Reporting quality and risk of bias in randomised trials in health professions education

ContextComplete reporting of research is essential to enable consumers to accurately appraise, interpret and apply findings. Quality appraisal checklists are giving way to tools that judge the risk for bias. ObjectivesWe sought to determine the prevalence of these complementary aspects of research reports (completeness of reporting and perceived risk for bias) of randomised studies in health professions education. MethodsWe searched bibliographic databases for randomised studies of health professions education. We appraised two cohorts representing different time periods (2008-2010 and 2014, respectively) and worked in duplicate to apply the CONSORT guidelines and Cochrane Risk of Bias tool. We explored differences between time periods using independent-samples t-tests or the chi-squared test, as appropriate. ResultsWe systematically identified 180 randomised studies (2008-2010, n=150; 2014, n=30). Frequencies of reporting of CONSORT elements within full-text reports were highly variable and most elements were reported in fewer than 50% of studies. We found a statistically significant difference in the CONSORT reporting index (maximum score: 500) between the 2008-2010 (meanstandard deviation [SD]: 242.755.6) and 2014 (mean +/- SD: 311.6 +/- 53.2) cohorts (p<0.001). High or unclear risk for bias was most common for allocation concealment (157, 87%) and blinding of participants (147, 82%), personnel (152, 84%) and outcome assessors (112, 62%). Most risk for bias elements were judged to be unclear (range: 51-84%). Risk for bias elements significantly improved over time for blinding of participants (p=0.007), incomplete data (p<0.001) and the presence of other sources of bias (p<0.001). ConclusionsReports of randomised studies in health professions education frequently omit elements recommended by the CONSORT statement. Most reports were assessed as having a high or unclear risk for bias. Greater attention to how studies are reported at study outset and in manuscript preparation could improve levels of complete transparent reporting.