Article Text
PDF
Abstract
Background Overuse of lumbar imaging is common in the emergency department (ED). Few trials have examined interventions to address this. We evaluated the effectiveness of a multifaceted intervention to implement guideline recommendations for low back pain in the emergency department.
Methods We conducted a stepped-wedge, cluster-randomised trial in four EDs in New South Wales, Australia. After a 13-month control phase of usual care, the EDs received a multifaceted intervention to support guideline-endorsed care in a random order, based on a computer-generated random sequence, every 4 weeks over a 4-month period. All sites were followed up for at least 3 months. The primary outcome was the proportion of low back pain presentations receiving lumbar imaging. Secondary healthcare utilisation outcomes included prescriptions of opioid and non-opioid pain medicines, inpatient admissions, length of ED stay, specialist referrals and re-presentations. Clinician beliefs and knowledge about low back pain care were measured before and after the intervention. Patient-reported pain, disability, quality of life and satisfaction were measured at 1, 2 and 4 weeks post ED presentation.
Results A total of 269 ED clinicians and 4625 episodes of care for low back pain (4491 patients) were included. The data did not provide clear evidence that the intervention reduced lumbar imaging (OR 0.77; 95% CI 0.47 to 1.26; p=0.29). It did reduce opioid use (OR 0.57; 95% CI 0.38 to 0.85; p=0.006) and improved clinicians’ beliefs (mean difference (MD), 2.85; 95% CI 1.85 to 3.85; p<0.001; on a scale from 9 to 45) and knowledge about low back pain care (MD, 0.48; 95% CI 0.13 to 0.83; p<0.01; on a scale from 0 to 11). There was no difference in pain scores at 1-week follow-up (MD, 0.04; 95% CI −1.00 to 1.08; p=0.94; on a scale from 0 to 10). A similar trend was observed for all other patient-reported outcomes and time points. This study found no effect on the other secondary healthcare utilisation outcomes.
Conclusion It is uncertain if a multifaceted intervention to implement guideline recommendations for low back pain care decreased lumbar imaging in the ED; however, it did reduce opioid prescriptions without adversely affecting patient outcomes.
Trial registration number ACTRN12617001160325.
- emergency department
- implementation science
- continuing education
- continuing professional development
Data availability statement
No data are available. Due to information governance restrictions imposed by organisations governing data access, we are unable to share the trial data unless applicants secure the relevant permissions.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
Introduction
The huge problem of low back pain has not yet been adequately addressed, as it continues to hold the lead for burden of disease globally.1 It is also a leading reason for presentation to the emergency department in the United States, with over four million presentations per year.2 Imaging, opioid medicines and hospital admissions are not recommended in guidelines for most cases of low back pain.3 Only about 5% of patients with low back pain who present to emergency departments have serious spinal pathology requiring urgent medical care.4 However, in the emergency department, about one-third of patients with low back pain are imaged,5 two-thirds receive an opioid medicine4 6 and up to one-third are admitted to hospital.4 7
These variations from best practice are wasteful of scarce health resources, lack evidence of benefits and can lead to patient harm.8–10 Imaging in low back pain is costly, in most cases does not improve outcomes and may lead to unnecessary invasive treatments.11 12 In some cases imaging can be considered iatrogenic and has been associated with increased disability in the future.11 Opioids commenced in the emergency department can be continued for prolonged periods13 and contribute to the opioid crisis.14 Admissions for low back pain are prevalent in Australia, with 664 per 100 000 population being admitted with low back pain in 2017–201815 even though hospital admissions incur significant economic cost and many patients could be adequately followed up as outpatients. The Lancet low back pain series outlined the urgent need for action to reduce these harmful practices for low back pain.9 As the initial site of care and the hub of acute diagnostic and treatment modalities, the emergency department is well positioned to address these issues.
There have been a number of evidence-based guidelines and care models developed to improve care for low back pain worldwide.3 16 At present, it is unclear if active implementation of these guidelines improves healthcare so that it better aligns with the evidence, without adversely affecting patient outcomes. In this trial, we aimed to test the effectiveness of a multifaceted intervention to implement a guideline-based model of care for low back pain in the emergency department, by reducing harmful practices, improving clinician knowledge and not adversely affecting patient outcomes.
