Abstract
Purpose :
The visual analog scale (VAS) and categorical rating scale (CRS) are commonly used rating scales in patient reported outcome (PRO) instruments. However, the effect of these rating scales on a PRO instrument’s psychometric properties is not clear. Utilizing Rasch analysis on the activity limitation scale of the Keratoconus Outcome Research Questionnaire (KORQ), we explored the hypothesis that the KORQ with VRS would demonstrate better psychometric properties and provide less noisy measurement than with VAS.
Methods :
The KORQ is a PRO measure for use in keratoconus; it has activity limitation and symptoms scales. The 15-item KORQ activity limitation scale with two different response scales (VAS and a discrete 4-response CRS), were completed by self-administration by people with keratoconus. The VAS was presented as a single line of 10cm anchored ends with 0 (Not at all) and 10 (so much, I can’t do it) for each item. For Rasch analysis VAS was transformed into a discrete scale by dividing it into 11 categories (1cm=1 category). Rasch analysis based psychometric properties were compared between the two versions the KORQ (precision as person separation, unidimensionality as variance explained by the measure). Rasch analysis was also used to optimize rating scale functioning when disordered thresholds were observed.
Results :
A total of 118 (mean age ±SD, 46.4±0.4 yrs; male, 53.6%) and 169 (45.4±14.7 yrs, male, 43.6%) completed the KORQ with VAS and CRS respectively. Both scales demonstrated very high measurement precision (3.24 for VAS and 3.19 for CRS). However, the VAS rating scale was disordered (6 out of 11 categories) and had 3 misfitting items. Conversely, the KORQ with CRS had ordered categories and no misfitting items. For the disordered VAS, ordering was achieved only after collapsing it into 4-response categories. This had significantly reduced precision (2.92). In comparison to the KORQ with CRS, the KORQ with VAS had a lower variance explained by the measure (60% vs 56.4%), misfitting items (2 vs 0), poor targeting (-1.31 vs -0.75 logit) and significantly reduced measurement range (4.12 vs 12.13 logit).
Conclusions :
The KORQ demonstrated superior psychometric properties when measured using a CRS than with a VAS. This demonstrates the validity a reliability of the KORQ questionnaire. These results illustrate the advantage of CRS scales. Visual analog scales should be avoided in PRO measurement.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.