Background
Clinical practices that are unnecessary or potentially harmful (i.e., low-value care [
1]) expose patients to avoidable risks of harm and are incongruent with global efforts to improve patient-centered care [
2]. Low-value care is a source of needless consumption of valuable healthcare resources and contributes to financial instability within healthcare systems [
3‐
5]. Recent estimates suggest that low-value care accounts for millions if not billions of dollars of wasteful healthcare spending within high income countries [
6‐
9]. Not only is low-value care a source of wasteful spending, but it may also be a source of unnecessary medical waste, a phenomenon recently recognized as contributing to global climate change [
10].
The recognition that certain aspects of medical care may be low-value has been acknowledged since the early twenty-first century when Fisher et al. suggested that up to 30% of all medical care in the United States may be unnecessary [
11,
12]. Other similar estimates subsequently catalyzed numerous initiatives seeking to decrease the use of low-value care (e.g. Choosing Wisely [
13]), many of which have arisen within the last 10 years [
14]. However, such extensive efforts have not translated into similar decreases in the use of low-value care [
15‐
20]. While this observation has many explanations, one important factor that remains is the lack of understanding of the process of de-implementation and the determinants that enable (i.e., facilitators) and impede (i.e., barriers) its success.
Most research examining determinants of practice change has focused on the implementation of high-value practices [
21,
22]. Although similarities likely exist between implementation of high-value care and de-implementation of low-value care, studies suggest that de-implementation is more difficult and may require different, nuanced considerations [
23]. Two studies recently examined determinants of reducing low-value care [
24,
25]. In a qualitative evidence synthesis that identified 81 articles focused mostly on low-value therapeutics, van Dulmen et al. identified provider, organizational, and patient-related factors as the most common determinants [
25]. For providers, most determinants related to their attitude towards de-implementation [
25]. For organizations it was having appropriate resources, and for patients it was knowledge of which potentially common practices may be low-value [
25]. Using scoping review methodology, Augustsson et al. found that within 101 relevant citations patients’ expectations and professionals’ fear of malpractice were prominent determinants of use and de-implementation of low-value care [
24]. Though these two recent evidence syntheses make important contributions to understanding the process of de-implementation, limitations applied to the searches may have missed potentially important determinants. Also, it is unclear how such determinants compare to those identified by stakeholders with lived experience with de-implementation, and whether they represent useful, actionable items that will improve de-implementation efforts. To address this gap, we conducted a two-phase multi-method study to identify determinants of de-implementation within the literature and compare these determinants to those identified by stakeholders within a test medical discipline, namely critical care medicine. This study is part of a broader program of research to develop a framework to guide de-implementation within acute care [
26].
Discussion
In this study, we employed several methodologies to develop and a comprehensive list of the determinants of de-implementation of low-value care from the published literature and to compare these to determinants described by those with lived experience with de-implementation within critical care medicine. From 172 articles, the systematic review with conventional content analysis identified 29 distinct barriers and 24 facilitators, of which semi-structured interviews independently cited approximately 80% of identified barriers and facilitators as determinants of de-implementation in critical care medicine. To our knowledge this is the first multi-method study to develop and corroborate a list of determinants of de-implementation. Although corroboration was confined to stakeholders from one medical discipline, the breadth of the literature from which the determinants were synthesized, combined with the similarities between the identified determinants compared to previous literature, and their otherwise actionable nature (e.g., stakeholder engagement), suggest applicability outside the test discipline, and an opportunity to influence de-implementation efforts more broadly. Collectively, the findings from this study may help to explain the observed effects of prior de-implementation interventions and inform the development of future initiatives that aim to de-implement low-value care.
Similar to prior studies examining barriers and facilitators to evidence use [
39‐
41], the current study underscores the powerful influence of the availability and credibility of scientific evidence on de-implementation. Cabana’s seminal work examining why physicians don’t follow clinical practice guidelines found that lack of awareness, familiarity or agreement with a guideline accounted for nearly 40% of identified barriers and three of seven major barrier categories [
39]. In a prior systematic review of barriers and facilitators to gaps between evidence and clinical practice, Cochrane and colleagues identified a lack of utility of evidence, in addition to lack of its awareness, as top barriers to behavior change [
41]. More recently, Tricco and colleagues examined barriers and facilitators to uptake of systematic reviews by healthcare managers and decision-makers [
40]. They found that lack of awareness of a systematic review, lack of agreement with systematic review methods in general, and lack of agreement with results of specific systematic reviews were the main determinants of systematic review utilization among managers and decision-makers. More specific to de-implementation, two recent studies examined determinants of use and de-implementation of low-value care [
24,
25]. van Dulmen et al. found a predominance of barriers related to individual healthcare provider and patient attitude and knowledge as well as availability of resources within an organization [
25]. Many articles cited patient preference and expectations combined with physicians’ communication and time as important barriers to de-implementation. Augustsson et al. also identified that patient expectations and physicians’ fear of malpractice are prominent determinants of de-implementation [
24]. Our study similarly identified patients’ knowledge and expectations and clinicians’ resistance to change as frequent determinants of de-implementation of low-value care. However, in our study, cited more frequently within the included articles were the importance of the quality and availability of evidence that underpins a de-implementation initiative as well as stakeholder collaboration. This was subsequently identified by interviews among stakeholders with lived experience with de-implementation. Taken together with these two prior studies, there is now a more advanced understanding of the determinants of de-implementation, how they are similar to implementation, and the nuanced differences. Future de-implementation initiatives should focus on clinical practices defined as low-value by strong scientific evidence, seek early stakeholder engagement including patients, healthcare professionals, and decision-makers, and broadly educate stakeholders regarding the risks and benefits of de-implementing clinical practices deemed to be low-value.
