Overuse of Magnetic Resonance Imaging in the Diagnosis and Treatment of Moderate to Severe Osteoarthritis

Abstract

Background

MRI in the evaluation of end-stage knee joint osteoarthritis (OA) is usually unnecessary when radiographic and clinical evidence of gonarthrosis is clear. The purpose of this study was to assess the prevalence of MRI scans ordered in patients with radiographically obvious gonarthrosis and to examine the characteristics of health care providers who ordered these imaging studies.

Methods

We retrospectively identified 164 patients diagnosed with moderate to severe OA who were referred for total knee replacement (TKA) over a one-year period. The percentage of patients who had an MRI scan with or without X-ray, within the preceding 3 months prior to referral, were calculated. Subgroups were analyzed to identify characteristics that may influence the decision to order an MRI, including K-L grade, provider type, level of training, and practice location.

Results

Of 145 patients, 19 (13.1%) presented with an MRI scan. Between the number of MRI scans ordered, there was a significant difference when comparing physicians versus non-physicians, with physicians ordering less MRI scans (p=0.018). There was a significant difference when comparing non-academic versus academic, with academic providers ordering less MRI scans (p=0.044). There was no significant difference with fellowship training or provider proximity to our academic institution.

Conclusions

In this study, 13.1% of patients with radiographically obvious knee OA obtained an MRI prior to referral for TKA. Non-physicians and non-academic physicians were more likely to order MRI scans. Improved education for referring providers may be necessary to decrease overuse of MRI in the diagnosis of moderate to severe arthritis.

Level of Evidence

Level II

Introduction

The disease burden of knee osteoarthritis (OA) affects an estimated 27 million people.¹ More than 600,000 of these patients undergo total knee arthroplasty (TKA) annually in the United States.^2,3 With such a high prevalence of disease and surgical treatment, it is critical to identify cost-effective strategies for accurately diagnosing and managing moderate to severe OA. Health care expenditures continue to increase and the overuse of diagnostic imaging including magnetic resonance imaging (MRI) is a significant contributor to costs.^4,5

Clinical examination of the patient with plain radiography is 91% sensitive and 86% specific for diagnosing knee OA.⁶ The diagnostic accuracy of this combination increases with worsening severity of OA. One meta-analysis calculated that the sensitivity and specificity of MRI for making the diagnosis of arthritis to be 61% and 82%, respectively.⁷ While MRI has the ability to visualize abnormalities in OA patients not present on radiography – including bone marrow lesions, ligamentous damage, and meniscal tears^8-12 – the clinical relevance of these findings and implications for surgical treatment in severe OA are not well understood. For example, MRI can detect meniscal tears, but these are frequent findings in patients with radiographic evidence of OA, with no difference in prevalence among those with and without symptoms.^13-15 Similarly, the presence of undiagnosed ACL tears among patients with OA is not associated with increased pain or functional instability.^16,17 Therefore, MRI has little to no role in the initial evaluation of patients with radiographic evidence of moderate to severe arthritis, and similarly minimal role in the pre-surgical planning or decision making prior to TKA.

Previous studies have demonstrated MRI overuse for the diagnosis and treatment of shoulder and foot and ankle pathology.^18,19 Our goal was to examine MRI overuse in patients with moderate to severe knee OA referred to our institution for TKA. Additionally, we sought to examine the characteristics of the providers who ordered these MRI scans. Our hypothesis was that many providers order MRI for evaluation of OA prior to referring to an orthopedic surgeon, and that providers with higher levels of training are less likely to order these imaging studies.

