1 Background

Pain management is an essential component of patient care. In acute hospitals, appropriate pain management contributes to earlier patient mobilisation, shortened lengths of stay, and reduced healthcare costs [1, 2]. Opioids are regarded as a vital part of the accepted standard of care to effectively manage moderate to severe acute pain, particularly after trauma or surgery [2, 3]. However, there is increasing evidence of long-term unintended harms associated with hospital prescribing of opioids [4,5,6]. As a result, multiple safety strategies, and hospital-based quality improvement initiatives, often referred to as ‘opioid stewardship’ interventions, have been implemented [7,8,9,10]. As with many healthcare quality improvement reports, published literature describing opioid stewardship interventions frequently lacks detail about the implementation process to allow for replication of the interventions elsewhere [11, 12].

The issue of absent implementation reporting is further complicated by the challenges of translating evidence into practice that stem from the complex nature of healthcare systems, which can respond differently to the same interventions [13]. Successful change to practice is not only dependent on the characteristics of the intervention itself, but also on the relationship of the intervention to the setting (context) and the process of introducing the intervention (implementation) [13, 14]. Murray et al. [15] describes this challenge in three stages. Firstly, implementing and securing acceptance of new solutions is difficult, even when armed with persuasive evidence(‘the take-up problem’) [15]. Secondly, disseminating knowledge of an intervention’s benefits across the entire system is hard (‘the diffusion problem’) [15]. Thirdly, even if a new model of care, technology, or practice is successfully adopted and widely spread, its shelf life will be short (‘the sustainability problem’) [15]. Aside from the complexity of the system, interventions themselves can be complex, that is, ‘built up from a number of components, which may act both independently and inter-dependently’ [15]. Therefore, the implementation of complex interventions is time-consuming and, without sufficient planning, can fail to create sustainable and meaningful change [13]. It is clear that considering and adapting implementation processes is critical to the success of an intervention.

Implementation science helps to systematically identify and describe the characteristics (the intervention, the stakeholders to engage, and navigating the context) required for successful change to practice and identify reasons why success or failure occurs [15, 16]. Normalisation Process Theory (NPT) is a commonly used implementation research theory [15, 17, 18]. It describes four domains: coherence (sense-making); cognitive participation (engagement); collective action (the work required to implement the intervention); and reflexive monitoring (formal and informal feedback on the intervention). These domains can be used to describe and summarise interventions and implementation strategies, including in reporting [15].

The rationale for using NPT in this review was based on several factors. Firstly, NPT has been used extensively for both implementation and evaluation of healthcare improvement initiatives [15, 17, 19]. Secondly, the structured process of NPT allows researchers to understand the dynamic relationship between the intervention and its recipients (e.g., clinicians, policymakers, and patients), provides insights into enablers and barriers affecting practice change and has been used as an explanatory framework [15, 17, 19, 20]. Finally, opioid stewardship interventions are complex by nature, because of the need to balance pain management with (opioid) safety. Given the current interest and volume of opioid-related improvement activities, there is an opportunity to scale and spread interventions that have been proven to improve patient outcomes, if sufficient detail on the intervention and implementation are provided in the published literature.

Our aim was to systematically review studies describing opioid stewardship interventions in acute hospital settings and identify those that included an evaluation and a description of their implementation.

2 Methods

2.1 Protocol

Our methodology was guided by similar systematic reviews on implementation research [17, 20,21,22]. The research questions used to assess the final studies identified were developed through consultation among the research team and key stakeholders with experience in healthcare improvement or opioid stewardship.

2.2 Concept

Our research questions were “what work is required to implement interventions to improve opioid prescribing in acute hospital settings?” and did the authors “provide sufficient implementation detail to allow decision makers to assess the potential adaptation and replication of the intervention in a different healthcare setting?”.

2.3 Inclusion and exclusion criteria

This review was limited to studies based in adult acute hospital clinical settings, including acute post-surgical, medical wards, and emergency departments. Interventions that aimed to improve communication between hospitals and community providers were included but those undertaken solely in community or outpatient settings were excluded. It was also limited to countries with comparable health systems and models of care in acute settings and with significant cultural overlap with Australia because pain experience is intrinsically linked to societal and cultural beliefs [23] (see Table 1 for included countries).

Table 1 Inclusion and exclusion criteria
Full size table

The date range for the review was limited to publications between 2000 and 2020, as this represents a time when contemporary opioid stewardship strategies have been developed and are relevant to the current regulatory and policy environment. Over this time there has also been growth in implementation science as a methodological discipline. Full eligibility criteria for inclusion are described in Table 1.

