Abstract
Purpose:
Determine a clinically meaningful endpoint in visual ability in patients receiving low vision rehabilitation services
Methods:
The Activity Inventory (AI) was administered adaptively by telephone prior to low vision rehabilitation (LVR), and again at follow-up, 6 to 9 months after usual LVR to 468 patients at 28 collaborating clinical centers throughout the U.S. Rasch analysis was performed on the AI difficulty ratings (510 items in a bank calibrated to 3200 low vision patients) at baseline and follow up, with item measures anchored to calibrated baseline values. Items rated “not difficult” at baseline for each patient were excluded from both baseline and follow up person visual ability estimates because those activities have no room for improvement and would not be targeted in the LVR plan of care. Means and standard deviations were calculated for the respective distributions in changes in overall visual ability and changes in each of the 4 functional domains (reading, visual information processing, visual motor and mobility). To evaluate whether individual changes in visual ability measures were clinically meaningful, a minimum clinically important difference (MCID) was calculated for each measure from the ratio of the change in visual ability to 1.95 * standard error of the estimate.
Results:
With item filtering and consideration to standard error estimates, nearly half (47%) of patients exceeded a MCID endpoint in overall visual ability. The prevalence rates of patients exceeding a MCID in functional domains were: 44.8% for reading, 38.7% for visual motor activities, 33.6% for visual information processing and 27.5% for mobility. Despite smaller unfiltered standard error estimates (SE unfiltered mean 0.41 vs. SE filtered mean 0.46), unfiltered LVR outcomes yields only 25% of patients exceeding the MCID endpoint.
Conclusions:
A clinically meaningful endpoint in LVR can be calculated precisely by using a filtered approach because the estimation error remains small and the change score is large enough to exceed the MCID criterion in nearly half of patients obtaining services. Including items at the ceiling of the measure dilutes the effects of the targeted items on the change score, reduces measurement accuracy and underestimates the effectiveness of intervention in nearly half of patients.