Predictive validity of obstacle-crossing test variations in identifying fallers after inpatient rehabilitation for stroke.

Background: The ability to step over an obstacle is often evaluated as part of fall-risk and balance assessments. Although different obstacle-crossing tests exist, their comparative predictive validity in stroke is unknown.

Objectives: To examine the predictive validity of different obstacle depths and different obstacle-crossing tests, including a novel, custom-height test and an existing "one-size-fits-all" obstacle test, for predicting post-stroke fallers.

Methods: 46 independently ambulatory adults with stroke completed a custom-height obstacle-crossing test with 3 depths (0.5-inch, 1.5-inch, 3.0-inch) and the Functional Gait Assessment (FGA) 1-3 days before hospital discharge. Falls were tracked prospectively for 3 months using fall calendars and fortnightly phone calls.

Results: 35% of participants fell at least once in 3 months. Test accuracy was not significantly different between obstacle depth conditions. However, the 0.5-inch obstacle depth condition demonstrated the highest sensitivity and specificity, and participants who failed were 9 times more likely to fall in the first 3 months after discharge than those who passed (95% CI 1.9, 42.1; p = 0.005). Performance on the obstacle item of the FGA at hospital discharge was not significantly associated with fall status at 3 months post-discharge and had a 50% floor effect.

Conclusions: The ability to step over a custom-height obstacle may be a good indicator of post-stroke fall status 3 months after hospital discharge. Subtle increases in obstacle depth did not significantly alter accuracy. The "one-size-fits-all" obstacle test from the FGA had poor predictive validity at discharge from inpatient rehabilitation for stroke.