The interview mode effect on the Center for Epidemiological Studies Depression (CES-D) scale an item response theory analysis

Background: Evidence of a mode effect has raised concerns about the comparability and validity of self- versus interviewer-administered versions of the Center for Epidemiological Studies Depression (CES-D) scale. Response anonymity has been proposed to explain this effect. However, the factors that contribute to this mode effect are not well understood. We used item response theory (IRT) to examine the nature of the CES-D mode effect. Methods: A sample of depressed primary care patients from the Partners-in-Care Study were randomized to receive either a phone interview (N = 139) or a mail survey (N = 139) of the CES-D. We used likelihood ratio tests to identify differentially functioning items in the 2 groups. Category response curves are used to describe these effects. Results: Twelve items manifested differential functioning. Category response curves consistently indicate that phone respondents had a lower probability of endorsing the third of 4 response categories than mail respondents, suggesting a possible cognitive effect. Conclusion: Although response anonymity could be important in mode effects observed in surveys of sensitive topics, cognitive factors appear more important to the mode effect in the CES-D.