TY - JOUR
T1 - Improving the quality and applicability of the Dutch scales of the communication profile for the hearing impaired using item response theory
AU - Mokkink, L.B.
AU - Knol, D.L.
AU - van Nispen, R.M.A.
AU - Kramer, S.E.
PY - 2010
Y1 - 2010
N2 - Purpose: The aim of this study was to improve the quality and applicability of the 6 Dutch scales of the Communication Profile for the Hearing Impaired (CPHI; Demorest & Erdman, 1986, 1987, 1988) using item response theory (IRT). IRT modeling can produce precise, valid, and relatively brief instruments, resulting in minimal response burden (Edelen & Reeve, 2007). Method: We investigated (a) whether items fit to the graded response model (GRM; Samejima, 1969) using Stone and Zhang's (2003) procedure to test goodness of fit and (b) whether items showed differential item functioning (DIF) with the IRT log-likelihood ratio approach for 5 group variables: gender, age, living arrangement, use of hearing aids, and degree of hearing impairment. We investigated the magnitude of DIF by calculating the maximum difference between expected item scores for subgroups of the group variables of items showing DIF. Results: In a cross-sectional study, 408 consecutive individuals with hearing impairment completed the Dutch CPHI scales. Twelve items were deleted because of misfit to GRMs, 14 items showed uniform DIF, and 2 showed nonuniform DIF. Five items showing DIF were deleted due to large magnitude of DIF. The short form of the Dutch CPHI scales contains 35 items. Conclusions: DIF analyses showed that the short form was relatively free of DIF with respect to the 5 group variables. Hence, there is no need to make adjustments for the calculation of subgroup scores. © American Speech-Language-Hearing Association.
AB - Purpose: The aim of this study was to improve the quality and applicability of the 6 Dutch scales of the Communication Profile for the Hearing Impaired (CPHI; Demorest & Erdman, 1986, 1987, 1988) using item response theory (IRT). IRT modeling can produce precise, valid, and relatively brief instruments, resulting in minimal response burden (Edelen & Reeve, 2007). Method: We investigated (a) whether items fit to the graded response model (GRM; Samejima, 1969) using Stone and Zhang's (2003) procedure to test goodness of fit and (b) whether items showed differential item functioning (DIF) with the IRT log-likelihood ratio approach for 5 group variables: gender, age, living arrangement, use of hearing aids, and degree of hearing impairment. We investigated the magnitude of DIF by calculating the maximum difference between expected item scores for subgroups of the group variables of items showing DIF. Results: In a cross-sectional study, 408 consecutive individuals with hearing impairment completed the Dutch CPHI scales. Twelve items were deleted because of misfit to GRMs, 14 items showed uniform DIF, and 2 showed nonuniform DIF. Five items showing DIF were deleted due to large magnitude of DIF. The short form of the Dutch CPHI scales contains 35 items. Conclusions: DIF analyses showed that the short form was relatively free of DIF with respect to the 5 group variables. Hence, there is no need to make adjustments for the calculation of subgroup scores. © American Speech-Language-Hearing Association.
U2 - 10.1044/1092-4388(2010/09-0035)
DO - 10.1044/1092-4388(2010/09-0035)
M3 - Article
SN - 1092-4388
VL - 53
SP - 556
EP - 571
JO - Journal of Speech, Language and Hearing Research
JF - Journal of Speech, Language and Hearing Research
IS - 3
ER -