Survival analysis was used to model different stages of alcohol progression using time-to-event data to examine the impact of psychosocial and psychiatric risk and protective factors on the rate of progression; that is, the duration of each of the three stages defined above. Specifically, three Cox proportional hazards regression models were constructed to predict (a) time to first drink, (b) time from first drink to onset of at-risk drinking, and (c) time from at-risk drinking to onset of AD. Models included psychosocial factors, notably R/S, and psychiatric risk factors as covariates influencing rate of progression through each stage. The dependent variable was the cumulative survival function; that is, the proportion of cases that had not reached a given alcohol transition, thus “surviving” to that point in time. With this method, all data up to the time of censorship (i.e., time of interview) were used in calculating hazard ratios. Basic univariate analyses were conducted with PASW Statistics Software, Release 18.0.0 (IBM-PASW, 2009). Advanced analyses were conducted with STATA, version 8.2 (StataCorp, 2001). Standard deviations of means and confidence intervals for advanced