Methods

With Institutional Review Board approval, we retrospectively examined the records of consecutive patients referred for TKA to one high-volume arthroplasty surgeon at a large academic institution over a 12-month period. All new patients with an ICD-9 code for OA (715) were identified. Knee radiographs were graded in severity using the Kellgren-Lawrence (KL) scale taking into account osteophyte formation, joint space narrowing, and subchondral sclerosis.²⁰ Patients with a KL grade 3 (moderate) or 4 (severe) in any compartment were included in the analysis. Exclusion criteria consisted of: 1) recent trauma, 2) history of systemic inflammatory disorder, and 3) previous ipsilateral knee arthroplasty. We next identified all patients within this cohort who had a knee MRI ordered by their referring provider within the 3 months prior to the index surgical consultation.

Subgroup analysis evaluated for characteristics which may have influenced the decision to order an MRI prior to referral. These included: Kellgren-Lawrence grade (III versus IV), physician versus non-physician (physician assistant or nurse practitioner) provider, non-academic versus academic physician, fellowship trained versus non-fellowship trained physician, and distance of referring provider from our institution.

Statistical analysis consisted of applying Fisher’s exact test to compare characteristics between those presented with MRI versus those who did not, with statistical significance set at p < 0.05.

Results

Of the 164 patients identified who met study criteria, 19 patients were excluded because no referring provider was listed. Subgroup analysis was therefore performed on a total of 145 patients (19 patients presenting with MRI (13.1%), 126 patients presenting without MRI). Provider demographic information can be found in Table 1. There were a total of 94% patients referred from a physician and 6% referred from a non-physician. There were 24% referred from academic providers and 76% from non-academic providers; 20% of the physicians were fellowship trained and 80% were non-fellowship trained. A total of 45% of the patients were referred from a provider greater than 40 miles from our academic institution.

Table I.

Subgroups of providers with the amount of patients presenting with an MRI vs. without an MRI

	MRI	Without MRI	Total
Physician	15	121	136
Non-Physician	4	5	9
Academic	1	34	35
Non-Academic	18	92	110
Fellowship Trained	4	23	27
Non Fellowship	15	103	118
≥40 miles	11	54	65
<40miles	8	72	80

Subgroup analysis demonstrates that a greater proportion of patients with KL grade 3 presented with MRI compared to those with KL grade 4 (16.1% vs. 10.0%, p=0.32) with no statistical difference found between the groups. (Figure 1)

Figure 1.

Graph shows that there was no difference between KL grade with the patients who presented with an MRI.

Out of the 9 patients who were referred from a non-physician provider, 44% presented with an MRI. Of the 136 patients referred from a physician provider, 11% presented with an MRI. The unequalized odds ratio of non-physicians ordering an MRI in our population was 6.45. When comparing physicians versus non-physicians, there was a statistically significant difference between the number of MRIs ordered based on provider type (p=0.018) (Figure 2).

Figure 2.

Characteristics of the percentage of patients referred with an MRI. * indicates significant difference p < 0.05. Physician vs. Non-physician: p < 0.018; Academic vs. Non-academic: p < 0.044.

Only 3% of the patients who were referred from an academic provider presented with an MRI as compared to 16% referred from non-academic providers. The unequalized odds ratio of non-academic providers ordering an MRI in our population was 6.65. When comparing non-academic providers to academic providers, there was a statistically significant difference between the number of MRI orders based on academic affiliation (p=0.044).

Of the 27 patients referred from a fellowship trained physician, 14% presented with an MRI as compared with 10% of the 109 patients referred from a non-fellowship trained physician. When comparing fellowship trained versus non-fellowship trained there was no statistical significance in the number of MRIs ordered based on the physician fellowship status (p=0.50).

Out of the 65 patients who presented from a provider greater than 40 miles from our academic institution, 17% presented with an MRI as compared to 10% patients who presented from a provider within 40 miles from our academic institution. When comparing the number of MRIs ordered based on provider distance, we found no statistical significance (p=0.32) between the groups.

Discussion

The financial burden of MRI imaging in the diagnosis and evaluation of pathologic orthopaedic conditions is well recognized.^21-23 We examined the incidence of MRI imaging ordered for patients with radiographically obvious OA prior to referral to an academic arthroplasty surgeon. We found that 13.1% of patients referred for knee arthroplasty had a MRI scan despite knee radiographs demonstrating K-L grade III-IV OA. Non-physician and non-academic providers were more likely to order MRI scans when compared with physicians or academic providers.

