Influence of question type, reference period and sensitivity on item reliability in the National Survey on Drug Use and Health

The National Survey on Drug Use and Health (NSDUH) is an annual, cross-sectional survey of the civilian, non-institutionalized U.S. population aged 12 and older. Respondents (N=67,500 per year) are interviewed in households or group quarters, using laptop computers. The survey obtains information on past month, past year, and lifetime use or non-use of illicit drugs, alcohol and tobacco, as well as a variety of other health-related topics. Sensitive items are administered in ACASI mode; less sensitive information is captured in CAPI mode. A Reliability Study was conducted in 2006, in which approximately 3,100 NSDUH respondents were re-interviewed after a 5 to 15 day interval. The study enabled assessment of response consistency for a wide range of survey item types within several content domains. Three item characteristics are of interest in this study: answer type (fact vs. attitude), sensitivity (as inferred by legal penalties associated with behavior being probed), and reference period (30 days, 12 months, lifetime). This paper will describe results obtained on several items that vary in these characteristics. It is hypothesized that increased sensitivity will be associated with decreased reliability, and factual questions will be more reliable than attitude questions. Longer reference periods are expected to have no effect on reliability of non-sensitive items, but should be positively associated with reliability of sensitive items. These results may inform researchers about possible differences in reporting errors across question types that vary in the cognitive, social and contextual factors that influence consistent and truthful reporting.