Abstract
Purpose:
Item banking provides the opportunity to explore item grouping to form latent traits within a core item set. This is not possible with existing vision-related quality of life instruments that have limited number of items. We aimed to explore whether item banking enables important latent traits, such as driving, to form stand-alone measures.
Methods:
The 88-item activity limitation (AL) domain of the glaucoma module of the Eye-tem Bank was interviewer administered to patients with glaucoma. Rasch analysis was used to calibrate all AL items on the same interval level scale and also to test its metric properties (namely: precision, principal component of analysis of residuals (PCA) to test dimensionality, item fit statistics and differential item function (DIF) to test item bias). Based on Rasch dimensionality metrics, the AL domain was separated into subscales. These subscales were subjected to separate Rasch analyses to test whether they could form stand-alone measures. Independence of these measures was tested with Pearson’s correlation co-efficient and Bland and Altman (B&A) Limit of Agreement (LOA).
Results:
The AL domain was completed by 293 patients (male, 55%; median age, 70 yrs; range 20-91 yrs). The domain demonstrated excellent precision (3.12). However, the PCA indicated that the AL domain had other dimensions which were driving, reading and luminance-related activities. The separate Rasch analyses revealed that the remaining 62 AL items, driving and luminance subscales were unidimensional, had excellent precision of 4.25, 2.94 and 2.22 respectively, with acceptable fit statistics and free of item bias (no DIF by age and gender). The reading subscale showed poor precision (1.66) so was discarded. The luminance subscale had a high correlation (r=0.78, p<0.0001) and excellent agreement (mean bias, 0.2 logit; 95% LOA, -2.2 to 3.3 logit) with the AL subscale. However, the driving subscale had a weaker correlation (r=0.48, p<0.0001) and poor agreement (mean bias, 1.1 logit; 95% LOA, -4.8 to 7.0 logit) with the AL subscale.
Conclusions:
These findings indicate that driving items were perceived and responded differently from the AL items, but the reading and luminance items were not. The driving subscale has excellent Rasch based metrics. Therefore, item banking enables stand-alone measurement of driving ability.