Background: The World Health Organisation stresses the need to collect high quality longitudinal data on rehabilitation and to improve the comparability between studies. This implies using all the information available and transparent reporting. We therefore investigated the quality of reported or planned randomised controlled trials on rehabilitation post-stroke with a repeated measure of physical functioning, provided recommendations on the presentation of results using regression parameters, and focused on the difficulties of adjustment for baseline outcome measures.
Methods: We performed a systematic review of the literature from 2011 to 2013 and collected information on the way data was analysed. Moreover we described various approaches to analyse the data using mixed models illustrated with real data.
Results: Eighty-four eligible studies were identified of which 61% (51/84) failed to analyse the data longitudinally. Moreover, for 30% (25/83) the method for adjustment for baseline is not known or not existent. Using real data we were able to show how much difference in results an adjustment for baseline data can make. We showed how to provide interpretable intervention effects using regression coefficients while making use of all the information available in the data.
Conclusions: Our review showed that improvements were needed in the analysis of longitudinal trials in rehabilitation post-stroke in order to maximise the use of collected data and improve comparability between studies. Reporting fully the method used (including baseline adjustment) and using methods like mixed models could easily achieve this.