A Systematic Review Protocol to Determine the Most Effective Strategies to Reduce Computed Tomography Usage in the Emergency Department

This study describes the protocol for a systematic review and meta-analysis. The primary objective of the review is to identify experimental studies assessing the effectiveness of interventions that aim to reduce the proportion of computed tomography (CT) in emergency departments (EDs). Data permitting, our secondary objectives will be to assess the impact of reduction in CT utilization on the length of stay, admission to hospital, and uptake/satisfaction with the intervention. When available, balancing measures such as readmission to hospital or ED revisit rates will be included. Pre-defined subgroup analyses include patient populations (adult or pediatric), type of ED, and the nature of the intervention. Through this review, the research team aims to inform knowledge translation initiatives aimed at lowering CT usage in the ED by identifying the most effective interventions to safely improve CT resource stewardship.


Introduction Description of the issue
Computed tomography (CT) is an important diagnostic tool used in the assessment of patients in the Emergency Department (ED). However, emerging evidence has shown dramatic escalations in CT utilization for adult ED patients, with reported annual increases of 6.0%-13.8% [1,2]. While CT is an important diagnostic imaging modality, it is also associated with radiation exposure, increased health system costs, longer ED lengths of stay, and impaired flow of patients through the department [3]. Data suggests that national CT utilization for ED patients is increasing at rates exceeding the increase in patient volumes. Some of these CTs are likely avoidable. As an example, one study looking at low-risk headaches suggested that more than 38% of patients may be imaged unnecessarily [4]. If the increasing use of CT was associated with improved patient-oriented outcomes, the risks may be worth the benefits but that is not always the case. As an example, the use of CT pulmonary angiograms to diagnose pulmonary emboli (PE) has been steadily rising but the rates of massive PEs and the mortality rate due to them have been on the decline [5][6][7].
Considering the risks of CT imaging, reducing the rate of non-urgent CTs is an important target for improving care and, as a result, has been identified as a priority in many health systems. A number of studies have attempted to develop clinical decision rules (CDRs) to safely forego advanced imaging for patients with certain clinical conditions. However, there is a fundamental difference between developing a sensitive and well-validated CDR and successfully implementing one to change clinical practice. Despite having validated CDRs for head and neck trauma, acute headache, and pulmonary embolism, as well as potential alternatives to CT imaging for many presentations, these studies have clearly not translated clinically to a reduction in the overall CT usage. Therefore, this review will focus on the knowledge translation aspect of practically implementing interventions that will lead to a reduction in ED CT utilization.

Description of the existing intervention data
Interventions aimed at decreasing unnecessary CT utilization can be grouped into three broad categories: pre-ordering, at the time of ordering, and post-imaging feedback [8][9][10]. Assessing these strategies and their effectiveness at reducing imaging is a popular topic in the literature with several systematic reviews having been published. However, this literature frequently focuses on specific pathologies (e.g., CT use in acute low-back pain patients), includes multiple imaging modalities, or reviews the incidence of low-value testing in general without addressing potential solutions [7,[10][11][12][13][14][15].
Furthermore, the data that can be extrapolated to the ED setting from these prior reviews is limited. Of the current knowledge translation literature, one review aimed at reducing CT use in low-back pain found targeted reminders of indications and a modified referral form with decision support reduced CT usage by 36.8% and 22.5%, respectively. The same study found no benefit from physician audits or feedback tools, education sessions, or guidelines dissemination [14]. However, this study included both outpatients and all hospitalized patients with acute and chronic pain, so its generalizability to the ED setting is limited. In contrast, a review in the pediatric population found that patient/family education was the single most effective intervention with a relative reduction of 57.9%, and that multi-faceted interventions were more effective than single focused ones [15].

Rationale
A review of the literature demonstrates that while there have been several individual and heterogeneous studies attempting to reduce CT utilization, there has not been an effective review to aggregate this data and make generalizable conclusions to guide quality improvement efforts. The limited data available is not specific to the ED population and presents contradictory evidence. This suggests that both the patient population and the clinical setting likely impact the effectiveness of interventions.
Through this systematic review, the authors aim to identify and summarize the literature available regarding the effectiveness of various interventions to reduce CT usage in the ED. We will also explore factors that predict the effectiveness of interventions based on a priori defined subgroups (e.g., the nature of the intervention, whether the intervention was co-designed by providers, whether it was multi-faceted, etc.). By identifying those interventions that have had the most significant impacts, targeted strategies can be designed to ensure physicians are ordering CTs in such a way that maintains timely access to CT while ensuring patient safety and resource stewardship.

