Comparative Predictive Validity of Fall-Risk Screening Assessments in Subacute Stroke

Background and Purpose: Poststroke fall prevention is hindered by inaccurate prediction of imminent fallers. We compared the predictive validity of the Berg Balance Scale (BBS), Functional Gait Assessment (FGA), 10-m gait speed, and a novel, step test + obstacle test (STOB) in subacute stroke. Methods: A total of 64 ambulatory stroke survivors discharging home from inpatient rehabilitation completed all tests before discharge. Falls were tracked prospectively for 3 months. Logistic regression estimated the odds of being a faller with a failed test. Sensitivity, specificity, likelihood ratios, area under the curve, accuracy, and posttest probabilities for classifying fall status for each test were calculated using existing and optimized cutoffs. Results: Twenty participants (31.3%) were fallers at 3 months. STOB had the best overall model fit with odds ratio 8.0 (95% CI 2.4-26.8). FGA ≤13 had the highest sensitivity of all tests (0.85), and gait speed ≤0.47 m/s had the highest specificity (0.89). A combination of positive tests for 10-m gait speed (≤0.47 m/s), STOB (≥3), FGA (≤13), and BBS (≤43) generated a cumulative diagnostic posttest probability of future falls of 94.5%. Discussion and Conclusions: No single test emerged with strong discriminative power for predicting future falls at discharge from inpatient rehabilitation for stroke, but a combination of gait speed, STOB, and FGA may enhance prognostic decision-making accuracy.