OBJECTIVE To evaluate the utility of the 12-item World Health Organization Disability Assessment Schedule (WHODAS-12) in predicting institutionalization after subacute stroke rehabilitation. METHODS Prospective observational study. METHODS On a specialized rehabilitation ward, discharge WHODAS-12 scores of 156 consecutive patients (24-h National Institutes of Health Stroke Scale (NIHSS) ≥ 15) and assessment from their proxies were compared, and receiver operating characteristic curves for predicting institutionalization were generated. Clinician-rated assessments of functioning were applied for comparison. RESULTS Thirty-three percent of the patients were unfit to respond, due to the consequences of major stroke. However, both patient and proxy WHODAS-12 sum scores differentiated the community (n = 70) and institution (n = 86) groups (p = 0.02 and p < 0.0001, respectively), the discriminative accuracy (area under the curve; AUC) being 0.63 and 0.79, respectively. In proxy assessments, the institutionalized patients were significantly more impaired in all item comparisons except for emotions and concentrating. Ability to participate differentiated the groups as accurately as activities (AUC 0.75 vs 0.78, respectively). The corresponding discriminative accuracy of the clinician-rated World Health Organization (WHO) minimal generic dataset sum score and modified Rankin Scale were 0.74 and 0.79 (p < 0.0001), respectively. CONCLUSIONS Despite its brevity and subjectivity, the WHODAS-12 from proxies has shown high accuracy in predicting institutionalization after subacute rehabilitation of individuals with major stroke, the impact of participation being as relevant as that of activities.