Inappropriate requests for magnetic resonance scans of the shoulder

Abstract

Purpose

Magnetic resonance imaging (MRI) scans are a useful investigation for some shoulder pathology. They are costly however and a significant burden on radiology departments. In most cases clinical examination, plain radiography or ultrasound scan (USS) are sufficient for a diagnosis. There are no current UK guidelines regarding MRI shoulder scan requests.

Methods

We reviewed 100 consecutive MRI shoulder scan requests and the associated formal reports; other investigations were also assessed.

Results

Overall, 56 % of MRI scans were ordered inappropriately. Shoulder consultant's requests were more appropriate than other groups (70 % vs. 38 %. p = 0.04). Excluding shoulder consultants 63 % of scans were inappropriately ordered. Shoulder consultants were more likely to order a preceding X-ray (80 % vs. 53 % respectively, p = 0.03). Of those with a clinical diagnosis of cuff pathology only 29 % had an USS.

Conclusion

A high percentage of MRI shoulder scans are performed inappropriately. Shoulder consultants are more appropriate in their ordering than other groups. If all groups performed as well 50 % less MRI scans would need to be performed.

Introduction

Magnetic resonance imaging (MRI) scans are commonly used to investigate shoulder problems [1, 2]. When investigating acute shoulder pain the American College of Radiology guidelines of 2010 recommend plain radiographs as the mainstay of investigations. They conclude that ultrasound scan (USS) is as effective as MRI in diagnosing cuff pathology and that MRI arthrogram (MRA) should be used in suspected instability [3].

The costs of MRI scans are significant, e.g. in England the NHS cost of a plain shoulder MRI is £153 (€180) and for MRI arthrogram £272 (€321) [4]. Costs elsewhere can be higher. In the United States a conventional shoulder MRI costs £1334 ($2,033) and MRA costs £1535 ($2,339) [5].

Alternative forms of investigation can be more cost effective. A plain shoulder X-ray (two views) costs £14.67 (€173) in our hospital and ultrasound of the rotator cuff £47 (€55) [4].

Given this information, it is important to make sure that the clinical indications for MRI and MRA requests of the shoulder are valid; currently up to 41 % of MRI scans ordered in the United States may be unnecessary [6].

The aim of our paper was to assess the validity of MRI and MRA requests in our hospital and to suggest guidelines to improve efficiency.

Methods

Study design

In our hospital we use the picture automated computer system (PACS) imaging system (Centricity PACS, GE Healthcare). We interrogated this database to retrieve 100 consecutive MRI shoulder scan requests submitted between 2nd October 2012 and 22nd January 2013. We obtained the data listed below from reviewing the data held on PACS (the scanned request form, the formal report of the scan and other imaging performed for that patient).

Data collected

1. The grade of doctor requesting the MRI

2. The diagnosis (indication for MRI) on the MRI request form

3. Investigations performed prior to the MRI request (X-ray, CT and USS)

4. Use of contrast in the MR scan

5. The diagnosis in the MRI report

6. Appropriateness of request (see below)

Appropriateness of request

We reviewed the literature regarding the most appropriate investigation for common shoulder problems and the accuracy of MR imaging in each case. The findings are summarised in Table 1. Based on this data we used the following definition of appropriate investigation.Patients with shoulder pathology should first have a clinical examination by an experienced clinician and an initial X-ray. Suspected rotator cuff pathology should be investigated with USS and instability should be investigated with MRA. MRI scan with intravenous contrast should be ordered for investigating tumours and infections [7].

Table 1.

Summary of evidence for each investigation

Condition	Investigation of choice	Sensitivity and specificity for investigation of choice		Sensitivity and specificity for MRI scan
		Sensitivity	Specificity	Sensitivity	Specificity
Instability	MRA	88 % [8]	93 % [8]	76 % [8]	87 % [8]
Cuff tear	USS	85 % [9]	92 % [9]	86 % [9]	90 % [9]
SLAP	Clinical examination	90 % [10]	97 % [10]	38 % [11]	94 % [11]
Biceps pathology	USS	49 % [12]	97 % [12]	27 % [13]	94 % [13]

Definitions

MRI and MRA were considered inappropriate unless there was a clinical diagnosis of instability, tumour or infection. MR1 for other diagnoses were only considered appropriate if other preceding investigations had failed to give a diagnosis.

Statistics

We compared groups using the Chi squared test. Statistical significance was set at the 5 % level.

Results

Of the 100 scan requests all but three of the scans were performed. Three scans were not performed because of patients failing to attend. We therefore obtained data regarding 100 requests and 97 completed scans.

