Background and Purpose- This study investigated the diagnostic accuracy of the Barthel Index (BI) in 206 stroke patients, measured within 72 hours, for activities of daily living at 6 months and determined whether the timing of BI assessment during the first days affects the accuracy of predicting activities of daily living outcome at 6 months. Methods- Receiver operating characteristic curves were constructed to determine the area under the curve and optimal cutoff points for BI at Days 2, 5, and 9. OR, sensitivity, specificity, positive predictive value, and negative predictive value were calculated to predict BI 19. Results- The area under the curve ranged from 0.785 on Day 2 to 0.837 and 0.848 on Days 5 and 9. Comparison of the receiver operating characteristic curves showed that the area under the curve was significantly different between Days 2 and 5 (P<0.001) and between Days 2 and 9 (P<0.001). No significant difference was found between Days 5 and 9 (P=0.08). Using a BI cutoff score of 7, the positive predictive value gradually increased from 0.696 on Day 2 to 0.817 on Day 2 to 0.864 on Day 9, whereas negative predictive value declined from 0.778 on Day 2 to 0.613 on Day 9. Conclusions- Assessment of the BI early poststroke showed good discriminative properties for final outcome of BI at 6 months. However, Day 5 proved to be the earliest time for making an optimal prediction of final outcome of activities of daily living. The BI should be measured at the end of the first week in hospital-based stroke units for early rehabilitation management. © 2011 American Heart Association, Inc.