Does a blood alcohol level of 0.15 or more identify accurately problem drinkers in a drink-driver population?

To determine whether a blood alcohol concentration (BAC) of 0.15 or more identifies accurately problem drinking in an apprehended drinking driver.

Cross-sectional analytic study of 86 drink drivers ("diagnostic test study").

St Vincent's Hospital, Melbourne, drink-driver education programme.

Men aged between 18-25 years, applying for relicensing after disqualification, who were participants at 10 consecutive drink-driver education courses conducted in early 1987.

Self-reports of the amount of alcohol consumed and the amount of money spent on alcohol per week, the standard Michigan Alcoholism Screening Test (MAST) score and the blood alcohol level at apprehension. The first three measures were used as reference standards for alcohol-related problems, against which the blood alcohol level was compared.

Scatter plots of BAC against amount consumed, amount spent and the MAST scores showed that a large proportion of the heaviest drinkers were not identified by the BAC. Log-transformed correlation coefficients were: for amount spent v. amount consumed, r = 0.73; for MAST score v. amount consumed, r = 0.52, and for MAST score v. amount spent r = 0.53. Correlating the BAC with the log of the MAST score gave r = 0.21, BAC with log of the amount consumed, r = 0.20, and BAC with log of the amount spent, r = 0.15. Comparing a BAC of 0.15 or more with a MAST score of greater than or equal to 5, we obtained a sensitivity of 0.36, a specificity of 0.76, a positive predictive value of 0.86, and a negative predictive value of 0.23. The prevalence of heavy drinking as measured by the MAST score was 0.80.

A blood alcohol concentration of 0.15 or more identifies only a third of the problem drinkers in this sample of drink drivers. Although it selected a group of drinkers with alcohol-related problems, it missed most of the problem drinkers and did not reliably identify the heaviest drinkers.