Abstract
Objective
To test the cost effectiveness of an implementation strategy to promote evidence-based practice for low back pain in the emergency department.
Methods
An economic evaluation was conducted alongside a stepped-wedge cluster-randomised controlled trial. The trial aimed to evaluate a strategy to implement a guideline-endorsed model of care in four emergency departments in New South Wales, Australia. The intervention targeted emergency clinicians and was compared to usual emergency care. The main trial outcomes were healthcare use that aligns with the main principles of the model of care. The outcomes explored in this economic evaluation were lumbar imaging referrals, opioid use, and hospital admissions for low back pain. Costs related to implementation development, delivery and healthcare utilisation were included. Bivariate linear multilevel regression analyses were conducted, adjusting for clustering, time and the correlation between cost and outcome to calculate incremental cost and effects and incremental cost-effectiveness ratios. Non-parametric bootstrapping with 5,000 replications of incremental cost and effect pairs was carried out and plotted on cost-effectiveness planes for each of the outcomes. Cost-effectiveness acceptability curves were generated to explore the probability of being cost-effective based on a range of willingness to pay thresholds for each of the outcomes. Sensitivity analyses were carried out to determine to what extent the decision to exclude episodes of care missing healthcare costs and including inpatients costs impacted the results.
Results
The implementation was more effective but more costly when considering the three key outcomes. The implementation was cost-effective in reducing opioid use with an incremental cost-effectiveness ratio of $3,574.29 per episode of care where opioids were avoided. The incremental cost effectiveness ratios for avoiding imaging and hospitalisation were $26,298.50 and $49,290.00 per episode of care, respectively. The first sensitivity analysis highlighted uncertainty with the hospital admission result and the second sensitivity analysis found that the implementation was more likely to be cost-effective when considering emergency department costs only.
Conclusion
This implementation of the model of care may be cost-effective for reducing opioid use in patients who present to emergency with low back pain, however more research into willingness to pay to avoid opioid use in emergency departments is required.
Supplementary Information
The online version contains supplementary material available at 10.1186/s12962-025-00652-6.
Key messages
What is already known on this topic. Low value care for low back pain is prevalent in the emergency department. A multifaceted implementation strategy of an evidence-based model of care did not change the primary outcome imaging, however it reduced opioid use in the emergency department without adversely impacting pain outcomes and satisfaction with care.
What this study adds. This study illustrates that in addition to this multifaceted implementation strategy being effective in reducing opioid use within the Emergency department, it is also likely to be cost effective in reducing opioid use.
How this study might affect research, practice or policy. This study gives a clearer picture to policy makers about whether implementing this model of care is cost-effective, and should be adopted by other hospitals.
Supplementary Information
The online version contains supplementary material available at 10.1186/s12962-025-00652-6.
Introduction
Many people seek care for low back pain in the emergency department [1]. It is advised in most low back pain clinical guidelines that imaging should be reserved for those where serious pathology is suspected [2]. Furthermore, simple analgesia should be used before opioids [2], and hospitalisation for low back pain is discouraged [3]. Despite this, about 35% of patients presenting to the emergency department with low back pain receive diagnostic imaging [4], opioid medicines are prescribed in up to 70% [5, 6], and about 20% are admitted to hospital [5, 7]. These practices may be detrimental to patient outcomes [8] and use scarce health care resources when they could be better used elsewhere [3, 9].
The Sydney Health Partners Emergency Department (SHaPED) trial evaluated the implementation of a new evidence-based model of care into four emergency departments in New South Wales in an attempt to combat this problem [10]. The trial found that a multifaceted strategy to implement this model of care did not change the primary outcome (i.e. lumbar imaging), or a key secondary outcome of hospitalisations, however it did significantly reduce opioid use without adversely affecting patient outcomes such as pain, function and satisfaction with emergency department care, when compared to usual emergency department care [11].
The objective of this economic evaluation is to calculate the cost-effectiveness of the implementation strategy in SHaPED compared to usual emergency department care. It is important to understand the cost-effectiveness of an intervention to inform decisions about future implementation in view of the scarcity of healthcare resources.
