Abstract
This study examines the differential item functioning (DIF) on both the standard 20-item scale and 10-item and 12-item subsets of the Center for Epidemiological Studies Depression (CES-D) inventory.
Two elderly groups in the United States are compared: one composed of Hispanics and the other of non-Hispanic Whites. We also compare item use among the elderly Hispanics over time. Survey data was analyzed from participants (N = 4499) in the Established Populations for Epidemiologic Studies of the Elderly (EPESE) with complete data on the CES-D. Partial correlation analysis was applied to test hypotheses regarding the DIF on the basis of ethnic background and among the elderly Hispanics across time points.
Hispanics appear to express positive affect differently than non-Hispanic Whites. On the 20-item scale, Hispanics under responded to items measuring positive affect. Item bias was also found for one positive affect question in the 10-item subset. Among the Hispanic sub-population, we also observed item response bias for the positive affect items across time. A 12-item subset tested for DIF displays an acceptable amount of item bias across ethnic groups and across time for the Hispanic sample.
Attempts to make substantive comparisons about the relative prevalence of depression symptoms in elderly Hispanic and non-Hispanic sub-populations appears to be limited by the lack of measurement invariance in the full CES-D. Comparisons across time for elderly Hispanics also appear problematic.