Main findings
This study classified 2,051 lumbar MRI referrals as compliant or non-compliant with the 2015 version of the ACR Imaging Appropriateness Criteria for LBP [3]. The classification was based on the narrative text from the MRI referral, and no patients were clinically evaluated in this study.
Three-quarters (75.5%) of the MRI referrals were deemed inappropriate, and 24.5% were classified as appropriate. The distribution was stable from 2014 to 2018. To our knowledge, no active implementation strategies targeting imaging for spinal pain conditions in general practice were carried out during that period.
Comparison with other studies
Many studies have investigated the appropriateness of lumbar MRI, and a variety of methods have been used [8, 15–18]. Some studies used imaging guidelines to clarify the appropriateness of MRI, and others used clinical symptoms and/or various standards of red flags [19, 20]. The considerable variation in methods for classification of appropriateness of MRI makes it difficult to make comparisons between studies.
An Australian systematic review from 2018 estimates the overuse and underuse of imaging in the management of LBP [8]. The review included 33 studies and assessed the use of X-ray, CT and/or MRI imaging referrals for patients presenting for care. Inappropriate referrals in people referred for imaging were assessed in 23 studies and showed a pooled effect of 34.8% when ‘Absence of red flag clinical features’ was used as the inappropriateness criterion and 31.6% when ‘No clinical suspicion of pathology’ was the criterion. The majority of studies assessed patients referred from general practitice. Four studies[20–23] assessed imaging referrals received by radiology departments from primary care physicians, which we considered were comparable to the setting of our study. Rates of inappropriate referrals ranged from 20–47.9% which were considerably lower than the 75.5% found in our study. One study[23] used the same ACR imaging guideline as in our study to assess appropriateness of MRI referrals and reported the highest rate of inappropriate referrals (47.9%) of the four studies.
A scoping review from 2019 included 23 studies describing adult LBP imaging appropriateness in general practice [19]. A range of red flag features was utilised to determine imaging appropriateness. Most studies considered appropriateness in a binary manner by the presence of any red flag feature. Ten different guidelines were referenced in 16 of the 23 studies, while seven studies (30%) used combined methods or modified guidelines. The method for calculating the proportion of inappropriate imaging varied. Ten percent of the studies used the total number of patients presenting with LBP as the denominator, suggesting most studies underestimated the rate of inappropriate imaging and did not capture where imaging was not performed for clinically suspicious LBP.
Both the Australian systematic review from 2018 [8] and the scoping review from 2019 [19] conclude that many different methodologies are used to assess LBP imaging appropriateness.
The current study is therefore not directly comparable to the studies mentioned above. Although they are based on data from MRI referrals, none of them uses the same guideline or checklist to classify referrals, and they do not report the same considerably high number (75.5%) of inappropriate MRI referrals as was found in our study. This large proportion of inappropriate referrals in the current study was probably due to the strict criteria for appropriateness. Only precise information on previous non-surgical treatment and the duration of that treatment in the referral text was used in the evaluation of appropriateness. If information about previous treatment or duration was not provided, the referral was classified as inappropriate, even if text like "the patients have for some time maintained training" was available. In cases like this, it was unclear what type or level of training had been performed or what period of time was involved (more or less than 6 weeks), which is essential information in the evaluation of appropriateness when using the ACR imaging guideline. If the criterion of duration was omitted, 38.3% (75.5 minus 37.2%) of the MRI referrals would have been inappropriate which is still higher but closer to the results reported in the previously mentioned literature [8, 19]. It is possible that important clinical symptoms were absent due to oversight in the referrals or that information about non-surgical treatment or duration was not mentioned because of a lack of knowledge about imaging referral guidelines for LBP. Furthermore, there was a tradition in the department of accepting all referrals despite the lack of information and no feedback procedures existed to inform the GPs that some referrals lacked proper information. This could partly explain why important information was not included in the referrals, which led to a substantially higher number (75.5%) of inappropriate MRI referrals in our study.
Demographic data
More than half of the appropriate referrals were categorised as either Variant 4) 'Candidate for surgery or intervention with persistent or progressive symptoms during or following 6 weeks of conservative management' or Variant 5) 'New or progressing symptoms or clinical findings with a history of prior lumbar surgery'. Both variants are related to clinical management rather than suspicion of serious pathology. In the current study, only 8.5% of all referrals (35% of the appropriate referrals) were referred due to suspicion of serious pathology as in Variants 2, 3 or 6 (fracture, infection, cancer, or cauda equina). In comparison, a study by Gidwani et al. [17] found that 24% of the appropriate MRIs for LBP (n=76.663) had suspicion of red flag conditions as identified by diagnostic codes (IDC-9-CM).
Mean age and sex distribution were relatively comparable between the six variants (Table2). Only Variant 2 (Suspicion of fracture) had a higher percentage of women (70.97%) compared with men (29.03%), and mean age (67.7) was higher when compared to the other variants. This finding is in line with an increasing risk of osteoporotic fractures among women with increasing age.
Perspectives
This study focused on referrals for MRIs of the lumbar spine, in which the narrative text of the referrals was not compared with clinical data. Therefore, it is not possible to determine if a referred patient truly had indications for MRI. Also, it is not possible to measure the proportion of cases where clinical symptoms indicated appropriate MRI but where MRI was not performed (underuse). However, the study reflects what referrals look like in a Danish clinical setting, and it shows that these were not aligned with clinical guideline recommendations. To investigate the 'true' prevalence of guideline-appropriate referrals, future research should contain clinical information at the patient level, including precise information of duration and type of non-surgical treatment. Ideally, appropriate referrals should demonstrate high sensitivity for the detection of serious pathology with a reasonably high specificity to limit unnecessary imaging of patients without serious pathology.
Danish National Clinical Guidelines for non-surgical treatment of patients with recent onset LBP or lumbar radiculopathy [24] was published in 2016. Although it recommends against routine use of MRI for LBP with or without radiculopathy, there was no change in the distribution of appropriate and inappropriate MRI referrals from 2014 to 2018. From a clinical perspective, it seems timely to develop an implementation strategy of imaging guidelines for LBP among GPs to ensure that only patients with a clear indication for MRI are referred and to increase the quality of the referrals.
Methodological considerations
The strengths of this study are the large study sample and the use of a previously tested data collection method that had shown substantial to high inter-rater reliability. The data collection method was based on a well-documented international method for the assessment of imaging guideline appropriateness (ACR)[3]. To ensure a homogeneous practice of referral from general practice clinics and triage by the receiving department, the data were collected from a single imaging department located at a medium-sized hospital with a catchment that included both city and rural areas. However, this could at the same time be a weakness as it is possible that local routines and agreements exist that are not applicable to other parts of the country.
In everyday clinical practice, imaging referrals are not controlled and monitored against clinical findings and symptoms. Although this study cannot tell if the patients had an appropriate indication that was not reflected in the narrative text of the referral, the results do reflect that in everyday practice there is no evidence that guidelines are implemented in either GP referrals or at the receiving imaging department.