Assessing uncertainty in forensic alcohol calculations: The case for using percentile ranges

Forensic alcohol calculations are used to determine the amount of alcohol consumed based on (a) a relevant biological sample (i.e., breath or blood), (b) the blood alcohol concentration after a certain period following the consumption of a known quantity of alcohol, or (c) the extrapolation of the analytical result of an alcohol analysis to a legally relevant time. Current best practice guidelines recommend the use of percentile ranges for the parameters that contribute the most uncertainty to the results derived from forensic alcohol calculations. These parameters are volume of distribution (V_d) and elimination rate of alcohol (β). To date, there has been a lack of a transparent and defensible framework for using percentile ranges in forensic alcohol calculations. This work uses simulated data and statistical principles to demonstrate both why and what percentile ranges should be used in forensic alcohol calculations. We recommend that the percentile range used should depend on the case type (civil vs. criminal) and the amount of data available. The percentile ranges should be determined from an appropriate population using bootstrapping to estimate the relevant percentile with a two-tailed 95% confidence interval: for civil cases (25th and 75th percentiles; 50% range) and for criminal cases (0.5th and 99.5th percentiles; 99% range or 2.5th and 97.5th percentiles; 95% range) depending on the amount of data available. Using relevant percentile ranges produces results that better reflect the underlying uncertainty in a specific case than using a population average when carrying out forensic alcohol calculations.