Abstract
OBJECTIVE: Underreporting of substance use is a frequent concern about studies based on self-report, but few robust studies have examined the agreement between different methods for capturing self-reported substance use. The current study therefore used repeated measures to compare self-reported substance use using (a) clinician interviewers and (b) self-administered computerized surveys in a sample that included both inpatients and community residents. METHOD: Adults age 18 years and older with problematic substance use were recruited from the community or an inpatient addiction treatment facility. At baseline (N = 588), 3-month (n = 469), and 6-month (n = 476) interviews, participants were asked whether they used alcohol, cannabis, cocaine, heroin, and prescription painkillers by two methods: semi-structured, clinician-administered interview, and computerized self-administered questionnaire. Agreement between these two methods was investigated using Cohen's kappa coefficient. Multivariable logistic regression assessed differences in the odds of discordance between the two measures by recruitment source, gender, age, race/ethnicity, employment status, marital status, and level of education. RESULTS: There was moderate-to-strong agreement between clinician-administered and self-administered surveys for alcohol (kappa = .70-.88), cannabis (kappa = .87-.92), cocaine (kappa = .81-.89), and heroin (kappa = .90-.92). However, there was only weak-to-moderate agreement for nonmedical use of prescription painkillers (kappa = .55-.71), with the self-administered questionnaire capturing a higher prevalence of use (percent difference = 2.4%). CONCLUSIONS: Clinician interviewers and self-administered surveys were shown to capture similar rates of self-reported use of alcohol, cannabis, cocaine, and heroin. Surveys assessing nonmedical prescription opioid use may benefit from using self-administered questionnaires.