PURPOSE: Differential item functioning (DIF) is said to exist in an item if a subject's response to the item is affected by other aspects than that which the test is intended to assess. DIF might affect the validity of a test. The aim of this study was thus to examine whether any of the items in the Epworth Sleepiness Scale (ESS) exhibits DIF regarding age or gender, and if so, to which degree. METHODS: Using previously collected cross-sectional ESS data from 1,168 subjects with different clinical characteristics (61% males, mean age 67.8 year (SD 12.2 year)), ordinal regression as well as Rasch-based DIF analyses were performed. RESULTS: Concerning age, both DIF analyses showed DIF for age in items 3 (inactive in a public place), 4 (passenger in a car), and 8 (in a car that has stopped in traffic). The Rasch model also showed DIF for gender in item 3. The DIF magnitudes as judged by McFadden pseudo-R (2) changes were, however, only minor. CONCLUSIONS: ESS has small but reproducible DIF for age in items 3, 4, and 8. The detected DIF might be worth to consider in large-sample studies, although it probably has no effect on an individual basis.
PURPOSE: Differential item functioning (DIF) is said to exist in an item if a subject's response to the item is affected by other aspects than that which the test is intended to assess. DIF might affect the validity of a test. The aim of this study was thus to examine whether any of the items in the Epworth Sleepiness Scale (ESS) exhibits DIF regarding age or gender, and if so, to which degree. METHODS: Using previously collected cross-sectional ESS data from 1,168 subjects with different clinical characteristics (61% males, mean age 67.8 year (SD 12.2 year)), ordinal regression as well as Rasch-based DIF analyses were performed. RESULTS: Concerning age, both DIF analyses showed DIF for age in items 3 (inactive in a public place), 4 (passenger in a car), and 8 (in a car that has stopped in traffic). The Rasch model also showed DIF for gender in item 3. The DIF magnitudes as judged by McFadden pseudo-R (2) changes were, however, only minor. CONCLUSIONS:ESS has small but reproducible DIF for age in items 3, 4, and 8. The detected DIF might be worth to consider in large-sample studies, although it probably has no effect on an individual basis.
Authors: Anders Broström; Anna Strömberg; Jan Mårtensson; Martin Ulander; Lena Harder; Eva Svanborg Journal: J Sleep Res Date: 2007-12 Impact factor: 3.981