Examining the relationship between heavy alcohol use and assaults: With adjustment for the effects of unmeasured confounders

Abstract

Background. Experimental studies suggest that alcohol can lead to aggression in laboratory settings; however, it is impossible to test the causal relationship between alcohol use and real-life violence among humans in randomized clinical trials. Objectives. (i) To examine the relationship between heavy alcohol use and assaults in a population based study; (ii) to demonstrate the proxy outcome method, as a means of controlling the effects of unknown/unmeasured confounders in observational studies. Methods. This study used data collected from three waves of the National Survey on Drug Use and Health (NSDUH). The effects of heavy alcohol use on assault were measured using multivariable logistic regressions in conjunction with the proxy outcome method. Results. Application of the proxy outcome method indicated that effect sizes of heavy alcohol use on the risk of assault were overestimated in the standard models. After adjusting for the effects of unknown/unmeasured confounders, the risk of assault remained 43% and 63% higher P < 0.05 among participants who consumed 5+ drinks/day for 5-8 days/month and 9-30 days/month, respectively. Conclusions. Even after adjustment for unknown/unmeasured confounders the association between heavy alcohol use and risk of violence remained significant. These findings support the hypothesis that heavy alcohol use can cause violence.