Methods
Study design and participants
The Sydney Health Partners Emergency Department (SHaPED) trial was an investigator-initiated, multicentre, pragmatic, stepped-wedge, cluster-randomised trial to evaluate the implementation of an evidence-based model of care for low back pain in emergency departments. 17 The SHaPED trial was coordinated by a steering committee and was funded by Sydney Health Partners and the New South Wales (NSW) Agency for Clinical Innovation. FlexEze Australia supplied the heat wraps used in the trial at no cost.
Sydney Local Health District hospitals and an affiliated rural hospital in NSW, Australia were eligible provided they had an emergency service and used the same electronic medical record system (Cerner PowerChart). These hospitals provide diversity with respect to patient ethnicity, demographic characteristics, hospital setting and clinical staffing models. We worked with local clinicians to develop a dashboard integrated into the electronic medical record system using Qlik Sense to quantitatively explore healthcare utilisation practices, such as lumbar imaging, pain medications and laboratory tests, and examine the problem locally. We used these data, as well as discussion with local clinicians, including rheumatologists, physiotherapists and emergency physicians, to determine whether active implementation of the model of care would be beneficial for quality improvement at each site. If so, the emergency department directors agreed to have their site participate in the study. Although these hospitals have a medical rotation programme for resident medical officers and registrars working in the emergency department, the trial’s start date was planned to avoid contamination arising from rotation of medical staff during the intervention period. However, there may have been a small number of clinicians that worked at two sites simultaneously. After a 13-month control phase of usual care, each emergency department (cluster) randomly transitioned to the intervention phase every 4 weeks (‘step’) and were followed up for at least 3 months, for a total of 20 months (periods).
Participants were emergency department clinicians (physicians, nurses and physiotherapists) that manage patients with low back pain. The criteria to include low back pain presentations were age of 18 years or older with a diagnosis code of non-specific low back pain or radicular syndromes (ie, presentations that were given a diagnosis code associated with serious spinal conditions such as fracture, spinal infection, cauda equina syndrome, malignancy or trauma as a cause were excluded). We used all relevant diagnosis codes from the Systematized Nomenclature of Medicine-Clinical Terms-Australian Version-Emergency Department Reference Set to identify eligible low back pain presentations through the electronic medical records (diagnosis codes are described in online supplemental appendix 1).18 If a patient re-presented within 48 hours, only the first presentation was included.
Supplemental material
Emergency department clinicians and patients were invited to participate in this study and provided informed consent; electronic medical record data were analysed without individual patient consent in accordance with the Health Records and Information Privacy Act 2002 (NSW). The authors vouch for the accuracy and completeness of the data presented. We followed the Consolidated Standards of Reporting Trials extension for stepped-wedge, cluster-randomised trials19 and published the trial protocol.20
Randomisation and masking
In this trial, the hospital site was the unit of randomisation. We collected data retrospectively from 1 July 2017 to 30 June 2018, and prospectively from 1 July 2018 to 31 July 2018 for the control phase of usual care (periods 1–13). Then emergency departments crossed from the control phase to the intervention phase in a randomised order, with a new emergency department receiving the intervention every 4 weeks. The randomisation schedule, based on a computer-generated random sequence, was generated by a statistician blinded to the sites. This timing was concealed from clusters and site investigators until 2 months before the implementation date. It was not possible to blind the clinician participants, patients or investigators, but outcome measures were extracted from electronic medical records by the data provider of each hospital independently of the research team. The intervention phase began on 1 August 2018 and ended on 16 November 2018 (periods 14–17). The follow-up phase ended on 17 February 2019 (periods 18–20). The core trial period was defined as periods 13–18 (ie, between 1 July 2018 and 16 December 2018).
