Alcohol-attributable fraction for injury in the U.S. general population: data from the 2005 National Alcohol Survey

Although studies of patients seen in emergency department (ED) settings have documented a strong association of alcohol with injury, such patients are not necessarily representative of the larger population, and less is known of alcohol's association with risk of injury in patient samples outside the ED.

Drinking before injury was analyzed in the 2005 National Alcohol Survey among the 1,149 respondents (18.5%, weighted) who reported an injury during the past year; analysis was by injury treatment type (ED-treated, 29.2%; other-treated, 47.8%; and nontreated, 22.9%).

Based on case-crossover analysis, the relative risk of injury from drinking was 1.85 (p < .01) for those with an ED-treated injury, 1.42 (ns) for those with an other-treated injury, and 1.43 (ns) for those with a nontreated injury. Alcohol-attributable fractions based on these relative risk estimates were 2.96% for an ED-treated injury, 1.59% for an other-treated injury, and 1.89% for a nontreated injury. Comparative attributable fractions based on the person's causal attribution of injury to his or her drinking were 3.06%, 1.61%, and 1.47%, respectively. Although these attributable fractions based on case-crossover analysis and subjective evaluation of causal attribution were not greatly different, all estimates were considerably smaller than those found in studies of ED patients.

The data suggest that alcohol plays a larger role in those injuries for which treatment is sought in EDs, and this may be related to the severity of the injury. Additional studies of alcohol and injury in general populations that take into account the intensity of exposure to alcohol before the event, as well as recall bias by eliciting data on the proximity of the event to the time of the respondent interview, are necessary for determining unbiased estimates of the attributable fraction of alcohol in injury morbidity.