Eleven RCTs were included (n=1,711 participants). Mean quality score was 7.9 out of 11 points (range 7 to 9). The main limitation in study quality was lack of blinding.
Compared to controls, Barthel Index scores were significantly higher in the groups treated at home at six to eight weeks post intervention (MD 1.00, 95% CI 0.12 to 1.88; two RCTs, n=245) and at three to six months (MD 4.07, 95% CI 0.81 to 7.34; two RCTs, n=245). There was no statistical heterogeneity (I2=0%).
At six months there was no statistically significant difference between the groups (six RCTs, n=912) and significant heterogeneity (I2=80%). Omission of one RCT (n=327) reduced heterogeneity and resulted in a mean difference that significantly favoured home rehabilitation (MD 1.04, 95% CI 0.05 to 2.04; five RCTs, n=585, I2=59%).
The narrative synthesis reported that seven RCTs found some significant benefit for home-based treatment, four RCTs found no significant difference between the groups and none found significant benefits from centre-based rehabilitation. No adverse events were reported.
Subgroup analyses did not alter the main findings.