There are many challenges in the population comparisons described here. Of course, suicide decedents may differ from survey respondents in innumerable ways. For example, decedent alcohol data were derived from blood tests, whereas respondent information came from self-reports. Numerous so-called coverage studies [29] have found that self-reported alcohol consumption accounts for (at most) about half the alcohol sold (based on tax data). Also, survey respondents were volunteers, who may well have changed their usual drinking behaviors in the day(s) prior to interview. Given the comparison challenges, it is reasonable to inflate the drinking probabilities estimated from the survey data.
