Results of the multivariate analysis are shown in Table 2. Combined motor function of the arm was not entered into the multivariate prediction models for upper limb function because there was a high correlation between severity of stroke and combined motor function of the arm (correlation this website between

NIHSS and sum of MAS Items 6, 7, and 8 were r = 0.64 in the model for moving a cup, and r = 0.70 in the model for feeding oneself). Age and NIHSS were statistically significant (p < 0.05) predictors of recovery in ambulation and moving a cup. For recovery in feeding oneself, only NIHSS was statistically significant. The final multivariate models ( Table 2) were used to estimate probabilities of recovery in ambulation and functional use of the arm. The probabilities are shown graphically in Figure 2. All three multivariate backwards prediction models had good discrimination (ability to differentiate between participants who did and did not recover). The AUC for the prediction models were 0.84 (95% CI 0.77 to 0.92) for ambulation, 0.73 (95% CI 0.59 to 0.87) for moving a cup, and 0.82 (95% CI 0.70 to 0.94) for feeding oneself. The Hosmer-Lemeshow test was not statistically significant for any model (0.70 for ambulation,

0.74 for moving a cup, 0.38 for feeding oneself), indicating that there was no evidence of a failure of fit. However with KU-55933 mouse the sample size used here the Hosmer-Lemeshow test lacks the statistical power needed to provide a strong test of goodness of fit. Calibration curves

are shown in Figure 3. This study provides estimates of incidence of recovery in independent ambulation and upper limb function in a representative not acute stroke cohort six months after stroke. Using age and NIHSS, we were able to develop models to predict independent ambulation and upper limb function six months after stroke. Our estimates of recovery in independent ambulation (70% of those initially unable to ambulate) and upper limb function (41 to 45% of those initially without upper limb function) are broadly consistent with previous estimates from acute stroke cohorts. In studies that followed patients up six months after stroke, 79–85% of patients have been reported to recover independent ambulation (Veerbeek et al 2011, Wade and Hewer 1987) with a smaller proportion of patients (32–34%) recovering upper limb function (Au-Yeung and Hui-Chan 2009, Nijland et al 2010). The small differences between our estimates and those from these previous studies may be due to differences in the characteristics of cohorts or differences in the definitions of recovery in upper limb function.