英语论文网

f baseline substance use and severity.The timeline followback24 was the primary measure of substance use and employment. We collected substance use data for each day from the date of the baseline to obtain a continuous record of use and to construct a dichotomous measure of abstinence (abstinent or not) for each month of the 24-month follow-up period. Monthly rates of absolute abstinence were the primary outcome measure. We confirmed selfreported substance use via collateral interviews and urine tests at 3, 9, and 15 months. Agreement between self-report and other methods ranged from 87.2% to 95.5%. Thus, findings suggest that self-report of substance use was valid. Along with substance use, we collected data on the number of days employed per month on the timeline followback. We constructed 3 monthly employment outcome variables: any employment within a month, number of days employed during the month, and employment of 19 days or more within the month (we considered this an indicator of full-time employment). We considered a month valid if there were at least 10 days of data available. For those partial months that we considered valid, we extrapolated missing data for the remainder of the month from the existing data. Fewer than 1% (79 of 10728 months observed) of data were imputed. Procedures Clinicians conducted baseline assessments for the substance-dependent sample, and research staff conducted all other interviews. A random number generator was used to determine assignment to a treatment group, and the assignments were sealed in an envelope. Envelopes were opened after the baseline assessment was completed to ensure that clinicians were blinded to group assignment while conducting the baseline assessment. All participants received in-person or telephone follow-up interviews 3, 9,15, and 24 months after baseline assessment. Details on follow-up retention rates through month15 have been published elsewhere.9 Of the original 302 substance-dependent participants who received a baseline assessment, 284 (94%) received a 24-month follow-up interview. Of the 150 participants in the comparison sample, 141 (94%) received a 24-month follow-up assessment. Analytic Plan We applied methods (e.g., generalized estimating equations) in our analysis appropriate for a longitudinal panel design.25 The analytic plan comprised 4 steps. In step 1, we tested whether abstinence outcomes associated with group assignment were maintained in months 16 through 24. With the same procedures as in our previous work,9 we modeled complete abstinence from all substances for each month of the follow-up period. In this model, we treated abstinence as a binary outcome, assuming a binomial distribution and logit link function. In step 2, we undertook a process of model building26 in which we examined the association between sets of sociodemographic, human capital, and substance use severity with employment. The following covariates had a marginal statistical association (P<.10) and were retained in subsequent models: having a high school diploma or the equivalent, recent work experience, and baseline drug use severity. In step 3, we examined whether group assignment was associated with employment outcomes. We selected days employed a priori as our outcome measure because it can be modeled as a count variable rather than a simple dichotomy, thus providing a more discriminating dependent measure. We determined a priori that if a group significantly predicted t