OBJECTIVES: Residual disability after stroke presents a major economic and humanistic burden. To quantify disability in patient, three disability instruments are globally used: the Barthel Index (BI), Motor component of Functional Independence Measure (M-FIM), and Modified Rankin Scale (MRS). Study objectives are: 1) to develop a crosswalk among these measures by categorizing scores presenting clinical significance; and 2) to examine the variation of the categorization scheme over time from stroke onset. METHODS: The Kansas City Stroke Study (KCSS) data was utilized for the current study. Disability measures at baseline (stroke index date), one, three and six month after index date were collected by trained clinicians. Polytomous logistic regression analysis was applied to produce probabilistic distributions of BI or M-FIM scores, representing clinically distinct disability states, which were defined by the six levels of the MRS. Model fit statistics were examined to verify logistic model appropriateness. A categorization scheme, which minimized the false positive response rate, was selected as the optimal categorizing system. RESULTS: BI and M-FIM differentiated disability well in lower ADLs than higher ADLs. For BI, four clinically distinct disability levels were differentiated with the following categorization scheme: 0 £ MRS5 <15, 15 £ MRS4 <70, 70 £ MRS3 <95, and 95 £ MRS(0+1+2) £ 100. For the M-FIM, three stages were differentiated: 13 £ MRS5 <26, 26 £ MRS4 <62, and 62 £ MRS(0+1+2+3) £ 91. Each wave, baseline, 1, 3, and 6 month showed similar trend in cut-offs. CONCLUSIONS: The categorization scheme developed can serve as a crosswalk between measures. However, due to the ceiling effect of the BI and M-FIM, the crosswalk could not be completed for all six distinct levels in the MRS. No apparent variation over time in the categorization scheme was observed.