Predicting repeat DUI offenses with the alcohol interlock recorder

The aim of this report has been to use information contained in the alcohol ignition interlock recorder to determine whether systematic analysis of it can be used to predict which DUI offenders will recidivate during the first 2 years after the interlock is removed. The interlock record was accumulated during a 4-year intervention study in Alberta, Canada. Data from more than 5.5 million breath tests collected during interlock use were analyzed retrospectively after allowing repeat DUI offenses to accumulate for up to 2 years post-interlock. The rate of interlock warns at low BAC (0.02–0.04%) and fails at higher BAC (>0.04%) were found to be predictive of later repeat DUI. The interlock record was used along with selected driver record variables and questionnaire data to identify predictor sets. CHAID segmentation analysis was used to identify combinations of predictor variables; these were joined with sensitivity analysis to compare different predictor combinations. Several variables, but primarily more prior DUIs and more interlock warns and fails logged during the first 5 months of interlock usage predict greater than 60% of repeat DUI with a false positive rate of less than 10%.