Background: Rehabilitation interventions are diverse - making decisions about pooling data in meta-analyses challenging. Intervention reporting templates such as the Consensus on Exercise Reporting Template (CERT) may help reviewers document intervention variability.
Aim: To assess inter-rater agreement and utility of CERT used to assess completeness of reporting of one rehabilitation exercise intervention: pelvic floor muscle training (PFMT).
Design: A non-experimental agreement study.
Setting: Update of the Cochrane systematic review comparing different approaches to PFMT for urinary incontinence in women.
Population: Two PFMT arms from 21 newly identified trials.
Methods: Five raters independently used CERT to assess sufficiency of reporting of each arm (experimental PFMT and control PFMT) of each trial. One rater, PFMT non-expert, rated all trials. Four raters, all PFMT experts, assessed a mutually exclusive subgroup of the trials. In addition to rating sufficiency - "Yes" compared to No" or "Uncertain" - raters also reported on CERT utility. Expert ratings were used to determine the proportion of CERT items rated as sufficiently reported. Rater agreement was estimated using coefficient kappa and McNemar's test.
Results: The range of CERT items rated as sufficiently reported was 0 to 15 of 19 items, and the mean for both trial arms was 5.5. For agreement, 11 of 19 items had sufficient data to estimate coefficient kappa and only 3 of 11 had a kappa >0.4 (moderate agreement). From the 12 of 19 items for which McNemar's test could be performed, five had evidence that PFMT experts more often rated the reporting as sufficient than the non-expert. Raters reported the CERT template was comprehensive but not complete and needed contextualizing for PFMT.
Conclusions: Completeness of reporting was poor for this example of a rehabilitation exercise intervention, and equally poor in both trial arms. Inter-rater agreement of completeness of reporting was also poor. Using a data extraction tool with poor rater-agreement may add unnecessary burden in a review. However, using a data extraction tool that enables assessment of intervention homogeneity has benefits in making decisions about which data to pool or not.
Clinical rehabilitation impact: Researchers reporting clinical trials must pay more attention to completeness of rehabilitation exercise reporting.