OBJECTIVE To determine whether the assessment of disability level including basic activities of daily living (ADL) and instrumental ADL varies between patients' and proxy respondents' reports, and to explore the factors influencing proxy agreement. DESIGN Patient-proxy agreement study. SETTING Hospital settings. SUBJECTS Fifty-two stroke patients with mild or without cognitive impairments and their primary caregivers voluntarily participated in this study. MAIN MEASURES The self-administered versions of the Barthel Index and Frenchay Activities Index were completed by the patients and their proxy respondents concurrently, yet separately, to assess the level of disability. RESULTS The proxy agreement on total scores of the self-administered Barthel Index and self-administered Frenchay Activities Index demonstrated moderate to good agreement (self-administered Barthel Index: intraclass correlation coefficient (ICC) = 0.71; self-administered Frenchay Activities Index: ICC = 0.6). There were no significant differences between patient and proxy reports on self-administered Barthel Index and self-administered Frenchay Activities Index scores (paired t = -0.22, P = 0.83; paired t = 0.88, P = 0.39, respectively). However, limits of agreement of patient and proxy reports on the self-administered Barthel Index and self-administered Frenchay Activities Index were notable. Stepwise regression analysis found no significant variables associated with proxy agreement. CONCLUSIONS These results indicate that it is appropriate for research purposes, but not appropriate for clinical usage, to use proxy reports to measure disability levels in patients with stroke. Furthermore, the two methods should not be used interchangeably to monitor patients because of the wide limits of agreement between patient and proxy reports on the self-administered Barthel Index and self-administered Frenchay Activities Index.