A Comparison of Two Case-Crossover Methods for Studying the Dose-Response Relationship between Alcohol and Injury

Abstract

This study compares dose-response injury risk estimates for two control periods defined as the same 6-hour period the week prior and the set of all non-sleeping 6-hour periods over the past year. Dose-response injury risk estimates for the multiple match controls are generated via the application of a maximum-likelihood approach. Injury risk associated with any (i.e., 1 drink or more) drinking 6 hours prior to injury was similar for the two control choices (last week and usual frequency). For 1–4 drinks, risk estimates were similar across control period definitions; for 5+ drinks, risk using the week prior as the control was nearly double that using the past 12 months as the control. Although studies with smaller n-sizes may benefit from the increase in precision from the use of the multiple control periods, results indicate that heavy drinking injury risk estimates should be used with caution.