Procedure
The intervention was the implementation of the guideline-based NSW Agency for Clinical Innovation model of care for acute low back pain16 using a multifaceted intervention strategy. The model of care was developed for primary care and the emergency department setting in collaboration with policymakers, clinicians, consumers and researchers and distils the high-quality evidence in this area to formulate key messages for practice. The key messages in the model are the following: (1) patients with non-serious low back pain do not require lumbar imaging; (2) where medicines are used, simple analgesics should be the first option; and (3) patients with non-serious low back pain should be managed as outpatients. The implementation intervention was guided by the Knowledge-to-Action framework21 and incorporated evidence-based implementation strategies for targeting the behaviour of healthcare professionals.22–25 We undertook barrier analyses, mapped locally developed treatment processes for low back pain, and met with local opinion leaders, key clinicians and patients to design an intervention to address identified barriers to uptake of recommendations in the model of care.
The multifaceted intervention was clinician-targeted. The clinicians included medical officers, nursing staff and physiotherapists that worked in the emergency departments. The intervention included five main components:
Education seminars: structured training from experienced rheumatologists and physiotherapists that focused on skills for assessing, managing, educating and referring patients according to the Agency for Clinical Innovation model of care for acute low back pain. These seminars were run on numerous occasions throughout the 4-week intervention period in protected teaching time, either in neighbouring teaching rooms or in the emergency department itself. Clinician participation in the education sessions was tracked through a logbook.
Educational materials: a hard copy of the model of care, a website and decision support tools for appropriate use of lumbar imaging and analgesic medicines were distributed to clinicians. Posters highlighting key messages about benefits and harms of lumbar imaging, opioid medicines and inpatient admission were displayed throughout the emergency departments and patient handouts were provided so that clinicians could use them to more easily educate patients.
Provision of non-opioid pain management strategies: non-opioid pain medicines and heat wraps were made more easily accessible to clinicians as an evidence-based alternative to opioid medicines.
Fast-track referral to outpatient services: clinicians were educated on referral pathways options to outpatient services such as specialist back clinics and physiotherapy follow-up.
Audit and feedback: clinicians were provided with structured real-time audit and feedback data on department-level imaging, opioid and inpatient admission rates through monthly email newsletters and real-time dashboard developed in Qlik Sense.
Further details on the intervention are provided in online supplemental appendix 2. Eligible low back pain presentations to the emergency departments during the control phase received care as usual.
Outcomes
We measured the effects of the multifaceted intervention with respect to three outcome domains: healthcare utilisation, clinician-reported outcomes and patient-reported outcomes.
The primary healthcare utilisation outcome was the proportion of low back pain presentations referred for any lumbar imaging in the emergency department. Secondary healthcare utilisation outcomes were advanced lumbar imaging (ie, CT or MRI), prescription of any opioid medicine, prescription of any strong opioid medicine (ie, buprenorphine, fentanyl, hydromorphone, morphine, oxycodone, pethidine or tapentadol), prescription of any non-opioid pain medicine, length of emergency department stay, consultation by a specialist in the emergency department, admission to hospital ward, admission to short stay unit and re-presentation to the emergency department within 48 hours. In the published protocol,20 we grouped pain medicines according to the Anatomical Therapeutic Chemical classification (ie, paracetamol, non-steroidal anti-inflammatory drugs, muscle relaxants, opioids and neuropathic pain medicines). In the statistical analysis plan,26 however, opioid medicines were analysed separately and non-opioid pain medicines were grouped to better reflect our aims stated in the protocol.20 We were unable to measure readmissions to hospital within 28 days, although this was a prespecified secondary outcome in the published protocol.20
Clinician-reported outcomes were beliefs about low back pain measured using the Back Beliefs Questionnaire (scores range from 9 to 45, with higher scores indicating more positive beliefs),27 and knowledge and attitudes about low back pain management using the Knowledge and Attitudes Questionnaire—a set of 11 questions of which the correct answers are based on current evidence and expert opinion, with higher scores indicating more accurate knowledge and positive attitudes.28 A mean difference (MD) greater than 2 points in the Back Beliefs Questionnaire has been considered clinically important in a population-based study,29 but the equivalent difference for the Knowledge and Attitudes Questionnaire has not been established. These questionnaires and the scoring criteria are presented in online supplemental appendices 3 and 4.
