Reliability and validity of life-course alcohol consumption measures: The 2005 National Alcohol Survey Follow Up

Background We undertook test-retest analyses of heavy drinking measures and examined their predictive validity in relation to current alcohol use disorders (AUDs). A 31.4% subset of the N11 US telephone adult sample was followed (N11T; 962 males, 1220 females; mean delay 2.7 years) by phone (n=1449) or postal pre-paid mailer (n=734); 2078 provided lifetime drinking both times. Results Ever drinking alcohol was reported consistently by 94.7% of N11 current drinkers, 85.5% of ex-drinkers and 74.4% of life-abstainers (overall status consistency 93.5%). Analyses of prior heavy drinking in teens, 20s, and 30s showed cumulative number of heavy days correlated (Spearman's ρ=.69, p<.001, n=1471), with some variations seen by age and ethnicity but not gender. Decade-specific categorical agreement rates were 84 to 86% with modest weighted kappas (0.52 - 0.55). Test-retest correlation for lifetime maximum/any day was 0.64 (p<.001). As regards validity, among drinkers aged 41+, in logistic regressions predicting Alcohol Use Disorder, controlling for age, gender ethnicity and current heavy (5+) drinking, lifetime cumulative days 5+ (teens through 30s) was significant (p=.02). In separate models, there were similar relationships seen for alcohol dependence and consequences. Relationships with age of drinking onset suggested that early initiation may help explain both a trajectory of lifetime heavy drinking over decades as well as subsequent AUD diagnosis. Conclusions Assessing earlier heavy drinking is challenging but may be of value in cross-sectional and baseline panel surveys adding information beyond current drinking patterns. Results suggest cumulative lifetime drinking may be important for developing AUDs.