Purchase this article with an account.
Robert W Massof, Kyoko Fujiwara, Theresa M Smith; Effects of item filtering on estimates of patient reported outcome measures for low vision rehabilitation. Invest. Ophthalmol. Vis. Sci. 2018;59(9):3409.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Low vision rehabilitation (LVR) outcome measures usually are estimated from difficulty ratings of selected activities (items). Patients differ in functional ability (P); items differ in the amount of functional ability required to perform the described activity (I). Theoretically, the difficulty of each activity for a patient scales inversely with the difference between P and I – called functional reserve (F=P-I). The aim of LVR is to increase F for those activities that are indentified in the plan of care (POC) as important to the patient and difficult or impossible to perform. An increase in F can be achieved by increasing P, decreasing I, or both. Most LVR is directed at reducing I for each activity goal in the POC. Thus, the choice of items could significantly impact measures of LVR outcomes. This study evaluates the effect of item selection on LVR outcome measures.
The Activity Inventory (AI) was administered to 100 low vision patients by telephone before and after LVR. LVR was provided by OTs who were shown each patient’s AI goals and the goals’ subsidiary tasks that were rated important and at least “slightly difficult. Rasch analysis was performed on the patient difficulty ratings of AI goals, and on functional domain subsets of AI tasks, with item and category-threshold measures anchored to previously calibrated values. Both Pre- and Post-LVR person measures were estimated using all rated items (unfiltered) and again using only items rated at least “somewhat difficult” at Pre-LVR (filtered).
LVR outcomes are expressed as changes in overall visual ability, estimated from difficulty ratings of AI goals, and as changes in different domains of functional ability, estimated from difficulty ratings of different subsets of AI tasks. Mean change scores were normalized to the respective change score SD (Cohen’s d). Overall visual ability had the largest item filtering effect: d=2.56 filtered vs 0.85 unfiltered. The effect of item filtering on outcome measures was smallest for reading: d=0.55 filtered vs 0.45 unfiltered. As shown in Figure 1, other functional domains also exhibited significant effects of filtering.
Filtering out items rated not difficult at Pre-LVR increases effect size. Items with no room for improvement dilute estimates of the average change in F. Items should be consistent with the POC.
This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.
This PDF is available to Subscribers Only