Patient-reported outcomes were back pain intensity measured on a 0–10 Numeric Pain Rating Scale, with 0 indicating no pain and 10 the worst possible pain; disability as measured by PROMIS Physical Function-4a (scale range from 4 to 20, with higher scores indicating better function); quality of life as measured by PROMIS Global Health-item 1 (scale range from 1 to 5, with higher scores indicating better quality of life); and satisfaction with care as measured by the Emergency Department Patient Experiences with Care Survey-item 31 (scale range 0–10, with 0 indicating worst care possible and 10 the best care possible).30
Healthcare utilisation outcome data were extracted from each hospital’s electronic medical record. At two sites where clinicians still used paper-based medication charts, pain medicine data were collected manually. Socioeconomic status was defined using patients’ postcodes and the Socio-Economic Indexes for Areas, developed by the Australian Bureau of Statistics. Clinician-reported outcome data were collected by questionnaire before and after the intervention. We collected patient-reported outcome data using an automated text message and online survey system supplemented by telephone calls at 1, 2 and 4 weeks after emergency department discharge for patients who consented to participate during the core trial period (ie, periods 13–18).
Statistical analysis
The statistical analysis plan was published in Open Science Framework Preprints after completion of the trial but prior to accessing the data.26 We estimated that a total number of 1920 eligible low back pain presentations would provide the trial with 80% power to detect a 10% absolute difference for the primary outcome of lumbar imaging referrals (a difference in imaging frequency considered to be worth detecting), assuming an alpha of 0.05 and an intraclass correlation coefficient (ICC) of 0.1. Assuming an SD of 2.5 and an ICC of 0.05, 600 patients would provide 80% power to detect a non-inferiority difference of 1 point in back pain intensity on a 0–10 point scale (a difference in pain intensity deemed to be clinically important). The decision for a non-inferiority analysis of patient-reported outcomes (prespecified in the statistical analysis plan26 but not in the published protocol20) was based on the rationale that implementation of the model of care should reduce the use of low-value care but not worsen health outcomes or satisfaction with care.
Data analyses were performed according to an intention-to-treat approach. All tests were two-sided with alpha set at 0.05. We conducted logistic (binary outcomes) and linear (continuous outcomes) regression analyses with a random effect for cluster (emergency department), a random effect for clinician nested within cluster, a fixed effect indicating the group assignment of each cluster at each step and a fixed effect of time. We investigated underlying temporal trends during the control phase (ie, periods 1–13) for all healthcare utilisation outcomes using visual inspection and by fitting a regression line to assess the significance of time slopes. Data collected during the intervention period were included in the intervention group, although we prespecified in the statistical analysis plan that these data would be excluded from the primary analysis.26 Prespecified sensitivity analyses were conducted by adjusting the main model for potential confounders (age, gender, diagnosis, presenting day, mode of arrival and triage category) and by analysing data from the core trial period only (ie, periods 13–18). A prespecified subgroup analysis for the primary outcome was conducted by age, socioeconomic status, triage category, diagnosis and mode of arrival. A post-hoc sensitivity analysis for healthcare utilisation outcome data was conducted by omitting repeat visits to the emergency department by the same patient during the trial period (ie, analysis of 4491 unique patients). Post-hoc subgroup analyses were conducted by site and for any opioid prescribing. The effects of the intervention are presented as OR or MD with 95% CI. Analyses were conducted in SAS V.9.4 software. This trial was registered with the Australian New Zealand Clinical Trials Registry.
Patient and public involvement
This study was completed with consumer involvement in study design, specifically with the patient-reported outcome measures. A consumer advisory group from the Australia & New Zealand Musculoskeletal Clinical Trials Network reviewed the protocol and gave feedback on the content of the intervention, the types of patient-reported outcome measures and time points for collecting the outcomes. Patients or the public were not invited to contribute to the writing or editing of this document for readability or accuracy.