Methods
We conducted a cost-effectiveness analysis alongside a stepped wedge cluster-randomised controlled trial [11]. The published protocol [10] for the main paper outlined the plan for this economic evaluation. A protocol for the analysis of this economic evaluation was completed prior to completing the analysis, however a Health Economic Analysis Plan was not published. This study is reported as per the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) [12]. Ethics approvals were gained from Sydney Local Health District Human Research Ethics Committee (X17-0419 and X17-0043). This cost effectiveness analysis was taken from a hospital perspective to better understand the cost and benefits of the implementation strategy to the hospitals, in view of determining whether it should be scaled up to other sites. In the SHaPED trial, there were four emergency department sites. The trial was conducted from July 1, 2017 to February 17, 2019. All presentations (or episodes of care) for non-serious low back pain during this period, where the patient was older than 18 years, were included. The episodes of care were identified using SNOMED CT® codes (Australian version - Emergency Department Reference Set) related to non-specific low back pain or radicular low back pain. There was a 12-month retrospective control period, a four-week intervention period at each of the sites that was rolled out in a randomised order, and a follow-up period which ended three months after the final site received the intervention. The intervention period (when the implementation strategy was rolled out) was between August 1, 2018 and November 16, 2018. The main trial results have been published elsewhere [11].
Description of the intervention
A multifaceted implementation strategy was directed at emergency department clinicians including doctors, nurses and physiotherapists and used to implement the model of care to the four emergency departments in New South Wales. The model of care [13] was developed by the NSW Health Agency for Clinical Innovation and involved input from clinical experts, researchers, and consumers. Implementation strategies included (i) clinician education seminars, (ii) educational materials including how to image appropriately, (iii) suggestions of non-opioid pain management strategies, (iv) fast-track referral options to outpatient services, and (v) audit and feedback on key healthcare utilisation outcomes (imaging, admissions, opioid use). The full description of the intervention can be found published with the trial results [11]. The control period consisted of usual emergency department care (i.e., before any implementation).
Outcomes
The key trial outcomes were healthcare utilisation outcomes that are considered low-value care and discouraged in the model of care [13]. Low-value care is defined as a treatment or intervention where the effectiveness does not adequately out weight the costs or harms associated with that treatment [14]. Lumbar imaging, opioid use, and hospital admissions have limited evidence of effectiveness in the low back pain population, are costly and may cause harm [3]. The key trial outcomes that were used as the effectiveness outcomes in this cost-effectiveness analysis are use of lumbar imaging (primary outcome), opioid use in the emergency department, and admission to hospital (key secondary outcomes identified for economic evaluation a priori). These healthcare use outcomes were extracted from the hospitals’ electronic medical record. For the economic analysis, the outcomes were framed as ‘outcomes avoided’.
Measurement and valuation of resource use
We estimated three broad categories of cost: (i) cost of implementation development, (ii) cost of implementation delivery, and (iii) costs associated with healthcare utilisation. Full descriptions of costs and sources of costing data are listed in Appendix 1. The implementation development costs, and implementation delivery costs were collected from trial logbooks, where the frequency of resource use was recorded, and in some cases estimated, and translated into monetary values as described in Appendix 1. These costs consisted of cost of developing the education materials and audit and feedback mechanism (i.e., staff time), cost of the education materials (e.g., printing costs, website fees), cost of training sessions (e.g., presenter and recipients time), and any travel required by the trial team to carry out the education sessions. Staff time was estimated based on trial logbooks and estimates determined by the authors. The intervention delivery and development costs were combined and calculated as a per episode of care cost, by dividing the total implementation cost (delivery and development) by the number of episodes of care in the intervention group of the main trial. This cost was then allocated to each of the episodes of care in the intervention group. Healthcare use costs were calculated from presentation (or episode of care) level costing data used for reporting and funding purposes by the hospitals (i.e., the actual healthcare cost attributed to each episode of care). These healthcare cost data include any care or resource used by the patient during that episode of care within the emergency department and hospital setting and is presented as the total cost for each episode of care. This includes resources associated with imaging, pathology, medications, staff costs (nursing, medical, allied health), administration costs, and hospital overheads. Costs are expressed in Australian dollars (AUD), adjusted for inflation using the Health Consumer Price Index for the financial year 2018–19, as published by the Australian Bureau of Statistics [15]. We have chosen the year 2018-19 as this was the year the model of care was implemented, and this was pre-specified in our published protocol.