For a study to be included in this review, we used the perspective of the potential for “scale to spread” of the intervention, which was defined as being able to answer the NPT question on coherence and at least two of the three remaining questions. When undertaking screening and data extraction about interventions and implementation strategies, we selected one key question for each of the domains from a list of questions provided by the NPT authors and described below in Table 2 [15].

Table 2 NPT domains and questions used for the assessment [15]
Full size table

2.4 Search strategy

A systematic search strategy to identify interventions that aimed to improve opioid prescribing for pain management in acute hospital settings was developed by the first author (CP) in collaboration with an academic medical librarian. Once the search strategy was finalised, a literature search of MEDLINE/PubMed, EMBASE, Web of Science databases and the Cochrane Library was undertaken by the librarian. (See Appendix 1 for search terms used in each database). The study was registered with the International Prospective Register of Systematic Reviews (PROSPERO) (Reg CRD42020202915).

2.5 Study selection

The database searches were conducted by the librarian in August 2020 and the results were imported to ENDNOTE X7 (Thomson Reuters, New York, NY) and sent to the first author (CP). CP transferred the data to Covidence (Covidence systematic review software, Veritas Health Innovation, Melbourne, Australia) and duplicate entries were eliminated. Reference lists of included full text articles were hand searched to identify any extra relevant literature.

Two researchers (CP and BS) independently performed the abstract and title screening, which was followed by a full text review of shortlisted articles. Where there was disagreement on eligibility for inclusion, further discussion occurred between the two reviewers to achieve consensus, and a third reviewer (PD or LH) was utilised for adjudication where consensus was not achieved.

2.6 Summarising and reporting of results

Similar to previous studies, we used a framework analysis approach for the full text review [17, 20]. During the initial familiarisation process, two researchers (CP and BS) independently assessed if the shortlisted listed articles contained sufficient description to undergo an NPT analysis. We were deliberately inclusive when assessing publications as implementation can be described in several ways, and in varying depth. The first author undertook the indexing, map**, and interpretation of the studies. Indexing and charting consisted of using relevant part of the text found anywhere in the publication that would support the four NPT questions. These assessments were then verified by one or more members of the research team. Where there was disagreement on an assessment, further discussion occurred between the reviewers to achieve consensus, and an additional reviewer provided adjudication where consensus was not achieved.

Relevant outcome measures to be included were, but not limited to, opioid prescribing rates balancing measures patient specific outcomes end process measures. Meta analysis was to be considered if there were sufficient number of studies intervention with similar interventions and outcome measures that would allow appropriate pooling of results.

3 Results

3.1 Included studies

The database searches identified a total of 4031 publications. Following the removal of duplicates, the title and abstracts of 2996 publications were screened for eligibility, with 372 undergoing full text review. Five of the 372 studies meeting the criteria for full-text screening could not be retrieved and were excluded. A total of 13 studies met the inclusion criteria for the final analysis. The main reasons for exclusion during full-text review were insufficient details of the intervention (n = 90/354, 25%) or implementation (n = 215/354, 61%) to undergo an NPT analysis. Figure 1 describes the review process using Prisma 2020 Flow Diagram [24].

Fig. 1
figure 1

Prisma 2020 flow diagram

Full size image

Of the 13 studies included in this review, four originated from acute hospitals in Australia and nine from the United States of America. No studies meeting the inclusion criteria from New Zealand, Canada, or the United Kingdom were found. Interventions were implemented either across the whole of the facility (n = 2) [9, 25] or limited to a specific clinical area, with one intervention in maternity [26], four in emergency [27,28,29,30], and six in surgical settings [31,32,33,34,35,36]. Table 3 provides the characteristics of the intervention and implementation of included studies following NPT analysis.

Table 3 Characteristics of the intervention and implementation of included studies
Full size table

Five of the 13 studies were educational interventions [9, 28, 30, 33, 34], four were guideline implementations [26, 29, 31, 36] and a further two studies [32, 35] were a combination of both (guideline implementation and education). Of the remaining two studies, the interventions were described as implementation of updated pain assessment tools [25] and the use of benchmarking [27].