Results
The trial profile is shown in figure 1. Four eligible emergency departments were randomised—one cluster did not have emergency services and was excluded. Between 1 August 2018 and 16 November 2018, all clinicians present in the participating emergency department during the multiple education session times that provided care for low back pain were invited to participate. This was estimated at 300 eligible emergency department clinicians. We estimated this retrospectively based on the number of individual clinicians that were recorded to have managed low back pain presentations during the intervention period. There were 269 clinicians enrolled in the trial. All 269 enrolled clinicians received the multifaceted intervention and completed the Back Beliefs Questionnaire and the Knowledge and Attitudes Questionnaire prior to receiving the intervention. The baseline characteristics of clinician participants are shown in table 1. Only 139 clinicians completed the Back Beliefs Questionnaire and the Knowledge and Attitudes Questionnaire after the intervention period. The characteristics of the clinicians who completed the follow-up questionnaires (n=139) and clinicians who did not (n=130) were generally similar, with the exception of sex, main profession and mean number of patients seen per week. Female clinicians, nurses and those seeing more patients per week were more likely to complete the follow-up questionnaire (differences in clinicians’ characteristics are described in online supplemental appendix 5).
Study flowchart.
Baseline characteristics of clinicians
Between 1 July 2017 and 17 February 2019, there were 5041 episodes of care for low back pain in the four emergency departments and 4625 eligible episodes (4491 unique patients) were included in the analysis of healthcare utilisation outcomes: 3233 in the usual care group and 1392 in the intervention group. Overall, the two groups were balanced with respect to baseline characteristics (table 2). The baseline characteristics of the low back pain presentations per emergency department are shown in online supplemental appendix 6. From 1 July 2018 to 16 December 2018 (ie, core trial periods 13–18), all patients who presented to the site emergency departments with diagnosis codes representing non-specific low back pain or radicular syndromes, who had mobile phone numbers (807 eligible patients), were invited to participate in the patient-reported outcome measure survey. A total of 425 patients consented to participate in the survey and were enrolled at week 1; 9 dropped out without reasons and 416 were included in the analysis of patient-reported outcomes. Of the 416 responders, 360 completed the follow-up survey at week 2 (86.5%) and 351 at week 4 (84.4%).
Baseline characteristics of low back pain presentations
The regression analysis to assess the significance of time slopes revealed no underlying time trends in any healthcare utilisation outcome. This is represented in table 3 through the values listed for ‘p for trend’. The results for the primary outcome of any lumbar imaging and other secondary healthcare utilisation outcomes are summarised in table 3. Our best estimate is that the intervention reduced the odds of lumbar imaging (OR=0.77), but the evidence is uncertain with a 95% CI of 0.47 to 1.26. The study found the intervention reduced the use of opioid medicines in the emergency department with an absolute reduction of 12.3%, from 62.8% to 50.5% of low back pain presentations (OR 0.57, 95% CI 0.38 to 0.85). The data did not provide clear evidence that the interventions had an effect on other secondary healthcare utilisation outcomes, including referrals for advanced lumbar imaging (OR 1.16, 95% CI 0.57 to 2.35), prescription of strong opioid medicines (OR 0.69, 95% CI 0.46 to 1.04), prescription of non-opioid pain medicines (OR 1.52, 95% CI 0.98 to 2.35), length of emergency department stay (MD –0.28, 95% CI –0.84 to 0.28), specialist consultations (OR 2.50, 95% CI 0.98 to 6.37), admission to hospital (OR 0.96, 95% CI 0.54 to 1.71), admission to short stay unit (OR 1.99, 95% CI 0.91 to 4.37) and re-presentation within 48 hours (OR 0.31, 95% CI 0.06 to 1.57).