Statistical analysis
Analysis approaches were based on recommendations for conducting economic evaluations alongside stepped-wedge trials [16]. Differences in costs and effects between the intervention and control phases were estimated using bivariate linear multilevel regression analyses. Clustering at the site level, adjusting for time based on the measurement period and correlation between cost and outcomes were incorporated in the multilevel models. An incremental cost effectiveness ratio (ICER) was calculated for each of the key outcomes and presented as the incremental costs per episode of care avoiding (1) lumbar imaging referrals, (2) opioid use, and (3) hospital admissions. We explored individual level sampling uncertainty through bootstrapping. Non-parametric bootstrapping with 5,000 replications was used to estimate 95% confidence intervals around the incremental cost and effect pairs and plotted on an incremental cost-effectiveness plane. Cost-effectiveness acceptability curves were also estimated to show the probability that the implementation is cost-effective when considering different decision maker’s willingness to pay for episode of care avoiding imaging, opioid use, and hospital admission. Episodes of care with missing healthcare costing data were excluded from the base case analysis (one of four sites); the effect was examined in sensitivity analyses, below. Costs and outcomes were not discounted due to the length of the intervention period. The intervention period (including the follow up period) lasted seven months. STATA version 18 (StataCorp LP, College Station, TX, USA) was used for analyses.
Sensitivity analyses
A sensitivity analysis examined the inclusion (with multiple imputation) of episodes of care with missing cost data. The methods of imputation were chosen based on guidance published by Faria et al. [17] Multivariate Imputation by Chained Equations was carried out considering all available variables (triage category, sex, socioeconomic status, ambulance status, emergency department length of stay, admission status, opioid use and imaging orders). Predictive mean matching was used. The imputation was run 30 times which is more than sufficient given there is 13.3% missing costing data. We then went on to follow statistical analysis consistent with the base analysis using the imputed data sets. A second sensitivity analysis (or scenario analysis) explored the impact of including only emergency department costing data, as the trial intervention only targeted emergency department care. As only total cost for hospital and emergency department care was available, mean conditional data imputation was used to estimate emergency department only costs for the episodes of care that also included a subsequent hospitalisation. We used age, triage category (as a measure of severity of low back pain), and ambulance arrival status to inform the imputation. These factors have been shown to be associated with higher emergency department costs in a previous study [18].
Patient and public involvement statement
This economic evaluation was completed without consumer involvement in study design. Patients or public were not invited to contribute to the writing or editing of this document for readability or accuracy.
Results
There were 4,625 episodes of care in the main trial. After excluding episodes of care that were missing costing data (n = 616, 13.3%), 4,009 eligible episodes of care were included in the base economic analysis and a complete case analysis was adopted. All missing costing data came from one of the four emergency department sites. The characteristics of the episode of care included in the base analysis are summarised in Table 1. Overall, the baseline characteristics were balanced between the two groups. The characteristics of episodes of care included in the sensitivity analyses (i.e., including the episodes of care with missing costing data with multiple imputation) are described in Appendix 2. The characteristics of episodes of care included in the sensitivity analysis are similar between groups and similar to the base analysis.
Table 1.
Characteristics of episodes of care included in the base economic analysis
| Characteristic | Intervention (n = 1,275) | Control (n = 2,734) |
|---|---|---|
| Age – years, mean (SD) | 53.8 (20.1) | 52.8 (20.4) |
| Female sex, n (%) | 621 (48.7) | 1,362 (49.8) |
| Primary diagnosis, n (%) | ||
| Non-specific low back pain | 1,140 (89.4) | 2,452 (89.7) |
| Radicular low back pain | 135 (10.6) | 282 (10.3) |
| Socioeconomic status, n (%) | ||
| SEIFA deciles 1–5 | 229 (19.0) | 627 (24.2) |
| SEIFA deciles 6–10 | 973 (81.0) | 1,963 (75.8) |
| Mode of arrival, n (%) | ||
| Ambulance | 404 (31.7) | 892 (32.6) |
| Other | 869 (68.3) | 1,842 (67.4) |
| Triage category, n (%) | ||
| 2 (emergency) | 29 (2.3) | 46 (1.7) |
| 3 (urgent) | 465 (36.5) | 1,042 (38.1) |
| 4 (semi-urgent) | 763 (59.8) | 1,569 (57.4) |
| 5 (non-urgent) | 18 (1.4) | 77 (2.8) |
| Length of emergency stay – hours, mean (SD) | 4.1 (2.7) | 4.1 (2.8) |
Data are missing for socioeconomic status (n=217), and mode of arrival (n=2). SEIFA=socio-economic indexes for areas. Lower numbers in socioeconomic status are more disadvantaged, and higher numbers are less disadvantaged. Triage categories are based on the Australasian Triage Scale; there were no eligible low back pain presentations in the triage category 1 (life threatening). SD=standard deviation
Outcomes
Healthcare use outcomes are presented in appendix 3. For the base economic analysis, where 616 episodes of care were excluded due to missing cost data, 23.7% (302/1275) of episodes of care in the intervention group received lumbar imaging referrals and 25.8% (705/2734) received lumbar imaging referrals in the control group. 50.4% (636/1275) of episodes in the intervention group were given an opioid, compared with 65.6% in the control. Finally, 16.9% (215/1275) of episodes of care resulted in a hospital admission in the intervention group compared to 16.3% (445/2734) in the control. In the sensitivity analyses, where all episodes were included, the effectiveness outcomes were slightly different to the base analysis due to the additional 616 episodes of care. These outcomes are the same as the main trial and reported in Appendix 3.
