Predicting Discharge Walking Function With High-Intensity Stepping Training During Inpatient Rehabilitation in Nonambulatory Patients Poststroke.

Objective: This cohort investigation identified primary predictors of discharge walking function of nonambulatory individuals poststroke with high-intensity training (HIT) during inpatient rehabilitation.

Design: Observational cohort investigation.

Setting: Inpatient rehabilitation.

Participants: Data were collected from individuals (N=257) <6 months poststroke who required assistance to walk at admission.

Intervention: Clinical physical therapy interventions attempted to maximize stepping practice at higher intensities.

Main Outcome Measures: Primary outcomes included the discharge level of assistance required during walking (minimal or no assistance) and attainment of specific gait speed thresholds (0.4 and 0.8 m/s) during the 10-m walk test. Independent predictors were demographics, training interventions (including steps/day), baseline Berg Balance Scale (BBS), and paretic leg strength.

Results: Participants performed a median (interquartile range) of 1270 (533-2297) steps per day throughout inpatient rehabilitation, with significant differences between those who walked with versus without assistance at discharge. Logistic regressions indicate steps per day was a primary predictor of unassisted walking recovery; removal of steps per day resulted in primary predictors of baseline BBS and strength. Receiver operating characteristic (ROC) analyses indicate significant areas under the curve for BBS and relatively low cutoff scores of 5.5 points at admission to walk without assistance at any speed. ROC analyses performed using 1-week outcomes indicate BBS scores of 5-17 points were needed to achieve locomotor thresholds.

Conclusion: Stepping activity, BBS, and paretic leg strength were primary predictors of walking outcomes in patients performing HIT, and ROC analyses indicated recovery of independent walking could be achieved in low functioning patients early poststroke.