Healthcare utilisation outcomes
The study showed that the intervention improved the accuracy of clinicians’ beliefs and knowledge regarding low back pain and its management (table 4). After the intervention, the mean Back Beliefs Questionnaire score increased from 32.1 to 35.0 (MD 2.85, 95% CI 1.85 to 3.85) and the mean number of correct answers in the Knowledge and Attitudes Questionnaire also increased from 7.1 to 7.6 (MD 0.48, 95% CI 0.13 to 0.83). Patient-reported outcomes were not adversely affected by the intervention (table 5). At 1 week after emergency department discharge, there was no difference in pain intensity (MD 0.04, 95% CI –1.00 to 1.08), physical function (MD 0.96, 95% CI –0.92 to 2.83), quality of life (MD 0.17, 95% CI –0.25 to 0.58) or patient satisfaction with emergency care (MD 0.16, 95% CI –0.72 to 1.03) between the intervention and usual care groups. We observed similar results at 2 and 4 weeks following emergency department discharge (table 5).
Clinician-reported outcomes
Patient-reported outcomes
The estimated effects of the intervention remained consistent in the prespecified sensitivity analyses (online supplemental appendices 7 and 8). Adjusting the main logistic regression model for potential confounders did not alter the effect of the intervention on the primary outcome of lumbar imaging referrals (OR 0.86, 95% CI 0.52 to 1.43) and continued to show significantly fewer opioid prescriptions after the intervention (OR 0.60, 95% CI 0.39 to 0.90), but revealed a significant increase in referrals for specialist consultations in the emergency department (OR 3.11, 95% CI 1.17 to 8.31) (online supplemental appendix 7). Restricting the analysis to the core trial period (ie, periods 13–18) also showed a reduction in opioid prescriptions in the emergency department (OR 0.61, 95% CI 0.39 to 0.97) and did not alter any other results (online supplemental appendix 8). A post-hoc sensitivity analysis omitting data from repeat visits by the same patient during the trial period (ie, analysis of 4491 unique patients) also showed similar results (see online supplemental appendix 9).
The prespecified subgroup analysis for the primary outcome of lumbar imaging referrals showed no interactions (online supplemental appendix 10). However, a post-hoc subgroup analysis revealed intervention-by-site interactions (p=0.0004) for use of any opioid in the emergency department (see online supplemental appendix 11). While the overall OR for use of any opioid was 0.57 (95% CI 0.38 to 0.85), the estimates at the four emergency departments were 0.46 (95% CI 0.30 to 0.73), 0.66 (95% CI 0.42 to 1.04), 0.40 (95% CI 0.24 to 0.66) and 1.09 (95% CI 0.61 to 1.93). The number and proportion of low back pain presentations receiving any lumbar imaging, any opioid medicine and admitted to hospital per period per emergency department are presented in online supplemental appendices 12, 13 and 14, respectively.
Discussion
Statement of principal findings
This trial showed that there was no clear evidence that a multifaceted intervention targeting emergency department clinicians reduced rates of lumbar imaging. Similarly, there was no evidence of a reduction in prescriptions of non-opioid pain medicines, admissions, length of stay, specialist referrals and re-presentations. However, the data showed the intervention reduced use of opioids by 12.3% (from 62.8% to 50.5% of presentations) in the emergency department and improved the accuracy of clinicians’ beliefs and knowledge. There were no adverse effects on patient-reported outcomes, such as pain and satisfaction with care despite the lower use of opioid medicines.
Strengths and limitations of this study
This trial had several strengths. First, the use of routinely collected data extracted from the electronic medical record allowed us to efficiently collect healthcare data for 4625 episodes of care for low back pain and give timely clinical feedback of outcomes to the participating clinicians as part of the multifaceted intervention strategy. Second, we had negligible loss to follow-up in the healthcare and patient-reported outcomes, minimising the risk of bias. Finally, the use of a stepped-wedge, cluster-randomised design allowed us to pragmatically implement the care model across all sites and test the intervention in a real-world setting.
