Performance-based vehicle ignition-interlock devices (IID) have the potential to directly detect deficits in cognitive ability, a key factor in driving safety, and are expected to play a pivotal role in traffic safety. However, as the absence of a standard method for evaluating performance tests is an obstacle to their further development, this paper proposes a new evaluation method. Regression analysis is used to calculate a border point (BP), where an upper 95% confidential interval (CI) for performance would not exceed the mean performance at a blood alcohol concentration (BAC) of 0 mg/ml, and a critical point (CP), where an upper 95% CI for performance would not exceed the lower 95% CI for performance at 0 mg/ml BAC. Compared to conventional analyses, it is found that the new method exhibits superior features, such as direct comparability not only across tests but across experiments and accurate capability estimations from degraded data set. Thus, we argue that the new method of evaluating performance tests in terms of BP and CP points is suitable for use as a standard in future IID studies.
View full abstract