Background and purpose: Early prediction of future functional abilities is important for stroke management. The objective of the present study was to investigate the predictive value of the 13-item National Institutes of Health Stroke Scale (NIHSS), measured within 72 h after stroke, for the outcome in terms of activities of daily living (ADL) 6 months post stroke. The second aim was to examine if the timing of NIHSS assessment during the first days post stroke affects the accuracy of predicting ADL outcome 6 months post stroke. Methods: Baseline characteristics including neurological deficits were measured in 188 stroke patients, using the 13-item NIHSS, within 72 h and at 5 and 9 days after a first-ever ischemic hemispheric stroke. Outcome in terms of ADL dependency was measured with the Barthel Index (BI) at 6 months post stroke. The area under the curve (AUC) from the receiver operating characteristic (ROC) was used to determine the discriminative properties of the NIHSS at days 2, 5 and 9 for outcome of the BI. In addition, at optimal cut-off odds ratio (OR), sensitivity, specificity, positive (PPV) and negative predicted values (NPV) for the different moments of NIHSS assessment post stroke were calculated. Results: One hundred and fifty-nine of the 188 patients were assessed at a mean of 2.2 (1.3), 5.4 (1.4) and 9.0 (1.8) days after stroke. Significant Spearman rank correlation coefficients were found between BI at 6 months and NIHSS scores on days 2 (r(s)=0.549, p<0.001), 5 (r(s)=0.592, p<0.001) and 9 (r(s)=0.567, p<0.001).The AUC ranged from 0.789 (95%Cl, 0.715-0.864) for measurements on day 2 to 0.804 (95%Cl, 0.733-0.874) and 0.808 (95%Cl, 0.739-0.877) for days 5 and 9, respectively. Odds ratio's ranged from 0.143 (95%Cl, 0.069-0.295) for assessment on day 2 to a maximum of 0.148 (95%Cl, 0.073-0.301) for day 5. The NPV gradually increased from 0.610 (95%Cl, 0.536-0.672) for assessment on day 2 to 0.679 (95%Cl, 0.578-0.765) for day 9, whereas PPV declined from 0.810 (95%Cl, 0.747-0.875) for assessment on day 2 to 0.767 (95%Cl, 0.712-0.814) for day 9. The overall accuracy of predictions increased from 71.7% for assessment on day 2 to 73.6% for day 9. Conclusions: When measured within 9 days, the 13-item NIHSS is highly associated with final outcome in terms of BI at 6 months post stroke. The moment of assessment beyond 2 days post stroke does not significantly affect the accuracy of prediction of ADL dependency at 6 months. The NIHSS can therefore be used at acute hospital stroke units for early rehabilitation management during the first 9 days post stroke, as the accuracy of prediction remained about 72%, irrespective of the moment of assessment. (C) 2010 Elsevier B.V. All rights reserved.