This trial had some limitations. SHaPED was conducted in four public hospital sites in NSW, Australia and so generalisability to other sites in Australia, including private hospital settings and other locations with different healthcare systems and cultures, is unclear. Further studies would be required to investigate if our results can be replicated in other settings. The limited number of sites and the short duration of the ‘core trial period’ also present limitations. Diagnostic coding is done by clinicians and hospital coders as part of standard practice. It is also possible that our intervention may not have long-term impact or may have a delayed impact, which highlights the need for investigation of more prolonged or ongoing strategies and measurement of longer-term outcomes. Another consideration is the lower imaging rates in the study hospitals when compared with imaging rates in the literature. There may not have been much room to move to allow for a reduction in imaging rates; however, without being able to judge the appropriateness of imaging, this remains unclear. We were also unable to track utilisation of the audit and feedback mechanisms. There may have been selection bias related to the clinicians who chose to participate in the intervention. Clinicians who participated in the education sessions and completed the follow-up surveys may be more likely to adhere to best practice and may be more interested in the topic. This may have impacted on the null primary outcome finding and the clinician-reported outcome results.
Comparison with other studies
A systematic review on the effectiveness of interventions to reduce lumbar imaging in the emergency department has been published.31 Five studies investigated the effects of implementation efforts to reduce lumbar imaging in the emergency department. The findings of these small studies suggested reduced referrals for specific imaging modalities, but they used non-randomised before-and-after study designs associated with a high risk of bias. Before this study, the effect of a multifaceted strategy to implement contemporary guideline recommendations for low back pain in the emergency department has not been assessed within a randomised trial. Despite the success of previous before-and-after studies, our trial demonstrated no clear evidence that the implementation of an evidence-based model of care for low back pain reduced lumbar imaging referrals in the emergency department.
Possible explanations and implications for clinicians, policymakers and future research
There may be several possible explanations for why we did not see change in imaging rates. One reason could be due to the limited alternatives to imaging offered by the intervention. Although the training emphasised the value of clinical assessment, rather than routine imaging in screening for serious spinal pathologies, there is evidence that patients’ expectations, clinicians’ concern for missing a serious pathology, fear of litigation, belief of minimal harm and time constraints contribute to overuse of lumbar imaging.32 We developed a clinician website that provided decision support to clinicians during the trial; however, we were unable to implement a decision support system within the electronic medical record as part of the intervention. This may have also contributed to the result. Another reason could have been because we were measuring the utilisation of imaging and unable to assess the appropriateness of imaging. The preintervention imaging rates in the study sites (~25%) were also lower when compared with the average imaging rates in the literature (~35%),5 which made reducing this further quite challenging. This may have made it difficult for clinicians to see a need for a change and therefore the audit and feedback mechanism may not have been as meaningful. Despite the lower baseline imaging rate, there were no previous efforts to attempt to reduce imaging at any of the sites, nor is lumbar imaging included in the Australasian College of Emergency Physicians Choosing Wisely recommendations.33 In countries like Canada, where discouraging the use of lumbar imaging in atraumatic low back pain is included in the Canadian Association of Emergency Physicians Choosing Wisely recommendations,34 emergency department imaging rates for non-serious low back pain are reported as low as 12%.35 However, in the USA, a large national initiative to reduce avoidable imaging found no significant change in imaging rates for non-serious back pain.36 It is evident that reducing imaging rates in low back pain is a complicated issue. Further studies assessing the appropriateness of imaging, implementing electronic medical record-based clinical decision tools and campaigns such as Choosing Wisely may be helpful in instilling behaviour change and reducing inappropriate lumbar imaging.
The trial’s success in reducing emergency department use of opioid medicines can be explained by some subcomponents of the intervention. For example, we provided feasible, evidence-based alternatives such as non-opioid analgesics as first-line care and access to heat wrap therapy, making it easier for clinicians to change their prescribing behaviour. The audit and feedback showed a reduction in opioid prescription almost immediately, and this positive reinforcement may have contributed to the ongoing reduction in prescriptions. The high preintervention rate likely was another facilitator for change.
While we provided improved rapid access to outpatient services, this was not sufficient to reduce length of emergency department stay, re-presentations and admissions. Reasons for admissions for low back pain are often multifactorial. For example, severe pain may limit function and ability for patients to cope at home and patients who live alone may be fearful of being alone. Key performance indicators, such as restrictions put on emergency department length of stay, and competing clinical demands, such as critically ill patients, make it difficult for emergency department clinicians to adequately address these issues during the emergency department stay. Nonetheless, admissions rates vary around the world, and in some places it is less of a problem than others. For example, a Canadian study reported that only 2.5%37 of patients with low back pain were being admitted from the emergency department, whereas an Australian study reported an admission rate of 34.1%.7 Further research is required to not only fully understand reasons for admission, but also address the multiple factors contributing to high rates of admission.