Costs
The implementation development costs were calculated as $10.39 per episode of care and the implementation delivery costs were $31.67 per episode of care (Appendix 3). These costs were allocated to each episode of care in the intervention group. We decided to include the implementation development costs in the analysis as there would be development costs at each new site to adjust the implementation strategy to the local context. For example, the audit and feedback dashboard would need to be developed at each site to be compatible with local electronic medical record systems. There would also likely be some re-development of the education material over years as best evidence changes. We also believe there may need to be some “top up” implementation over time due to staff rotation within the emergency department. For these reasons, we decided to include development costs to give a more conservative estimate of the implementation strategy costs. Each episode of care had an individual total cost which included the individual emergency department and inpatient cost if in the control period, and the individual emergency department and inpatient cost, the implementation development cost, and the implementation delivery cost if in the intervention period. These individual episodes of care costs were used in the multilevel models to estimate adjusted incremental costs of the intervention over the control (Table 2). The mean cost per episode of care is presented in Appendix 3.
Table 2.
Cost effectiveness outcomes – adjusted differences in outcomes and costs between intervention and control groups, bootstrapped incremental cost effectiveness ratio (ICER) and proportion of bootstrapped observations in each quadrant of the cost effectiveness (CE) plane
| Episodes of care | Adjusted incremental difference of intervention over control (95% confidence interval)* | ICER (per episode of care avoiding the low value care outcome) |
Proportion of bootstrapped observations in each CE quadrant | ||||||
|---|---|---|---|---|---|---|---|---|---|
| Intervention | Control | Cost difference (95% CI) | Effect difference (95% CI) | NE % | SE % | NW % | SW % | ||
| Base Economic Analysis | |||||||||
| Lumbar imaging | 1,275 | 2,734 | 525.97 (−411.85 to 1425.14) | −0.02 (−0.07 to 0.03) | $26298.50 | 67.6% | 12.0% | 19.0% | 1.4% |
| Opioid use | 1,261 | 2,532 | 500.40 (−424.41 to 1389.83) | −0.14 (−0.19 to −0.08) | $3574.29 | 85.4% | 14.6% | 0 | 0 |
| Hospital admission | 1,275 | 2,734 | 492.90 (−435.55 to 1383.67) | −0.01 (−0.05 to 0.03) | $49290.00 | 54.7% | 14.1% | 30.3% | 0.9% |
| Sensitivity Analysis – Multiple imputation for missing costs | |||||||||
| Lumbar imaging | 1,392 | 3,233 | 476.12 (−321.96 to 1227.23) | −0.02 (−0.06 to 0.03) | $23806.00 | 68.2% | 10.6% | 19.8% | 1.4% |
| Opioid use | 1,378 | 3,031 | 457.51 (−339.26 to 1212.62) | −0.13 (−0.18 to −0.08) | $3519.31 | 87.1% | 12.9% | 0% | 0% |
| Hospital admission | 1,392 | 3,233 | 456.25 (−337.67 to 1211.30) | −0.005 (−0.04 to 0.03) | $91250.00 | 49.4% | 11.8% | 37.7% | 1% |
| Sensitivity Analysis–emergency department costs only | |||||||||
| Lumbar imaging | 1,392 | 3,233 | 4.51 (−56.04 to 63.40) | −0.02 (−0.06 to 0.03) | $225.50 | 40.3% | 37.4% | 15.3% | 7% |
| Opioid use | 1,378 | 3,031 | 3.73 (−56.85 to 62.87) | −0.13 (−0.19 to −0.08) | $28.69 | 54.7% | 45.3% | 0 | 0 |
| Hospital admission | 1,392 | 3,233 | 3.61 (−56.72 to 62.82) | −0.005 (−0.04 to 0.03) | $722.00 | 30.4% | 29.2% | 24.3% | 16.1% |
*Differences in costs and effects (intervention minus control) were estimated using bivariate linear multilevel regression analyses adjusting for clustering, correlation between cost and outcomes and time. N.B. minus for effectiveness outcomes indicate a reduction in the low value care, therefore indicating effectiveness of intervention (i.e., outcomes avoided). Minus in the cost difference indicates the adjusted costs in the control group were more than the intervention group. ^ Cost effectiveness plane quadrants: NE (North East) - new intervention more effective but more costly; SE (South East) – new intervention more effective and less costly (intervention dominates); NW (North West) – new intervention less effective and more costly (control dominates); SW (South West) – new intervention less effective and less costly
