OBJECTIVES To assess EQ-5D-5L validity in patients with acute stroke, in comparison to EQ-5D-3L, EQ VAS, modified Rankin Scale (mRS) and Barthel Index (BI). METHODS Cross-sectional study of 408 patients (51.5% males; mean age 69 years), after median 8 days from stroke onset. We assessed: construct validity in terms of known-groups validity, convergent validity of EQ-5D-5L dimensions with other stroke outcome measures, and criterion-related validity in terms of concurrent validity, with mRS as a gold standard. RESULTS A total of 2.9% EQ-5D-5L and 3.7% EQ-5D-3L questionnaires had at least one missing answer, indicating good feasibility of both instruments in patients with stroke. The proportion of patients reporting ‘no problems’ was 38.2% for BI, 6.1% for EQ-5D-3L, 5.6% for EQ-5D-5L, 5.0% for mRS and 2.5% for EQ VAS. Results of the known-groups validity tests confirmed prior hypotheses: health state utilities were lower in females, patients with high mRS score, low BI or VAS score, patients with subarachnoid hemorrhage or intracerebral hemorrhage, and when proxy respondent was used. Convergence of EQ-5D-5L dimensions with mRS, BI and EQ VAS was improved or at least the same as EQ-5D-3L dimensions. For predicting outcome in patients with stroke, the sum of mRS related EQ-5D-5L dimensions (Mobility, Self-care, Usual activities), gave 1% of false positive and 0% of false negative results. CONCLUSIONS Results support the validity of the EQ-5D-5L descriptive system as a generic measure assessed by self-report and proxy in patients with acute stroke, demonstrating some psychometric advantages in comparison to EQ-5D-3L and substantially lower ceiling effect in comparison to Barthel Index.