Health services research questions often require longitudinal designs and long-term follow-up of subjects. The follow-up portion of a design, particularly if face-to-face interviews are conducted, can represent significant added costs, both in time and dollars, to the overall cost of a study. Using alternative survey modes (telephone follow-up or tracking by mail) may substantially reduce study costs. However, data on reporting differences for sensitive information, such as drug use or criminal behavior, indicates that there are significant differences in responses by mode. Additional research is needed to fully examine the impact of survey mode on mental health outcomes. In situations where data from different sources are combined, it may be necessary to adjust the responses for mode effects. This proposal (responding to PA-94- 060 and PAR-99-140) aims to analyze data collected by the Depression PORT (Patient Outcomes Research Team) to test for discrepancies in responses between phone interviews and mail self-administered questionnaires. Preliminary analysis indicates that there are significant differences in responses to the two modes, depending on the content of the measure. We are interested in determining how phone responses differ from mail responses and whether these differences vary across subgroups of the depressed population. We will also identify appropriate methods of adjustment for mental health outcomes. This study has the following specific aims: 1) Analyze differences between phone and mail SAQ reports of mental health status (SF12), depression status and symptoms, quality of care and employment scales using summary statistics and standard tests; 2) Apply linear regression and item response theory to calibrate mail responses against phone responses and test these methods using the PORT/QID data; and 3) Further analyze mode effects on individual items within scales, sensitive to mode effects, using differential item functioning methods.
| Chan, Kitty S; Orlando, Maria; Ghosh-Dastidar, Bonnie et al. (2004) The interview mode effect on the Center for Epidemiological Studies Depression (CES-D) scale: an item response theory analysis. Med Care 42:281-9 |
