Abstract
This report studies the number of drink/drive accidents that escape detection because the driver involved is not breath-tested, and revises estimates for the number of non-fatal drink/drive accidents. A statistical model is described which estimates the number of drink/drive accidents missed because of inappropriate decisions not to require a breath test. The report presents results which were achieved when the model was fitted to data from individual police forces for 1985-91. The report also attempts to estimate the number of drink/drive accidents which are undetected because the circumstances of the accident make a test impracticable.