Title

The title has been amended to clarify that the review focuses on the physical aspects of rehabilitation.

Study criteria

The original intention was to limit inclusion to studies that undertook follow up at a minimum of one month. However, because of a lack of such studies, this criterion was not applied.

Outcome measures

We clarified our meaning of function in activities of daily living to include specific measures of performance in physical ADL function, e.g. mobility.

We specified economic outcomes and additional adverse outcomes.

The original intention was to assess outcomes at the scheduled end of each trial (after follow up). However, as many studies lacked follow up, we assessed outcomes at the end of the intervention for consistency. In the narrative synthesis, we also reported follow‐up data.

We had planned, in the face of varied assessment tools, to dichotomise outcomes into deteriorated versus maintained or improved prior to meta‐analysis. For the same purpose, we specified a global poor outcome (death or deterioration). However, we have not included such varied measures in meta‐analysis because we lack the individual level data required to do this.

Search methods for identification of studies

We originally planned to search SIGLE (System for Information on Grey Literature in Europe), but we did not do so because it is no longer accessible. Furthermore, we planned to handsearch relevant journals, but because of their inclusion in electronic databases and the extensive results returned through electronic searches, we considered this unnecessary. In addition to the planned searches, we searched Google Scholar.

Data collection and analysis

We replaced the assessment of methodological quality described in the protocol and undertaken in the original review with The Cochrane Collaboration’s new 'Risk of bias' assessment tool (Higgins 2011). We reassessed all studies in the original review in line with these criteria. We performed data collection on a standardised electronic database, rather than a paper form. We clarified our approach to analysing data from cluster trials. We originally intended to combine results in a fixed‐effect meta‐analysis where sufficient homogeneity existed. However, because of the extensive heterogeneity in interventions (contents, intensity, and duration), we used a random‐effects meta‐analysis as our primary approach, but we still report the results of fixed‐effect models as sensitivity analyses. We did not perform all the subgroup analyses originally proposed in the protocol because there are too few pathology‐specific interventions for any one pathology, and studies often include both nursing and residential care homes. However, both of these groupings were partly intended to split participants by functional ability. Therefore, we instead grouped studies by baseline function in the measure being analysed. In addition to the subgroups suggested in the protocol, we added gender, duration of intervention, and risk of bias. We specified all of these subgroups before analysis commenced and presented and reported all of them for each measure.