All studies used education as the principal implementation strategy. This includes one study that used academic detailing [35] and one [29] that required staff to undergo specific training. Clinicians facilitated implementation through their roles as project lead [9], clinical champions [28,29,30, 32, 33, 35, 36], and as opinion leaders [9, 30]. These core implementation strategies were often supported by audit and feedback [9, 26, 27, 29,30,31, 33,34,35,36], benchmarking of opioid prescribing [9, 31] and the use of reminders [26, 29, 30]. Implementation was guided using a formalised quality improvement framework [36], leveraging on existing improvement culture, [27, 32, 35] or by using staged implementation [26, 35, 36].

All studies described resource development, including guideline development [26, 29, 31, 35, 36], patient resources [9, 25, 26, 28, 31, 34,35,36], updating pain assessment tools [25], and electronic medical record (EMR) enhancements [9, 25, 26, 31, 33, 34].

3.2 Coherence (or sense-making): is the intervention easy to describe?

The primary focus of the interventions was to provide a description of the need (patient safety) and value (the evidence) for the proposed change. Five authors provided a more detailed description of intervention, including the content of relevant guidelines and education [26,27,28, 32, 36].

3.3 Cognitive participation (or engagement): are target user groups likely to think it is a good idea?

The majority of studies provided limited detail on cognitive participation. For those that did, they described leveraging existing culture for improvement, utilising small or large group discussions as engagement strategies. In addition, the integration of pain management to education instead of focusing on opioids alone was also described, this would likely have been received positively by medical and nursing staff.

The enabler of clinician influence as an aspect of implementation appeared to be common, with author affiliations of all studies indicating a clinician lead from within the department involved. However, none of the authors explicitly described as an aspect of the implementation.

3.4 Collective action (work done to enable the intervention to happen): is it compatible with existing work practices?

A range of strategies were described that would have likely increased compatibility with existing work practices. This included the direct involvement of clinicians whose practice was affected by the change in the design of the intervention. Implementation strategies that were frequently tailored to suit the busy real-life clinical environment was described across all studies. e.g., short education sessions and utilising “existing” mechanisms to integrate the intervention into practice.

3.5 Reflexive monitoring (formal and informal appraisal): can users/staff contribute feedback about the intervention once it is in use?

Three studies described changes based on end-user feedback, Andereck et al. [27] changed how the prescriber emails were sent, Burgess et al. [26] developed extra resources and Tedesco et al. [36] made guideline modifications. No details on the use of reflexive monitoring were identified in the remaining 10 included studies.

All studies included in this review provided sufficient information for map** to three NPT domains (coherence, cognitive participation and collective action), but reflexive monitoring was poorly described. None of the studies described the use of implementation science and relevant evidence (of implementation) was scattered throughout the publications (e.g., methods, discussion, author affiliations).

4 Discussion

4.1 Principal findings

This is the first systematic review to include an evaluation of how the implementation of opioid stewardship interventions in acute hospital settings is described in the literature.

Only a small number of these studies (n = 13) provided a description of both the intervention and the implementation process, such that they may be potentially replicable in other settings. A significant number (n = 305) of studies were excluded because either the intervention or implementation process was not discussed in sufficient detail.

Despite the many advantages of using implementation science frameworks and theories to systematically describe the “work” done to create change, we did not find implementation science to be explicitly used among the studies included in this review. This could either be because implementation science wasn’t used or because it was used but not described in the resulting reports.

Previous studies have highlighted barriers to using implementation science, which includes a lack of widespread knowledge of the methodologies, confusion arising from the volume of implementation science tools available, and difficulties in describing the detail that allows for replication but will still fit within journal publication requirements [11, 37, 38]. Regardless, there are growing number of publications in the literature describing the use of implementation science including the use of NPT to support implementation and evaluation of healthcare improvement [18, 20, 39].

Most importantly, even though we didn’t find any opioid stewardship studies describing their “work” using an implementation science framework, we did find 13 examples which described implementation in some detail, and these provide insights about how such interventions could be described in future literature.

The complexity of designing and implementing opioid stewardship interventions, due to the need to balance pain management with (opioid) safety, was highlighted across all the studies. We found that the interventions were often multifaceted and tailored to the context into which they were being implemented (e.g., bundles of care with a focus not restricted solely to opioids but to improve overall pain control in obstetric patients, education delivered opportunistically, extensive stakeholder engagement which included patients and public). This is in kee** with current literature highlighting the benefits of a multifaceted approach to healthcare improvement in preference to passive or active strategies alone [40]. However, this further establishes the need for more succinct reporting of core and adaptable components of interventions and implementation strategies to support meaningful replication in other settings.

