Discrete Time Survival Analysis (DTSA) was used to analyze the time to first remission and time to sustained remission from AUD. Hazard curves were developed as a function of measurement occasion of the outcome (e.g. no longer meeting criteria for an AUD during the prior 5 year interval), recording an outcome as present or absent at each follow-up (Muthén and Masyn, 2005). Follow-up epochs prior to the outcome were scored as zeros, the age of remission was recorded, and (because the individual was no longer at risk for the event) ages subsequent to an event were coded as missing (Trim et al., 2009; 2010). All continuous variables were first centered (average score subtracted from each raw score) to reduce multicollinearity, and binary variables were recorded as 0 and 1. Models were run in MPlus v6.12 (Muthén and Muthén, 1998–2011) with full information maximum likelihood (FIML) estimation under the assumption of data missing at random (MAR) with robust standard errors. To increase confidence in the values, automatically generated starting values with random perturbations (100 random sets of starting values with 10 full optimizations) were used for all models.
