Reducing bias in telephone survey estimates of the prevalence of drug use: a randomized trial of telephone audio-CASI.

Aim: To assess the impact of telephone audio computer-assisted self-interviewing (T-ACASI) on reporting of alcohol use, alcohol problems and illicit drug use in telephone surveys of the general population. Prior research suggests that illicit drug use is underreported in traditional, interviewer-administered, telephone surveys.

Design: Randomized experiment embedded in telephone survey of probability samples of populations of USA and Baltimore, MD. Survey respondents were randomly assigned to be interviewed either by human telephone interviewers or by T-ACASI after household screening, recruitment, and informed consent procedures were completed.

Setting: Respondents were interviewed by telephone in their homes.

Participants: Probability samples of 1543 English-speaking adults ages 18-45 residing in telephone-accessible households in USA and 744 similarly defined adults residing in Baltimore, MD, USA.

Measurements: Nine questions on alcohol, marijuana, cocaine, and injection drug use adapted from 1994 NHSDA and four CAGE questions on alcohol problems. Crude odds ratios and odds ratios controlling for demographic factors calculated to test for differences between responses obtained by T-ACASI and human interviewers.

Findings: T-ACASI had mixed effects on reporting of alcohol use, but it did increase reporting of one of four CAGE alcohol problems: feeling guilty about drinking (23.0% in T-ACASI vs. 17.6% in T-IAQ, OR = 1.4, P < 0.01). T-ACASI also obtained significantly more frequent reporting of marijuana, cocaine, and injection drug use. The impact of T-ACASI was most pronounced for reporting of recent use of 'harder' drugs. Thus T-ACASI respondents were more likely to report marijuana use in the past month (10.0% vs. 5.7%, crude OR = 1.9, P < 0.001), cocaine use in the past month (2.1% vs. 0.7%, crude 3.2, P < 0.001) and injection drug use in the past five years (1.6% vs. 0.3%, crude OR = 4.8, P < 0.01).

Conclusions: Telephone survey respondents were more likely to report illicit drug use and one alcohol problem when interviewed by T-ACASI rather than by human telephone interviews.