Cost-effectiveness analysis
After adjusting for clustering, time and correlation between cost and outcomes, the differences in effectiveness and costs between groups are reported in Table 2. The base analysis found the incremental cost effectiveness ratio to be $26,298.50 per episode of care avoiding lumbar imaging, $3,574.29 per episode of care avoiding opioids and $49,290.00 per episode of care avoiding hospitalisation, when compared to usual care. 5,000 bootstrapped replications for the base analysis are plotted on the cost-effectiveness planes in Fig. 1 and the proportion of estimates in each quadrant are reported in Table 2. For the outcomes of imaging and admission, there was heterogeneity in the effect, i.e. the implementation ranged from being less effective to more effective in the cost effectiveness plane, whereas for opioids, the implementation was always more effective than the control. The bootstrapped mean incremental cost and effect pairs showed that the implementation was more likely to be effective but more costly for each of the outcomes. The cost-effectiveness acceptability curves for each outcomes in the base analysis are presented in Fig. 2. These show the probability the implementation was cost-effective when considering decision makers willingness to pay. For example, if the decision maker was willing to $30,000 for each additional episode of care avoiding imaging, there is a 55% probability that the implementation is cost-effective when compared to usual care. Similarly for opioids and hospital admission, if the decision maker was willing to pay $30,000 for each additional episode of care avoiding these outcomes, the probability that the implementation is cost effective would be 100% for opioids and 43% for hospital admission.
Fig. 1.
Cost-effectiveness planes for base analysis. Incremental in costs and effects (intervention minus control) were estimated using bivariate linear multilevel regression analyses and non-parametric bootstrapping with 5,000 replications was used to estimate 95% confidence intervals around the incremental cost and effect. The x-axis has been reversed to more easily visualise effectiveness outcomes so that aligns with convention for cost effectiveness planes i.e., minus effectiveness equates to outcome avoided, so therefore more effective. Cost effectiveness plane quadrants: NE (North East) - new intervention more effective but more costly; SE (South East) – new intervention more effective and less costly (intervention dominates); NW (North West) - new intervention less effective and more costly (control dominates); SW (South West) – new intervention less effective and less costly
Fig. 2.
Cost-effectiveness acceptability curves for base analysis. Cost-effectiveness acceptability curves show the probability that the intervention is cost-effective when considering different decision maker’s willingness to pay for episode of care avoiding imaging, opioid use, and admission
In this first sensitivity analysis, it was found that it cost $23,806.00 per episode of care avoiding lumbar imaging, $3,519.31 per episode avoiding opioids and $91,250.00 per episode of care where a hospital admission was avoided, when compared to usual care (Table 2 and Appendix 4). These incremental cost effectiveness ratios are similar to the base analysis for imaging and opioids avoided but almost double for hospital admissions. Regarding willingness to pay, if the decision maker was willing to pay $30,000 per episode of care avoiding the low-value care outcome, the probability of the intervention being cost-effective is the same as the base analysis but drops to 36% for hospital admissions (Appendix 6). When exploring cost-effectiveness when considering emergency department only costs (second sensitivity analysis), it showed the intervention was more cost effective compared to the base analysis: $225.50 per episode of care avoiding lumbar imaging, $28.69 per episode avoiding opioids and $722.00 per episode of care avoiding hospital admission (Table 2 and Appendix 5).
