Computational Conception of a Reduced Quality of Vision Questionnaire

Tuesday, April 21, 2015: 1:31 PM
Room 5A (San Diego Convention Center)
Mohammed Ziaei, MBChB (Hons), FRCOphth
Bruce D. Allan, MD
Gary S. Rubin, PhD

Purpose
To assess the possibility of shortening the Quality of Vision (QoV) questionnaire.

Methods
Data from a prospective multicentre randomized controlled clinical trial comparing two multifocal intraocular lenses, Acri.LISA 366D and the Acrysof Restor SN6AD1 was analyzed. Principal components analysis (PCA) was performed for each of the Frequency, Severity and Bothersome subscales of the QoV questionnaire. We also explored the correlation and interchangeability of the three QoV questionnaire subscales using multivariate correlation and Bland-Altman limits of agreement (LoA) analysis respectively. Principal component analysis of Rasch model residuals (PCAR) was finally performed to assess for the presence of subdimensions.

Results
In total 188 completed questionnaires were included in the final analysis. Multivariate correlation revealed that all three subscales are highly correlated, especially severity and frequency (r = 0.89). The severity component was found to contribute 89.5% of the total variance in the dataset using PCA analysis (eigenvalue = 2.68). PCAR analysis suggested that the questionnaire may be multifactorial but the dimensions do not align with existing subscales.

Conclusion
The significant correlation level between the three subscales as well as the high contribution of the severity subscale to the overall dataset variance indicates that a reduced version of the questionnaire focused on the severity component (QoV10) can be used as a reliable representative of the QoV questionnaire. The QoV10 can be employed to achieve a comprehensive assessment of visual quality without substantial loss of information. Further evaluation of the QoV10 with different target populations is warranted.