scores >800 and underestimate it at scores <500. STI likelihood is overestimated when CGS exceeds 40 g. Receiver operating characteristic curves demonstrated poor sensitivity and specificity of HIC and CGS in predicting injury. The actual MVC injury probability estimates did not vary between vehicles of different NCAP rating.

CONCLUSION

HIC and CGS are poor predictors of TBI and STI in actual MVCs. The NCAP five-star rating system is unable to differentiate vehicles of varying crashworthiness in actual MVCs. More sensitive parameters need to be developed and incorporated into vehicle crash safety testing to provide consumers and automotive manufacturers with useful tools with which to measure vehicle safety.