Objectives
Through this systematic review, the authors' primary objective is to identify and summarize the literature available on the effectiveness of various interventions to reduce CT usage in the ED.
When data is available, we will assess a variety of secondary outcomes that include the association between reduction in CT usage and ED length of stay, admission to hospital, and satisfaction/uptake of the intervention. Additionally, balancing measures such as readmission to hospital or ED revisit rates will be captured when the data is presented in individual studies.
Secondary objectives will include the exploration of predictors of effective interventions based on a priori defined subgroups: target population (adult versus paediatric, or both), study context (academic versus community centre, country of the study), nature of the intervention (implementation of clinical decision rule, education-only focused, policy, or workflow changes), and single versus multi-faceted intervention.

Study inclusion factors
The various factors taken into consideration for the inclusion of studies are described in Table  1.

Participants All ED patients undergoing CT (all modalities). Adult and pediatric patients will be included
Intervention ED-based intervention to reduce low-value CT utilization  Our search methodology was developed and piloted by a librarian/information specialist (DKL) in partnership with the broader review team and peer-reviewed by another librarian.
To ensure the strategy identified relevant articles, a short literature review was conducted a priori (EL) that included several notable studies in this area. The results of the strategy were checked to ensure all studies were captured in our search.
The search strategy was then adapted for several other databases, including Embase (Ovid), CINAHL (EBSCO), and Cochrane Central Register of Controlled Trials (CENTRAL). Appendix 1 outlines the search strategy used for MEDLINE (Ovid).

Grey Literature
Grey literature search strategies included ProQuest Dissertation & Theses Global™, conference proceedings, Google Scholar web search (including the first 10 pages of results), contacting experts in the field for unpublished and ongoing trials, and contacting study authors for additional information and unpublished results, if necessary. Additionally, the bibliography of the searches retrieved from the electronic search were reviewed and forward reference searching of cited articles was also conducted.

Study selection and quality assessment
All identified studies will be imported into Covidence systematic review software (Veritas Health Innovation Ltd, Melbourne). A standardized tool ( Figure 1) has been developed and piloted by the authors and will be used to aid in the selection of studies. Each title and abstract will be screened by two independent reviewers to identify potential articles and remove any duplicates. Following this, a full text review of applicable studies will be conducted by two independent reviewers who will apply the eligibility criteria to each study using a standardized form. In cases of disagreement, a third reviewer will adjudicate and reach a final decision on inclusion. This strategy will be used to limit the risk of selection bias.
Any excluded articles will have a reason captured and reported. This information will be presented in the final report in the form of a Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) diagram [16].
Two independent reviewers will also conduct a formal risk of bias assessment. The tool used will be based on the nature of the study (i.e., randomized or non-randomized/observational); however, either will be validated and supported by the Cochrane Collaboration [17][18]. Any discrepancies in risk of bias assessment will be decided upon by a third reviewer. We will be employing the Grading of Recommendations Assessment, Development and Evaluation (GRADE) methodology for certainty of evidence relating to the primary outcome where applicable.

Data extraction and analysis
Data extraction will occur by two separate investigators in duplicate using a standardized data extraction form. A third member of the team will adjudicate any discrepancies.
Descriptive statistics and narrative analysis will be used to summarize the data. Heterogeneity of the data (for the primary and secondary outcomes) will be measured by I 2 while publication bias will be assessed using a funnel plot if more than 10 studies are included in the analysis. When possible, we will pool the results and undertake a meta-analysis utilizing a random effects model using the Review Manager Version 5.3 (Nordic Cochrane Centre, Cochrane Collaboration, Copenhagen). Data will be separated by study design with dichotomous data reported using risk ratios and continuous data reported as mean difference, both with 95% confidence intervals.

