This study has limitations. Because we restricted our outcome definition to alcohol-induced deaths, we excluded causes known to be associated with alcohol but not 100% attributable to alcohol (eg, traffic collisions, alcohol-associated cancers, infections and organ system diseases known to be associated with alcohol use). Thus, although our analysis of trends in alcohol-induced deaths provides an important indicator of the consequences of alcohol on population health, our findings substantially underestimate the full mortality burden.26,36 Our study also bears limitations associated with the underreporting of alcohol-attributable deaths on death certificates.37 Polednak37 observed that use of multiple-cause death records, as opposed to the underlying cause of death, may enhance surveillance of premature mortality because of chronic (although not acute) disease resulting from alcohol use; we encourage future research that considers the effects of this approach. Conversely, we recognize that deaths classified as alcohol-induced deaths may have been influenced by coexisting conditions otherwise affecting the liver (eg, hepatitis C infection or nonalcoholic fatty liver disease).38,39 Misclassification of race/ethnicity data within death certificates is also possible in our study but is minimal for groups
