The quality of trials in quality improvement: Systematic review of risk of bias in diabetes studies

Article type
Authors
Ivers N1, Tricco A2, Turner L3, Moher D4, Grimshaw J5
1Department of Family Medicine, Women’s College Hospital, University of Toronto, Canada
2Li Ka Shing Knowledge Institute of StMichael’s Hospital, Toronto, ON, Canada
3Ottawa Methods Centre, Ottawa Hospital Research Institute, Canada
4Department of Epidemiology and Community Medicine, Faculty of Medicine, University of Ottawa, Canada
5Clinical Epidemiology Program, Ottawa Hospital Research Institute, Canada
Abstract
Background: Of the 356 studies included in seven recent Cochrane reviews evaluating interventions to improve health care professional practices, 14.3% were deemed as having a high risk of bias. Despite repeated calls for reducing the risk of bias in trials, no previous reviews have assessed whether risk of bias in quality improvement (QI) trials has changed over time.

Objectives: To assess for changes in risk of bias of QI trials over time.

Methods: This was a secondary analysis of data from a systematic review of 142 trials of QI interventions for diabetes. Risk of Bias was abstracted using the Cochrane 'Risk of bias' tool. We also grouped studies such that those with high risk of bias in one of allocation sequence, concealment, blinding, or outcome reporting were deemed high risk of bias overall.

Results: In the diabetes QI review, 25trials were published before 2001, 50 from 2001-2005, and 68 from 2006-2010. There was no improvement over time for 8 of 9Risk of Bias domains; only 'similarity in baseline covariates' showed a trend for improvement, with the proportion of trials at high risk of bias decreasing from 47% prior to 2001 to 13% in those published 2006-2010. Loss to follow up was a common source of bias in recent trials, with 23% of trials published 2006-2010 judged high risk of bias. Allocation sequence (58% unclear) and allocation concealment (56% unclear) were poorly reported. Overall, 14% of the trials were deemed high risk of bias; this has not improved over time.

Conclusions: Although other methodological reviews of RCTs have shown improvement over time, this is not reflected in the QI literature. There is a need to reduce the risk of bias and improve the quality of reporting of QI trials so that stakeholders have adequate evidence for action.