Abstract
Abstract: :
Purpose: To apply Item Response Theory (IRT) techniques to evaluate the psychometric performance of the NEI-VFQ-25 in a population based sample of Latinos. This model explicitly assesses the performance of the rating scale. Methods: The LALES is a population-based prevalence study to assess the presence of eye disease and quality of life (QOL) in Latinos, aged 40 and older. The NEI-VFQ-25 was used to assess self-reported visual function. NEI-VFQ-25 scores were calculated using standard algorithms, which assumed that ordered response choices were equidistant. For example in a scale including five response choices, ranging from 'none of the time' to 'all of the time', the distance between 'none of the time' and 'a little of the time' is assumed to be the same as the distance between 'most of the time' and 'all of the time'. For this analysis, the total pool of 25 questions was assumed to measure a single underlying trait-visiion related quality of life. The assumptions of unidimensionality and independence were tested through factor analysis. Rasch rating scale model methods were used to value the ordinal consistency of response choices. This model tests the assumption that all categories of response for a particular item are ordered. Results: 1938 participants that completed the NEI-VFQ-25 were included in this analysis. Sixty-nine percent of the participants chose to complete the questionnaire in Spanish. Factor analysis revealed that there was one dominant factor (eigenvalue 9.3) which explained 37% of the variance in the NEI-VFQ-25 composite scores. IRT revealed that items with six response choices did not give consistent results for different participants and for different questions. This pattern was evident when the entire sample was assessed as well as for participants who completed the instrument in either English or Spanish. Conclusion: We found that six response choices produced inconsistencies in assessing vision related quality of life in Latinos. Our findings suggest that fewer response choices per question may be preferable.
Keywords: 353 clinical (human) or epidemiologic studies: outcomes/complications • 350 clinical (human) or epidemiologic studies: biostatistics/epidemiology methodology