Comparison of breath-alcohol screening test results with venous blood alcohol concentration in suspected drunken drivers

Abstract Hand-held electronic breath-alcohol analyzers are widely used by police authorities in their efforts to detect drunken drivers and to improve road-traffic safety. Over a three month period, the results of roadside breath-alcohol tests of drivers apprehended in Finland were compared with venous blood alcohol concentration (BAC). The mean (median) time between sampling blood and breath was 0.71 h (0.58 h) with a range from 0 to 6 h. Some hand-held instruments gave results as the concentration of alcohol in breath and were converted into BAC assuming a blood–breath alcohol ratio (BBR) of 2260. The mean venous BAC (1.82 g/kg) in traffic offenders was higher than the result predicted by the hand-held breath analyzers (1.72 g/kg). In 1875 roadside tests, the relationship between venous BAC ( x ) and BrAC ( y ) was defined by the regression equation y = 0.18 + 0.85 x . The coefficients show both a constant bias ( y -intercept 0.18 g/kg) and a proportional bias (slope = 0.85). The residual standard deviation (SD), an indicator of random variation, was ±0.40 g/kg. After BAC results were corrected for the time elapsed between sampling blood and breath, the y -intercept decreased to 0.10 g/kg and 0.004 g/kg, respectively, when low (0.1 g/kg/h) and high (0.25 g/kg/h) rates of alcohol elimination were used. The proportional bias of 0.85 shows that the breath-alcohol test result reads lower than the actual BAC by 15% on average. This suggests that the BBR of 2260 used for calibration should be increased by about 15% to give closer agreement between BAC and BrAC. Because of the large random variation (SD ± 0.40 g/kg), there is considerable uncertainty if and when results from the roadside screening test are used to estimate venous BAC. The roadside breath-alcohol screening instruments worked well for the purpose of selecting drivers above the statutory limit of 0.50 g/kg.