Strategic Design for Users with Diabetic Retinopathy: Factors Influencing Performance in a Menu-Selection Task
Paula J. Edwards, Leon Barnard, V. Kathlene Emery, Ji Soo Yi, Kevin P. Moloney, Thitima Kongnakorn, Julie A. Jacko, François Sainfort, Pamela R. Oliver, Joseph Pizzimenti, Annette Bade, Greg Fecho, Josephine Shallo-Hoffmann · 2004 · Proceedings of the 6th International ACM SIGACCESS Conference on Computers and Accessibility (Assets 04) · doi:10.1145/1028630.1028652
Summary
This paper examines how interface design features affect menu selection performance for users with Diabetic Retinopathy (DR), a progressive eye condition affecting 40-45% of the approximately 18 million Americans with diabetes. The study involved 25 volunteers from Nova Southeastern University's College of Optometry, divided into three groups: a control group with no ocular disease (n=8), Group 1 with evidence of DR but normal visual acuity (20/20-20/30, n=9), and Group 2 with diabetic retinopathy and reduced acuity (20/20-20/30 to 20/30-20/50, n=8). Participants performed menu selection tasks under four interface conditions created by crossing two factors: multimodal feedback (normal visual-only vs. visual plus auditory plus haptic) and Windows accessibility settings (standard vs. accessibility-enhanced with larger text, high contrast, larger cursors). The multimodal feedback used a Saitek TouchForce optical mouse providing both auditory (abstract sound) and haptic (mechanical vibration) feedback when the cursor crossed menu boundaries. The menus were three-item pull-down menus positioned in left, middle, and right locations, with the target item appearing at top, middle, or bottom positions. A linear regression model incorporating visual, cognitive, physical health, and task variables was developed to predict performance.
Key findings
The regression model accounted for approximately 69% of the variation in task completion time (R²=69.5%, adjusted R²=69.3%, p<0.001). Windows accessibility settings significantly improved performance overall (F=50.51, p<0.001), but critically, this effect interacted with acuity group (F=9.91, p<0.001): the settings had a much larger positive effect on Group 2 (more progressed DR with reduced acuity) than on the Control Group or Group 1. The Windows accessibility settings reduced predicted task time by approximately 11% (472ms). Surprisingly, multimodal feedback had a statistically significant but practically negligible main effect on task time (mean increase of only 137ms, or 0.137 seconds), contradicting prior research showing stronger benefits. This may be because the feedback was not "intelligent" — sounds and vibrations were triggered by menu boundary crossings regardless of proximity to the target, providing no directional guidance. For errors (Missed Opportunities — instances where the correct item was highlighted but not selected), Windows accessibility settings significantly reduced errors for Groups 1 and 2, while multimodal feedback had no significant effect on errors for any group. The regression model identified significant predictors including: menu location (right-side menus 20% slower), item position (bottom items 14% slower), age (40 to 70 years increased time by 35%), computer experience (low to high decreased time by 18%), LogMAR visual acuity (20/20 to 20/50 increased time by 35%), visual field mean deviation, cognitive status (MMSE score), and learning effects (21% improvement from first to last trial).
Relevance
This research provides valuable quantitative evidence for accessibility practitioners designing interfaces for users with low vision, particularly the large population affected by diabetic retinopathy. The key practical finding is that built-in operating system accessibility settings (larger text, high contrast, larger cursors) provide meaningful performance improvements, especially for users with more progressed vision loss — suggesting that raising awareness of these existing settings could be a cost-effective intervention. The unexpected finding that multimodal feedback provided negligible benefit challenges assumptions about the universal value of adding sensory channels, and highlights that feedback must be contextually relevant ("intelligent") rather than merely present to be useful. The regression model's identification of multiple interacting factors — age, computer experience, visual acuity, menu position, cognitive status, and physical health — demonstrates that designing for users with visual impairments requires accounting for the complex intersection of visual, cognitive, and experiential factors. A limitation is the controlled laboratory setting with simple menu tasks, which may not generalize to real-world application use with more complex interfaces.
Tags: diabetic retinopathy · low vision · visual impairment · multimodal feedback · menu design · user interface design · accessibility settings · haptic feedback · auditory feedback · aging