The Development and Evaluation of Performance-Based Functional Assessment: A Methodology for the Measurement of Physical Capabilities
Katherine J. Price, Andrew Sears · 2009 · ACM Transactions on Accessible Computing · doi:10.1145/1530064.1530068
Summary
This paper introduces PB-FACT (Performance-Based Functional Assessment for Computer Technology), a methodology for objectively measuring the physical capabilities of individuals with motor impairments to inform assistive technology selection. Traditional approaches rely on medical diagnoses or subjective self-reports, neither of which adequately captures an individual's actual functional capabilities for technology interaction. The authors argue that two people with the same diagnosis may have vastly different abilities to use specific input devices. The study recruited 31 participants—11 without motor impairments and 20 with various conditions including stroke, spinal cord injuries at different levels, cerebral palsy, Parkinson's disease, and multiple sclerosis. Participants wore a P5 gaming glove equipped with motion sensors tracking hand position in 3D space plus finger bend sensors. They performed 14 everyday motions (grasping objects, tracing lines, waving, making fists, etc.) while the glove captured raw movement data at 60 samples per second. From this raw data, 45 metrics were computed measuring movement characteristics: amplitude, duration, variability, direction changes, and pause patterns. The researchers used stepwise regression to identify which metrics best predicted independent observer ratings of movement quality, producing task-specific predictive models.
Key findings
The predictive models explained substantial variance in observer ratings, with the best models achieving up to 92% accuracy. A critical finding was that metrics measuring extraneous or unintended movement (76% of selected metrics) were far more predictive than metrics measuring the intended motion itself (24%). This suggests that variability and tremor-like movements are key indicators of functional capability. The research demonstrated that medical diagnosis alone is insufficient for technology matching. Tables in the paper show participants with the same diagnosis (e.g., post-stroke) having dramatically different PB-FACT scores, while participants with different diagnoses sometimes showed similar functional capabilities. Self-reported surveys uniformly showed participants rating themselves as having no difficulty, even when observers noted clear challenges. The complete assessment takes approximately 30 minutes, making it practical for clinical use. The three-fold validation process (metric selection, model building, validation using separate participant groups) provides confidence in the methodology's robustness.
Relevance
PB-FACT addresses a fundamental gap in assistive technology provision: the lack of objective, quantifiable measures of functional capability. For practitioners, this methodology offers a path toward evidence-based device recommendations rather than trial-and-error approaches. The finding that diagnosis does not predict capability challenges common assumptions in AT assessment. The emphasis on measuring movement variability rather than just task completion has implications for interface design—systems could potentially adapt based on detected tremor or inconsistency rather than waiting for errors. While the P5 glove hardware is dated, the methodology could be implemented with modern motion capture technology. Future work extending this to cognitive and sensory capabilities could create comprehensive user profiles for technology matching.
Tags: motor impairments · functional assessment · assistive technology · motion sensors · user capabilities · performance-based testing