Discussion
On average, when adjusting for time, clustering and correlation between cost and outcome, the multifaceted strategy to implement a new guideline-endorsed model of care for low back pain is most likely more effective but more costly when compared to usual care. However, it is important to note that there is heterogeneity in the bootstrapped incremental costs, and most of the incremental effects are small and not statistically significant for both lumbar imaging and hospital admissions. In lumbar imaging for example, although 80% of bootstrapped observations indicate a reduction in referrals, 88% of the observations indicate a reduction in proportion of lumbar imaging referrals of less than 5%. This is similar in hospitalisations where 96% of the observations indicate a reduction in proportion of hospitalisations of less than 5%. The intervention was clearly effective in reducing the proportion of episodes of care where opioids were used with 100% of estimates indicating that the implementation was more effective in avoiding opioid use than usual care. The base analysis found an incremental cost effectiveness ratio of $26,298.50 per episode of care avoiding lumbar imaging, $3,574.29 per episode of care avoiding opioids and $49,290.00 per episode of care avoiding hospitalisation, when compared to usual care.
The probability of the intervention being cost-effective depends on the decision makers willingness to pay for reducing opioid use, imaging referrals, and hospital admissions in this population. There are currently no accepted thresholds for willingness to pay for these outcomes. To understand if the intervention might be acceptable, it is helpful to explore factors influencing willingness to pay for each of the outcomes. There is currently limited evidence in the literature on the cost-effectiveness of reducing each of the outcomes. For imaging, a recent observational study that found that in a program to reduce Computerised Tomography (CT) lumbar imaging amongst Australian GPs the incremental cost-effectiveness ratio was $2.82 per CT scan avoided, however they only considered the program costs and did not consider health care costs or other costs from a health care provider perspective [19]. They considered a willingness to pay threshold of $231, and based this on how much on average the Medicare Benefits Schedule (MBS) is reimbursed for CT imaging [19]. This may be too simplistic given the overall health care costs and known harms associated for those who receive imaging [8]. For opioids, most studies in reducing opioids look at the cost-effectiveness of reducing opioid use are in patients that already have opioid use disorders, rather than pain populations [20]. Similar to other conditions, a commonly reported willingness to pay threshold is $50,000 per quality-adjusted life-year (QALY) gained, however there is no accepted willingness to pay threshold for natural units, such as additional patient in treatment [20]. In the context of our study, given the societal costs and potential for significant harms (including overdose, hospital admission and death) associated with opioid misuse [21], the ICER of $3,574.29 per additional episode of care avoiding opioids may offer good value for money. However, we cannot know for sure without an accepted willingness to pay. For hospital admissions, given that on average a hospitalisation costs approximately $14,000 in this population [11]together with the risk of hospital-acquired complications, we may be willing to pay a considerable amount for a hospitalisation avoided. However, the probability that this intervention will avoid hospitalisation is relatively low, for example, when paying $50,000, there probability the intervention is cost-effective is only 51%.
Understanding whether reducing the proportion of these low-value care practices in the emergency department would have long-term benefits or reduction in harms, such as hospital-acquired complications or opioid misuse, would provide further insights on whether this implementation provides good value for money. Future research into these long-term benefits and costs or cost savings associated with these, including modelling studies, would assist in determining willingness to pay for imaging, opioids and hospitalisation avoided in the emergency department setting. In addition, future studies should consider quality adjusted life years (QALYs) as an outcome so that willingness to pay is more easily estimated and more meaningful comparisons can be made to other studies and conditions.
The first sensitivity analysis was carried out to determine to what extent the choice of excluding the episodes of care with missing data impacted the results. For the opioid and imaging outcomes, the incremental cost-effectiveness ratios were similar to the base analysis. For hospitalisations, the incremental cost-effectiveness ratio jumped from $49,290.00 in the base analysis to $91,250.00, making the probability of the intervention being cost-effective, when considering willingness to pay, very low. This is likely partly due to the heterogeneity of incremental effect and the small proportion of change in both directions, indicating that the intervention had likely no true effect on hospitalisation. The second sensitivity analysis explored the choice of including inpatient costs in the base analysis, by exploring only emergency department healthcare costs. These incremental cost-effectiveness ratios were $225.50 per episode of care where lumbar imaging was avoided, $28.69 per episode of care where opioids were avoided and $722.00 per episode of care where an admission was avoided, when compared to usual care. This indicates the there is a higher probability that the implementation had an impact on and was cost-effectiveness within the emergency department. This is unsurprising, given the implementation only targeted clinicians within the emergency department.
