Item-Level Psychometrics for the Functional Gait Assessment in Persons With Stroke

To determine the item-level psychometrics of the Functional Gait Assessment (FGA) for persons with chronic stroke and create a keyform (or score sheet) for clinicians.

Retrospective cohort.

Archival item-level data from a research database.

One-hundred-one ambulatory persons (N=101) with chronic stroke (44% women, 58% right hemiparesis, average age 59y, lower extremity Fugl-Meyer 25, and overground self-selected walking speed 0.76 m/s).

Not applicable.

A principal component analysis of the residuals from the Andrich Rating Scale Model (RSM) was used to evaluate unidimensionality and item local dependence. The RSM was also used to examine the rating scale structure, item and person fit, item difficulty hierarchy, and person separation index and to generate a keyform.

Principal component analysis of the residuals confirmed the FGA's unidimensionality and that no items had local dependence. The category rating scale met the criterion and advanced monotonically. The item difficulty hierarchy was similar to that of community-dwelling older adults. The sample's mean ability level (ie, person measure) was 0.28 logits (SE=0.63). The FGA had high person reliability (0.90) despite 10% of persons misfitting. There were no floor or ceiling effects, and the FGA separated people into 4 strata. The scored FGA keyform visually showed an individual's response pattern relative to their measure value.

Rasch analysis supports the use of the FGA to measure walking balance ability in ambulatory persons with chronic stroke. An FGA keyform can provide instantaneous interval measurement for individuals.