The questions reflect performance on activities covering domestic chores,

household maintenance, service to others and social activities over the last three months. Each activity is rated with four possible responses from 0–3, where a higher score reflects more participation. For the purposes of this study, and in line with recommendations, community participation was reported as a score out of 72. Further details on study protocols and data collection are in Appendix 1 on the eAddenda. We undertook an see more a priori power calculation to determine sample size based on primary outcome measures. About 50% of non-ambulatory patients walk independently at discharge ( Dean and Mackey 1992). We designed the study to detect a 25% increase in the proportion of non-ambulatory patients walking independently, ie, from 50% to 75%. The smallest number of participants to detect this difference between two proportions estimated from independent samples with 80% power at a two-tailed 5% significance level was 65 participants per group, ie, 130 participants in total ( Fleiss 1981). The secondary

outcomes were analysed using independent sample t-tests with a significance level of p < 0.05. The mean difference between the groups and a 95% CI was calculated for all the outcome measures. For participants who withdrew or died, data were censored at the time of withdrawal or death. One hundred and twenty-six participants were below recruited to the study between August 2002 and September 2008. The baseline characteristics of the participants are presented in Table 1. Sixty-four participants Buparlisib research buy were allocated to the experimental group and 62 to the control group. Two participants in the experimental group withdrew because of anxiety when using the treadmill. At 6 months after admission to the study, there were 59 participants in the experimental group and 60 in the control group. Figure 1 outlines the flow of participants through the trial. Twenty-five physiotherapists, on average 10 years (SD 9) since graduating, provided the

intervention. Six (24%) had relevant postgraduate qualifications and 12 (48%) had research experience. On average, therapists were involved in the study for 3 years (SD 2, range 1 to 6) and trained 5 participants (SD 5, range 1 to 19). Most therapists trained both experimental and control participants, although 8 (32%) trained only one participant each. Rehabilitation units at six centres participated in the trial: three had on-site acute stroke units, two were rehabilitation units only, and one had its acute stroke unit at a different location. The annual throughput of stroke patients averaged 314 (SD 121, range 118 to 444), and the physiotherapist: patient ratio averaged 1:8. The number of participants in each group was similar within each centre (Table 1).