Abstract
BACKGROUND: Phase angle (PhA) is a potential indicator of nutritional status and functional outcomes. However, optimal regional PhA cutoff values for predicting activities of daily living (ADL) post-stroke are unclear. We aimed to determine these cutoffs and assess whether PhA adds prognostic value to established clinical models. METHODS: In this retrospective cohort study, stroke patients undergoing inpatient rehabilitation had body composition and PhA measured via multi-frequency bioelectrical impedance analysis. The primary outcome was ADL independence at discharge (motor Functional Independence Measure >78). Receiver operating characteristic (ROC) curves identified optimal PhA cutoffs; logistic regression assessed their predictive value, adjusting for confounders. Incremental discriminative ability was evaluated using area under the ROC curve (AUC) comparisons with DeLong's test. RESULTS: Among 1,080 patients (median age, 75.6 years; 54.1% men), optimal PhA cutoffs were: whole-body (women 3.90°, men 4.60°), non-paretic upper limb (4.45°, 4.90°), and non-paretic lower limb (4.00°, 4.80°). Higher PhA values above these cutoffs were independently associated with ADL independence (all p<0.05). Adding whole-body PhA to the baseline model improved AUC from 0.937 to 0.970 (ΔAUC=0.033; p=0.011); significant gains were observed for limb PhA. CONCLUSIONS: Optimal site-specific PhA cutoffs predict ADL independence at discharge in post-stroke patients. PhA provides significant incremental prognostic value beyond established clinical models, supporting its use in tailoring rehabilitation strategies.