Education and clinician influence was commonly used implementation strategies, however, in many cases authors didn’t provide sufficient information for us to understand its impact. This was particularly challenging when the intervention was also education-based. From an implementation perspective, authors described the person delivering education, the audience, and the timing of education, but not the impact of these different components. For example, when examining the study by Khorfan [34], the reader may attribute impact of the intervention solely to the educational content, without understanding the influence of the surgeon champion leading the medical education. In terms of sustainability, neither the benefit of using clinician influence to improve acceptability nor the risk to the ongoing success of the intervention if those clinicians leading the change are unable to continue to dedicate time after the initial work of implementation was discussed. Using NPT or another implementation science framework should assist in improving the description and recognition of the role of clinician champions in the successful implementation and ongoing effectiveness of opioid stewardship interventions.

Several authors described integration of the intervention into an existing educational strategy or guideline implementation to maximise sustainability [25, 26, 29,30,31,32,33,34,35,36]. However, they did not acknowledge that practice improvement isn’t solely dependent knowledge acquisition, but also optimising integration into existing workload and workflow. Learnings from programs facing similar challenges are useful to consider here. For example Currie et al. [19] describe the importance of “workflow” in the implementation of guidelines in the antimicrobial stewardship setting. They highlighted that without understanding the real-world environment into which this intervention was being implemented with multiple prescribers being involved in patient care, the review of antimicrobial prescribing ‘fell through the cracks’ [19]. The same principle is highly likely to apply in opioid stewardship where surgeons, anaesthetists, and junior doctors may all have a role in care and prescribing. To address this challenge, identifying, describing and designing interventions to counter the real-world impact of competing priorities affecting decision making at the bedside is vital.

4.2 Strengths and limitations

The strength of this review is the rigor afforded through employing a systematic approach guided by an implementation theory. While other systematic reviews have looked at opioid stewardship interventions [7] and persistent use after hospital admissions [6], to our knowledge this is the first to provide a focus on the implementation of such interventions. This offers the benefit too of providing information and knowledge about implementation reporting in healthcare interventions which may be of interest and benefit in settings beyond opioid stewardship in the acute hospital.

This review had several limitations, partly driven by the characteristics of the identified literature. Firstly, many studies that may have had an impact on opioid prescribing were excluded (n = 305) after full text review as they did not provide sufficient details of the intervention and/or an implementation strategy. Secondly, although measurable improvements to either opioid prescribing rates and/or other outcome measures were seen for all studies, a meta-analysis was not feasible due to the diversity of measurements and evaluation approaches used. For example, prescribing rates, when measured, used a variety of metrics, making a direct comparison between the studies difficult. In addition, because of the lack of structured implementation reporting and the diversity of the included studies which described any implementation, it was not possible to make any comparison about which aspects of implementation were beneficial or important to the outcomes.

Although a comprehensive search strategy was employed and reference lists were manually searched to include all relevant studies, some relevant studies may have been inadvertently omitted. Grey literature was not searched, thus potentially introducing a level of publication bias. Lastly, using a framework other than NPT may have resulted in other studies being included or excluded or having the included studies described differently.

4.3 Implications of the findings for future directions

In kee** with observations by Luoto et al. [41] we found the quality of reporting on implementation to be generally low. In particular, the description of specific characteristics of interventions, and implementation identified as being important in previous studies (particularly for scaling and spreading), was lacking [16, 42, 43]. This review highlights the opportunity for those sha** the future opioid stewardship literature to better facilitate the translation of research into practice through considered, holistic description of both the intervention and implementation process. It also highlights value of NPT as an explanatory framework which could be used for this purpose.

Using and describing implementation in a structured fashion would help clinicians and researchers to identify barriers and enablers and explore factors vital not only for the initial success but also for scalability and sustainability of initiatives. It is the hope that this review will provide a nudge to encourage future authors to carefully consider how they describe implementation of opioid stewardship activities. If we begin appraising opioid steward intervention studies on the sufficiency of detail provided to permit sustainability and replicability—i.e., the potential for real long term and/or widespread change—perhaps the quality of reporting will begin to improve and provision of such detail will itself become normalised.

5 Conclusion

All the studies included in this review reported on opioid stewardship interventions in an acute hospital setting, however, there was variability in the description of implementation processes, potentially limiting the transferability of study findings and interventions to other settings. Implementation science emphasises that complex intervention development should be an iterative process; similarly, the process of normalising implementation science in the opioid prescribing literature will require widespread and sustained effort. This review provides the opportunity for future opioid stewardship and healthcare literature more broadly to consider how to describe both the intervention and implementation processes more holistically to best enable the translation of research into practice.