As more non-physicians are engaged in the initial management of patients, cost savings from using non-physicians may be overcome by the increased use of imaging, such as knee MRI scans.²⁴ We did not find that MRI use differed by referring physician specialty, or by referring physician fellowship training. More studies may shed light on the relationship, if any, between the incidences of knee MRI scans between providers with different levels of training.

Referring providers employed by the academic institution itself were less likely to order MRI (27%). It is possible that early communication between providers may have clarified what type of imaging was preferred at the time of referral. Such communication is often easier amongst providers within the same institution or setting via electronic medical records or more direct methods of interaction. Providers from an outside institution often do not have the liberty and freedom to easily access the tertiary specialist to whom they refer.

No patient referred with knee OA was felt to require a MRI for diagnosis, at least when evaluated by the orthopaedic surgeon. Knee MRI scans are usually reserved for younger patients without radiographically apparent OA, in whom the diagnosis is unclear. Bernstein et al. found that 45% of the knee MRI scans ordered by physicians outside the orthopedic specialty (non-orthopedic surgeons) were normal or showed only OA, compared to 27.6% of the scans ordered by orthopedic surgeons.²⁵ These results suggest that non-orthopedic physicians may use knee MRI scans to screen painful knees for a diagnosis more often than orthopedic surgeons.²⁶

Song et al. found a pre-obtained knee MRI rate of 27% in 185 of 680 patients. Their results suggest a “useful MRI” rate was assessed in sports-related injury (84%) than in degenerative joint disease (18%). The study, however, does not contain practice patterns of referring physicians.²⁷

There are several limitations to this study. Our study was limited to the patients referred to one surgeon and it is possible the findings may have been different if patients referred to other arthroplasty surgeons were included. Also, the academic center was in a semi-rural community with many referrals from adjacent rural clinics; the average distance travelled by the patients in this study was 53.7 miles. Referring physicians may have ordered knee MRI scans simply to expedite treatment and reduce the burden of travel on their patients. We did not identify any differences between the distances travelled by patients with and without knee MRI scan prior to referral (Table 2). Further studies may identify if patient convenience and expediency of treatment are related to referring physician readiness to order a knee MRI prior to evaluation by a surgeon.

Table II.

Referral characteristics of patients in the specifically defined population

	With MRI	Without MRI
KL Grade 3	50.0%	37.0%
KL Grade 4	50.0%	63.0%
Age	65.68	63.49
Average Distance	55.96	53.39
Prop MU physician	9.0%	0.27*
Prop fellowship trained	21.1%	18.3%
Physician specialty
Family Practice	47.0%	64.0%
Internal Medicine Orthopedic Surgery	11.0% 16.0%	20.0% 7.0%
Other	5.0%	5.0%
Non-Physician	21.0%	0.04*
Unavailable	1	18

^*= statistically significant prop = proportion

At our institution, outside imaging is brought by the patient to the visit, and the images are uploaded to the electronic record. Since some patients may not have brought pre-referral imaging with them, it is possible that the incidence of MRI scans was underestimated in the patient cohort studied. Since only 9.7% of referred patients had knee imaging (X-rays or MRI) performed before the first visit with an arthroplasty surgeon, some pre-referral imaging studies may have been missed in our analysis.

Our study showed that 13.1% of patients with moderate-to-severe knee OA (KL grade 3 and grade 4, respectfully) presented to our arthroplasty surgeon with an MRI ordered by the referring provider. Non-physician providers and non-academic providers were more likely to order knee MRI scans in patients with radiographically identifiable knee OA. While prospective clinical data is needed to corroborate these observations, provider education in the appropriate use of MRI scans in patients with radiographically obvious OA may be a cost-effective strategy.