Breath Alcohol Test Program, Washington State Patrol, 811 East Roanoke, Seattle, WA 98102, USA.
J Breath Res. 2011 Mar;5(1):016004. doi: 10.1088/1752-7155/5/1/016004. Epub 2011 Feb 15.
Several factors contribute to the variability observed among repeated measurements of breath alcohol concentration. Identifying these factors and the magnitude of their contribution is the focus of this study. Large breath alcohol data sets consisting of duplicate test results from drivers arrested for driving while intoxicated were obtained from four jurisdictions: Sweden, Alabama, New Jersey and Washington State. The absolute difference between duplicate results were fitted to a multivariate linear regression model which included the following predictor variables: mean breath alcohol concentration, absolute exhalation time difference between repeated measurements, absolute exhalation volume difference, gender and age. In all data sets considered here, the breath alcohol concentration was the most statistically and practically significant predictor of absolute difference between the duplicate results. The next two most important predictors to enter models for all jurisdictions were exhalation volume difference and exhalation time difference. The maximum multivariate R² for any jurisdiction, however, was only 0.24, suggesting that other factors not considered here may be of importance. Two predictors over which the subject would have some influence included exhalation time and volume. When these were set at values expected to have maximum impact, the effect on duplicate test differences was very small, 0.008 g/210 L or less in all jurisdictions, indicating that subject manipulation of exhalation time and volume would have at most a very small systematic effect on estimated breath alcohol concentration. This study presents multivariate models useful for identifying the impact of five variables that may influence breath test variability.
若干因素导致了呼气酒精浓度重复测量中的可变性。本研究的重点是确定这些因素及其对呼气酒精浓度测量值的影响程度。从四个管辖区（瑞典、阿拉巴马州、新泽西州和华盛顿州）获得了包含因醉酒驾车被捕的驾驶员重复测试结果的大型呼气酒精数据。将重复测试结果之间的绝对差值拟合到一个多元线性回归模型中，该模型包括以下预测变量：平均呼气酒精浓度、重复测量之间的呼气时间绝对差值、呼气体积绝对差值、性别和年龄。在所有考虑的数据集，呼气酒精浓度是重复测试结果之间绝对差值的最具统计学和实际意义的预测因素。对所有管辖区的模型而言，下两个最重要的预测因素是呼气体积差值和呼气时间差值。然而，任何管辖区的最大多元 R²仅为 0.24，这表明此处未考虑的其他因素可能很重要。两个预测因素是呼气时间和呼气量，受测者对此有一定的影响。当将这两个因素设置为预期对测试结果有最大影响的值时，对重复测试差异的影响非常小，所有管辖区的影响均小于 0.008 g/210 L，这表明受测者对呼气时间和呼气量的操纵对估计的呼气酒精浓度的影响最多只有非常小的系统影响。本研究提出了多元模型，可用于识别可能影响呼气测试变异性的五个变量的影响。
