Utility of different dimensional properties of drinking practices to predict stable low-risk drinking outcomes of natural recovery attempts

•Four drinking practice variables were used to predict natural recovery outcomes.•Less variation in pre-recovery quantities consumed predicted moderation outcomes.•Mean drinking frequency and quantities per drinking day did not predict outcomes.•Assessing variability in quantities consumed may aid drinking goal choice.•This dimension adds to established predictors of low-risk drinking outcomes. Functional measures indicating lower drinking problem severity predict stable low-risk drinking outcomes of recovery attempts, but findings for drinking practices are mixed. Because low-risk drinking outcomes are more common in natural than treatment-assisted recovery attempts, five studies of natural recovery attempts were integrated. Multiple dimensions of drinking practices during the year before recovery initiation were evaluated as predictors of post-recovery drinking (continuous abstinence, stable low-risk drinking, or unstable recovery involving relapse). Community-dwelling problem drinkers (N = 616, 68% male, mean age = 46.5 years) were enrolled soon after stopping alcohol misuse and followed prospectively for one year. A Timeline Followback interview assessed daily drinking during the year before recovery initiation and yielded four dimensions for analysis: frequency of heavy drinking days (4+/5+ drinks for females/males), mean ethanol consumption per drinking day, variability in days between heavy drinking days, and variability in ethanol consumed per drinking day. Multinomial logistic regression models showed that variability in ethanol consumed per drinking day was the sole significant predictor of 1-year outcomes when all dimensions were evaluated together. The low-risk drinker group showed less fluctuation in quantities consumed on pre-recovery drinking days compared to the groups that abstained or relapsed (ps