Subgroup and Sensitivity Analysis
A priori subgroups have been defined. These include target population (adult versus paediatric, or both), study context (academic versus community centre, country of the study), nature of the intervention (Implementation of clinical decision rule, education-only focused, policy, or workflow changes), and single versus multi-faceted intervention. The statistical approach for subgroups will be the same as for the primary outcome.
The effect of methodological design on summary estimates of sensitivity will be assessed. If enough studies are included, sensitivity analysis will consider whether studies were design for research or quality improvement.

Systematic versus scoping review
Scoping reviews can be useful for identifying current and emerging evidence in a topic area, while systematic reviews seek to rigorously appraise and synthesize the evidence base on a highly specific question to inform policy, practice, and further research [19]. While this is a broad topic in its nature, and simply identifying what strategies exist in the current literature may lend itself to a scoping review design, emergency medicine practitioners are already aware of the multitude of strategies employed by ED leadership, hospital administration, and government to limit the overuse of imaging. The implementation of these strategies, however, has resulted in mixed results on the their effectiveness. For example, efforts exist to reduce low-value CT imaging for specific conditions such as low back pain and cervical spine injuries, and there are a number of clinical decision rules for CT use in the ED [20][21][22]. However, these studies have reported mixed results in terms of impact, and there is a lack of clarity about which specific implementation strategies are most effective for achieving reductions in lowvalue CT imaging.
As researchers specifically interested in knowledge translation, our objective is to identify the strategies that are most effective for achieving clinical practice change in the ED. A systematic review will allow us to not only identify which strategies have been studied, but may also allow us to determine, with our sensitivity analysis, what setting and what conditions may contribute to the effectiveness of interventions.

Potential impact for emergency medicine
Resource stewardship and improving resource usage has been a focus of many campaigns over the past decade. Choosing Wisely Canada (CWC) and the Canadian Association of Emergency Physicians (CAEP) have partnered to create 10 recommendations where this may be possible in the ED. Six of these directly focus on reducing unnecessary image ordering [23]. Similarly, the American College of Emergency Physicians (ACEP) has released a position statement on the topic [24]. This highlights the need for improvement in how we decide to use decision supports and other tools that have already been validated and readily disseminated.
Safely decreasing CT usage in the ED will have an important impact on our patients, our department, and our health care system. From a system accessibility perspective, reducing lowvalue CTs will increase ED capacity as patients undergoing CT experience increased ED lengthof-stays and inhibit access to the CT for other ED patients who have an urgent/emergent indication. CWC states that as many as 30% of tests, treatments, and procedures in Canada are potentially unnecessary, and that reducing inappropriate tests can save money, reduce wait times, and improve patient experience [25]. From a patient perspective, reducing CT utilization will decrease their length of stay in the ED, decrease their exposure to ionizing radiation, and reduce the risk of harm from intervention/investigation of incidental findings (potentially leading to over-diagnosis and over-treatment). Finally, given the current constraints of the health care system, reducing health care costs is an important consideration, especially in light of the aforementioned benefits.
However, more data is needed to help translate successes from narrowly targeted studies to interventions that are globally effective. Table 2 identifies several of the strategies that have previously been suggested as interventions to achieve the goal of CT utilization reduction. This review aims to further inform these strategies as the identification and assessment of the available literature is completed.  Interventions that have been implemented in an effort to reduce unnecessary CT usage [5][6][7].

Conclusions
This report describes the systematic review and potential meta-analysis protocol that will be utilized to determine and assess the evidence available regarding the effectiveness of various interventions at reducing CT usage in emergency departments. The results of this review will help inform ED leadership, hospital administration, and governments looking to balance patient safety with resource stewardship as they implement strategies to reduce the proportion of low-value CTs.

Additional Information Disclosures
Human subjects: All authors have confirmed that this study did not involve human participants or tissue. Animal subjects: All authors have confirmed that this study did not involve animal subjects or tissue. Conflicts of interest: In compliance with the ICMJE uniform disclosure form, all authors declare the following: Payment/services info: The project received funding from Alberta Health Services' Emergency Strategic Clinical Network as part of their 2020 systematic review grant competition (https://www.albertahealthservices.ca/scns/Page13008.aspx). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Financial relationships: All authors have declared that they have no financial relationships at present or within the previous three years with any organizations that might have an interest in the submitted work. Other relationships: All authors have declared that there are no other relationships or activities that could appear to have influenced the submitted work.