De-Implementation Review

A 2014 review by Vinay Prasad and John Ioannidis in the Journal of Implementation Science reviewed strategies for de-implementing unproven or harmful medical procedures. The authors conducted a review of the available evidence to determine the best approach for removing practices that are not supported by science. They identified and made recommendations for three categories of healthcare practice that may need de-implementation: Practices that are known to be ineffective, practice without evidence proving effectiveness, and novel medical practices that are undergoing validation.

New evidence that suggests current practice is ineffective frequently comes under attack in one of several ways. Proponents of current practice will publish contradictory editorials, critique the evidence using subgroup analysis to achieve different results, or will conducting opposing studies with modified designs of weaker methodology. The problem with this process Prasad and Ioannidis argue is that weaker evidence and contradictory editorials are creating the false appearance that the established practice is still under debate, this creates delays to de-implementation allowing the practice to remain entrenched. The authors discuss the example of gown and glove controls as prevention for MRSA/VRE transmission to illustrate their point. The evidence for these control measures comes from a non-experimental before-after design study, that hasn't been reproduced. The intervention has failed to change colonization rates in subsequent cluster randomized trials; but the results from the weaker initial study have become entrenched in practice. Clinicians in this example have gone from accepting weak evidence, to being ambivalent about contradictory evidence, to continuing the use of an unsupported intervention until it has been not only been proven ineffective, but proven harmful. The authors highlight that in these instances de-implementation should begin with high quality evidence and progress until contradictory evidence comes from studies that are, at the very least, of equivalent rigor, and preferably of higher quality.

Unproven medical practices, or practices where we have no evidence to support our action are another area where de-implementation can be difficult. The authors discuss a Cochrane review that determined that the "existing evidence base was unable to support or refute 49% of interventions, and [that] 48% of American College of Cardiology recommendations are supported by expert opinion only". They suggest that a lack of evidence is pervasive and that much of clinical practice is unsupported by evidence. The key to de-implementation in these instances the authors suggest is through systematic testing of existing standard practice, with preference being given to testing practices that are common, that have optional approaches, have high costs, and have the weakest level of evidence.

The final recommendation for de-implementation suggests using preventative measures to ensure that the novel approaches that are being validated don't become entrenched. Prasad and Ioannidis use the example of experimental intracranial arterial stents to illustrate their point. In this instance a single uncontrolled study was used to gain FDA approval to use stents for arterial stenosis. Funding was approved by the Centers for Medicare & Medicaid Services (CMS), but only for RCT; there was fierce pressure from industry to open up the use of the stents, but CMS refused to pay for a liberalized use of the procedure. The result was after 6 years the only RCT on the procedure proved the practice almost tripled 30 day mortality, the CMS move likely saved thousands of lives. In this instance the authors argue the move prevented the need to pursue extensive de-implementation, and they recommend a similar strategy be used with all new treatments.


This review offers some guidance on how to categorize practices that need to be discontinued, Prasad and Ioannidis offer some general approaches to de-implementation, unfortunately however they don't offer any concrete strategies for how to implement these approaches. The review failed to discuss the selection criteria or review methodology for practices requiring change, how they determined their categories of practices that need discontinuation, or to directly link evidence to what the best approaches for de-implementation are. It was also somewhat lacking in the area of specific recommendations for how to action their general recommendations. From a nursing perspective this article doesn't provide concrete steps for de-implementing practices; but it does offer some global considerations for evaluating clinical practice, it suggests some metrics by which to evaluate practice, and has uses clinically relevant case studies, the discussion about isolation precautions alone makes this worth reading.

http://www.ncbi.nlm.nih.gov/pubmed/24398253