We observed a difference in effects on use of opioid medicines between the sites in our post-hoc subgroup analysis, including no clear reduction in opioid use at the rural site. This suggests that local factors probably play a key role in enabling the success of the implementation of a model of care in the emergency department. However, the lack of influence of age, socioeconomic status, triage category, primary diagnosis or mode of arrival on the intervention’s ability to reduce opioid use indicates that it is suitable for the broad range of patients with non-serious low back pain who present to the emergency department.
In this trial, there was no clear evidence that a multifaceted intervention to implement an evidence-based model of low back pain care in the emergency department reduced lumbar imaging referrals, but there was strong evidence that the intervention substantially reduced emergency department use of opioid medicines without adversely affecting patient outcomes and it also improved clinicians’ beliefs and knowledge about low back pain care. While this trial failed to show a definitive reduction in the rate of lumbar imaging (the trial’s primary outcome), it did show important reductions in opioid prescribing in the emergency department and shows promise for addressing one aspect of the opioid crisis.
Data availability statement
No data are available. Due to information governance restrictions imposed by organisations governing data access, we are unable to share the trial data unless applicants secure the relevant permissions.
Ethics statements
Patient consent for publication
Ethics approval
The trial protocol was approved by the Sydney Local Health District Human Research Ethics Committee (protocol number X17-0043).
Acknowledgments
We wish to thank all SHaPED trial investigators. FlexExe Australia supplied the heat wraps used in the trial at no cost. Qlik provided training and technical support in the development of the low back pain dashboard using Qlik Sense at no cost. The trial was sponsored by The University of Sydney (Sydney, Australia). We also acknowledge support from the Australia & New Zealand Musculoskeletal (ANZMUSC) Clinical Trials Network, NSW Agency for Clinical Innovation, NSW Emergency Care Institute and Sydney Local Health District. SHaPED investigators were entirely responsible for study design, conduct and data analysis.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
DMC and GCM are joint first authors.
X @gustavocmachado, @CGMMaher
Correction notice The licence to this article was updated to CC-BY-NC on the 09/09/2024.
Collaborators The SHaPED Investigators: Danielle M Coombs, Gustavo C Machado, Bethan Richards, Chris Needs, Rachelle Buchbinder, Ian A Harris, Kirsten Howard, Kirsten McCaffery, Laurent Billot, James Edwards, Eileen Rogan, Rochelle Facer, Qiang Li, Chris G Maher, Hannah Storey, Lorraine Ho, Noel Baidya, Marco Reedyk, Alfred van der Walt, Sharon Taylor, Matthew Oliver, Helen Broughton, Leonard Ladia, Boris Waldman, Matthew Chu, Timothy Tan, Mona Marabani, Glenn Wiseman, David Roberts, Lindy Collins, Randall Greenberg, David Lord-Cowell, Sokol Nushaj, Kristy Hatswell, Melinda Collett, Jeniffer Fiore-Chapman, Robyn Speerin, Teresa Anderson.
Contributors GM, BR, CN, RB, IAH, KH, KM, LB, JE, ER, RF and CM contributed to protocol development and design of the SHaPED trial. DC, GM, BR, CN and CM designed and delivered the SHaPED multifaceted intervention with input from JE, ER and RF. LB and QL led the statistical analysis with input from DC, GM, RB, IAH and CM. DC, GM, BR and CN were responsible for the conduct of the trial. GM and DC are the guarantors. All authors read and approved the final manuscript.
Funding This was an investigator-initiated study funded by Sydney Health Partners and the NSW Agency for Clinical Innovation. FlexEze™ Australia supplied the heat wraps used in the trial at no cost.
Competing interests GM, CM, RB and KM report grants from the Australian National Health and Medical Research Council during the conduct of this trial.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.