Predicting the Impact of Alcohol Taxation Increases on Mortality-A Comparison of Different Estimation Techniques

Aims To examine how standard analytical approaches to model mortality outcomes of alcohol use compare to the true results using the impact of the March 2017 alcohol taxation increase in Lithuania on all-cause mortality as an example. Methods Four methodologies were used: two direct methodologies: (a) interrupted time-series on mortality and (b) comparing predictions based on time-series modeling with the real number of deaths for the year following the implementation of the tax increase; and two indirect methodologies: (c) combining a regression-based estimate for the impact of taxation on alcohol consumption with attributable-fraction methodology and (d) using price elasticities from meta-analyses to estimate the impact on alcohol consumption before applying attributable-fraction methodology. Results and Conclusions While all methodologies estimated reductions in all-cause mortality, especially for men, there was substantial variability in the level of mortality reductions predicted. The indirect methodologies had lower predictions as the meta-analyses on elasticities and risk relations seem to underestimate the true values for Lithuania. Directly estimated effects of taxation based on the actual mortalities seem to best represent the true reductions in alcohol-attributable mortality. A significant increase in alcohol excise taxation had a marked impact on all-cause mortality in Lithuania. Four different analytic approaches were used to estimate the impact of a recent alcohol excise tax increase in Lithuania on all-cause mortality. While all resulted in decreases of mortality, the true effect was best estimated via a comparison between predictions derived from time-series methodology compared with actual number of deaths. The indirect analytic approaches based on alcohol-attributable fraction methodology seem to underestimate the true effect.