Estimating alcohol-attributable fractions for injuries based on data from emergency department and observational studies: a comparison of two methods.

Abstract

To compare the injury alcohol-attributable fractions (AAFs) estimated using emergency department (ED) data to AAFs estimated by combining population alcohol consumption data with corresponding relative risks (RRs). Comparative risk assessment. ED studies in 27 countries (n = 24 971). AAFs were estimated by means of an acute method using data on injury cases from 36 ED studies combined with odds ratios obtained from ED case-cross-over studies. Corresponding AAFs for injuries were estimated by combining population-level data on alcohol consumption obtained from the Global Information System on Alcohol and Health, with corresponding RRs obtained from a previous meta-analysis. ED-based injury AAF estimates ranged from 5% (Canada 2002 and the Czech Republic) to 40% (South Africa), with a mean AAF among all studies of 15.4% (18.9% for males and 8.4% for females). Population-based injury AAF estimates ranged from 21% (India) to 51% (Spain and the Czech Republic), with a mean AAF among all country-years of 36.8% (42.5% for males and 22.5% for females). The Pearson correlation coefficient for the two types of injury AAF estimates was 0.09 for the total, 0.06 for males and 0.32 for females. Two methods of estimating the injury alcohol-attributable fractions-emergency department data versus population method-produce widely differing results. Across 36 country-years, the mean AAF using the population method was 36.8%, more than twice as large as emergency department data-based acute estimates, which average 15.4%.