That a frequent barrier to de-implementation is a lack of credible evidence demonstrating a given clinical practice to be low-value helps explain why after nearly a decade speaking about low-value care [
42], consensus has yet to be reached on what constitutes low-value care, and how it should be identified [
14]. Naturally occurring clinical heterogeneity creates a spectrum of value within clinical practice; a test or treatment may be considered low-value in one clinical context but not in another, and it is difficult for science to adequately examine efficacy or effectiveness of all clinical practices in all contexts. In contrast to prior reviews examining determinants of evidence use [
39‐
41], methods for identifying clinical practices that are low-value and prioritizing them for de-implementation was more commonly cited as a barrier to de-implementation compared to implementation and appears critically important to stakeholder receptiveness to de-implementation initiatives. It is thus clear that defining methods for identifying and prioritizing low-value clinical practices for de-implementation should be a priority within de-implementation research. This process needs to be systematic, grounded in evidence and contemporary data demonstrating overuse of the low-value practice, and from the beginning engage relevant stakeholder groups, and not simply be the distillation of experts hand-picking from available literature, or their own personal lists [
15].
Our findings also demonstrate that framing can impact the effectiveness of de-implementation efforts. Framing de-implementation as an opportunity for cost-savings or reallocation of resources was identified as a facilitator in the systematic review, whereas framing it as cost-cutting was identified as a barrier. These slight but important differences in framing can substantially impact stakeholder attitudes towards behavior change. Patient demands and preferences were also identified as a common barrier to de-implementation in the systematic review. This suggests that even if clinicians intend to change their behavior and reduce low-value care, their intentions could be derailed by patient preferences for low-value tests (e.g. MRI for low-back pain [
43]) or treatments (e.g. antibiotics for viral infections [
43]), and greater attention to the importance of patient-engagement in de-implementation interventions is needed [
44].
The results of this study need to be interpreted in the context of its limitations. First, the search strategy restrictions by date and language may have caused omission of relevant articles. However, the 172 included articles as well as 437 raw text barriers and 280 raw text facilitators that were synthesized into 29 and 24 unique barriers and facilitators, respectively, are larger in number than that described in prior reviews on determinants of evidence use [
39‐
41], and describe concepts congruent with the main results of those reviews. Furthermore, the search was conducted in 2016, and articles published since this time will not have been included. However, recent evidence syntheses by Augustsson et al. and van Dulmen et al., found similar challenges facing de-implementation initiatives [
24,
25], suggesting potentially missed citations are unlikely to change our main results. Second, synthesizing a list of distinct barriers and facilitators from individual articles was a potentially subjective process. To mitigate this challenge, we had two reviewers code and review emerging representative barriers and facilitators. Third, is the nature of the included studies. A number of included citations were non-original research (only one randomized clinical trial), and the majority of original research citations were of low-to-moderate methodological quality from high-income countries. Therefore, the list of distinct barriers and facilitators derives from a cohort of mostly low methodological quality articles specific to the high-income country context. Despite this, the majority of the distinct barriers and facilitators were derived from data extracted from original research articles, and the final list is a comprehensive representation of what is reported in the literature. Fourth, the specific nature of our interview sample (critical care medicine stakeholders from one province) may limit transferability of the findings to other stakeholder groups. However, the high level of agreement between the interviews and the systematic review suggest that interviews exploring determinants of de-implementation in other medical disciplines are likely to yield similar results. Lastly, while there was an extended time period between conducting the search for the systematic review and conducting stakeholder interviews, the fact that determinants identified from the literature overlapped with determinants noted by stakeholders suggests that the extended timeframe between phases is unlikely to have influenced the main results.
Conclusions
Using a multi-method approach, this study identified 29 distinct barriers and 24 distinct facilitators to the de-implementation of low-value care from the published literature, of which the majority were also cited in interviews among stakeholders with lived experience with de-implementation in critical care medicine. Lack of credible evidence defining a practice as low-value, entrenched norms and clinicians’ resistance to change, and challenges with securing, mobilizing and maintaining stakeholder support were identified as frequent barriers, while stakeholder collaboration and communication, availability of credible evidence, and execution of de-implementation at the system-level were the most frequent facilitators. Additional work is required to determine if the identified list of determinants to de-implementation is relevant to stakeholders working in other medical disciplines, and to develop a comprehensive, evidence-informed model for de-implementation. However, in the meantime, de-implementation determinants identified in this study may be used to inform future de-implementation initiatives.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.