Strengths and limitations
The strengths of the study include that the analysis took in to account the stepped wedge cluster design and was adjusted for time, clustering and correlation between cost and outcome. Another strength was the use of administratively linked healthcare costs for three of the four sites in the analysis rather than self-reported estimates, which have the potential for bias. This leads to the limitations to this study. Firstly, there was missing costing data for one of the sites (n = 616) and it was decided to exclude these episodes of care from the base analysis. A sensitivity analysis indicated that excluding these episodes had some impact on the results for the hospitalisation outcome, however it did not change the conclusion. Finally, although this was not an aim, a limitation included that the study was unable to measure out-of-hospital healthcare costs, indirect costs, and societal costs, as well as out-of-hospital effects. In addition, due to the trial design, quality of life was not measured and therefore we were unable to estimate the interventions effect on QALYs. This would give further information to support whether this multifaceted strategy to implement a guideline-endorsed model of care should be adopted by other emergency departments.
In conclusion, this economic evaluation indicates that the multifaceted implementation strategy of a model of care for low back pain in emergency departments was likely more effective and more costly when compared to usual care for opioid use, imaging, and hospital admissions avoided. It is, however, likely that this implementation is cost-effective for reducing opioid use. Future research on willingness to pay will give a clearer picture of the cost-effectiveness of this intervention, and similar interventions designed to reduce these practices in emergency departments. Nonetheless, these results indicate promise in our strategy for implementing this model of care for improving the cost-effectiveness of care in the emergency department.
Supplementary Information
Acknowledgements
We would like to thank the University of Sydney, Sydney Musculoskeletal Health Flagship Centre and biostatistician Nicola Jones for their statistical support in implementing the sensitivity analyses.
Author contributions
All authors contributed substantially to concept and design, and interpretation of the data. TL, MBP, DMC, KH and GCM contributed to the data analysis. DMC drafted the manuscript. All authors participated substantially in the critical revision of the manuscript. Study supervision was conducted by CGM and GCM.
Funding
The main trial was an investigator-initiated study funded by Sydney Health Partners and the NSW Agency for Clinical Innovation. FlexExe™ Australia supplied heat wraps used in the main trial at no cost.
Data availability
Data sharing 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.
Declarations
Ethics approval
Ethics approvals were gained from Sydney Local Health District Human Research Ethics Committee (X17-0419 and X17-0043).
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Edwards J, Hayden J, Asbridge M, et al. Prevalence of low back pain in emergency settings: a systematic review and meta-analysis. BMC Musculoskelet Disord. 2017;18(1):143. 10.1186/s12891-017-1511-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Oliveira CB, Maher CG, Pinto RZ, et al. Clinical practice guidelines for the management of non-specific low back pain in primary care: an updated overview. Eur Spine J. 2018;27(11):2791–803. 10.1007/s00586-018-5673-2. [DOI] [PubMed] [Google Scholar]
- 3.Foster NE, Anema JR, Cherkin D, et al. Prevention and treatment of low back pain: evidence, challenges, and promising directions. Lancet. 2018;391(10137):2368–83. 10.1016/s0140-6736(18)30489-6. [DOI] [PubMed] [Google Scholar]
- 4.Downie A, Hancock M, Jenkins H, et al. How common is imaging for low back pain in primary and emergency care? Systematic review and meta-analysis of over 4 million imaging requests across 21 years. Br J Sports Med. 2020;54(11):642–51. 10.1136/bjsports-2018-100087. [DOI] [PubMed] [Google Scholar]
- 5.Ferreira GE, Machado GC, Abdel Shaheed C, et al. Management of low back pain in Australian emergency departments. BMJ Qual Saf. 2019;28(10):826–34. 10.1136/bmjqs-2019-009383. [DOI] [PubMed] [Google Scholar]
- 6.Kamper SJ, Logan G, Copsey B, et al. What is usual care for low back pain? A systematic review of health care provided to patients with low back pain in family practice and emergency departments. Pain. 2020;161(4):694–702. 10.1097/j.pain.0000000000001751. [DOI] [PubMed] [Google Scholar]
- 7.McCaughey EJ, Li L, Georgiou A, et al. Imaging for patients presenting to an emergency department with back pain: impact on patient pathway. Emerg Med Australas. 2016;28(4):412–8. 10.1111/1742-6723.12602. [DOI] [PubMed] [Google Scholar]
- 8.Webster BS, Bauer AZ, Choi Y, et al. Iatrogenic consequences of early magnetic resonance imaging in acute, work-related, disabling low back pain. Spine (Phila Pa 1976). 2013;38(22):1939–46. 10.1097/BRS.0b013e3182a42eb6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Hartvigsen J, Hancock MJ, Kongsted A, et al. What low back pain is and why we need to pay attention. Lancet. 2018;391(10137):2356–67. 10.1016/s0140-6736(18)30480-x. [DOI] [PubMed] [Google Scholar]
- 10.Machado GC, Richards B, Needs C, et al. Implementation of an evidence-based model of care for low back pain in emergency departments: protocol for the Sydney health partners emergency department (SHaPED) trial. BMJ Open. 2018;8(4):e019052. 10.1136/bmjopen-2017-019052. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Coombs DM, Machado GC, Richards B, et al. Effectiveness of a multifaceted intervention to improve emergency department care of low back pain: a stepped-wedge, cluster-randomised trial. BMJ Qual Saf. 2021. 10.1136/bmjqs-2020-012337. [DOI] [PubMed] [Google Scholar]
- 12.Husereau D, Drummond M, Augustovski F, de Bekker-Grob E, Briggs AH, Carswell C, Caulley L, Chaiyakunapruk N, Greenberg D, Loder E, Mauskopf J, Mullins CD, Petrou S, Pwu RF, Staniszewska S. CHEERS 2022 ISPOR good research practices task force. Consolidated health economic evaluation reporting standards 2022 (CHEERS 2022) statement: updated reporting guidance for health economic evaluations. BMJ. 2022;376:e067975. [DOI] [PubMed] [Google Scholar]
- 13.NSW Agency for Clinical Innovation. Management of people with acute low back pain: model of care. Chatswood; NSW Health; 2016. 39 p.
- 14.Scott IA, Duckett SJ. In search of professional consensus in defining and reducing low-value care. Med J Aust. 2015;203(4):179–81. 10.5694/mja14.01664. [DOI] [PubMed] [Google Scholar]
- 15.Australian Bureau of Statistics. Consumer Price Index, Australia 2020. [Spreadsheet] ABS [Accessed March 2020] https://www.abs.gov.au/statistics/economy/price-indexes-and-inflation/consumer-price-index-australia/mar-2020#data-download
- 16.Lung T, Si L, Hooper R, et al. Health economic evaluation alongside stepped wedge trials: a methodological systematic review. Pharmacoeconomics. 2021;39(1):63–80. 10.1007/s40273-020-00963-x. [DOI] [PubMed] [Google Scholar]
- 17.Faria R, Gomes M, Epstein D, White IR. A guide to handling missing data in cost-effectiveness analysis conducted within randomised controlled trials. Pharmacoeconomics. 2014;32(12):1157–70. 10.1007/s40273-014-0193-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Coombs DM, Machado GC, Richards B, et al. Healthcare costs due to low back pain in the emergency department and inpatient setting in Sydney, Australia. Lancet Reg Health. 2021;7:100089. 10.1016/j.lanwpc.2020.100089. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Morgan T, Wu J, Ovchinikova L, Lindner R, Blogg S, Moorin R. A national intervention to reduce imaging for low back pain by general practitioners: a retrospective economic program evaluation using medicare benefits schedule data. BMC Health Serv Res. 2019;19(1):983. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Beaulieu E, DiGennaro C, Stringfellow E, Connolly A, Hamilton A, Hyder A, et al. Economic evaluation in opioid modeling: systematic review. Value Health: J Int Soc Pharmacoeconomics Outcomes Res. 2021;24(2):158–73. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Degenhardt L, Grebely J, Stone J, Hickman M, Vickerman P, Marshall BDL, et al. Global patterns of opioid use and dependence: harms to populations, interventions, and future action. Lancet (London England). 2019;394(10208):1560–79. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Data sharing 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.