The grade of doctor requesting the MRI is shown in Table 2.

Table 2.

Grade of requestor

Requestor	Total
Orthopaedic consultant—shoulder specialist	20
Orthopaedic consultant—not a shoulder specialist	20
Orthopaedic middle grade (registrar or staff grade)	17
Physician consultant	16
General practitioner	21
Physiotherapist	4
Unknown	2
Total	100
Requestor sub-groups	Total
All non-specialist	80
Specialist	20

The diagnoses from the MRI request forms are shown in Table 3.

Table 3.

Clinical diagnoses

Clinical diagnosis	Number
Instability	27
Cuff tear	25
Pain	21
Tumour	10
Frozen shoulder	5
Infection	3
Calcific tendonitis	2
No diagnosis	2
LHB pathology	2
Other (OA, fracture, SLAP)	3
Total	100

Investigations performed prior to the scan request (X-ray, CT and USS)

Plain radiography

Only 44 % (44/100) of patients had an X-ray prior to MRI scan. Compared to all others requesting scans, shoulder consultants were more likely to order a preceding X-ray (80 % vs. 53 % respectively, p = 0.03).

Computed tomography

Only 7 % (7/100) of the cohort had a CT scan of the shoulder. These revealed instability related fractures (6/7) and one tumour.

Ultrasound

Of those with a clinical diagnosis of cuff pathology only 29 % had an USS prior to MRI scan of the shoulder. Only 32 % (8/25) of MRI scans ordered for a rotator cuff tear had this diagnosis in the report. In the 14 who had an ultrasound scan as well as an MRI the imaging diagnosis was the same in 71 % (10/14).

Use of contrast in the MR scans

Intra-articular contrast was used in 29 scans, 22 for suspected instability, five for cuff tears, two for pain and one for a SLAP lesion. Intravenous contrast was used five times, all for suspected tumours. No contrast was used in the remaining 66 scans.

Diagnosis in the scan report

This data obtained from the written report is given in Table 4.

Table 4.

Scan report diagnosis

MRI primary diagnosis	Number
Normal	37
Partial cuff tear	11
Instability	17
SLAP	6
Complete cuff tear	5
Non-malignant tumour	5
Not performed	3
Inconclusive	3
OA	3
Fracture	2
ACJ disease	2
Infection	2
Malignant tumour	2
Calcific tendonitis	1
LHB pathology	1
Total	100

Appropriateness of request

Overall 56 % of MRI requests were inappropriate. Shoulder consultant's requests were more likely to be appropriate than other groups (70 % vs. 38 %, p = 0.04). Excluding shoulder consultants 63 % of scans were inappropriately ordered. In total, non-shoulder consultants ordered 50 scans inappropriately (Table 5).

Table 5.

Appropriateness by requestor grade

Requestor	Appropriate	Inappropriate	Total
Orthopaedic consultant—shoulder specialist	14 (70 %)	6 (30 %)	20
Orthopaedic consultant—not a shoulder specialist	5 (25 %)	15 (75 %)	20
Orthopaedic middle grade	10 (59 %)	7 (41 %)	17
Physician consultant	4 (25 %)	12 (75 %)	16
General practitioner	8 (38 %)	13 (62 %)	21
Physiotherapist	2 (50 %)	2 (50 %)	4
Unknown	1 (50 %)	1 (50 %)	2
Total	44 %	56 %	100
Requestor sub-groups
All non-specialist	30 (38 %)	50 (62 %)	80
Specialist	14 (70 %)	6 (30 %)	20

Discussion

Cost analysis

Imaging costs for the 100 patients included in our study were £20,868 (66 MRI, 29 MRA, five MRI with contrast, 29 USS and five CT scans at £87). If the patients in our study had been investigated according to the guidelines suggested above the total cost of investigations would have been £12,154 (Table 6). Over the course of a year this equates to a potential saving of £36,599 (£51,047 vs. £87,646) in our hospital.

Table 6.

Cost analysis of appropriate investigations

Clinical diagnosis	Number	Cost of appropriate investigation
Instability	27	MRA @ £272 = £7344
Cuff tear	25	USS @ £47 = £1175
Pain	21	aUSS @ £47 = £987
Tumour	10	MRI with contrast @ £182 = £1820
Frozen shoulder	5	Clinical examination and X-ray
Infection	3	MRI with contrast @ £182 = £546
Calcific tendonitis	2	USS @ £47 = £94
No diagnosis	2	aUSS @ £47 = £94
LHB pathology	2	USS @ £47 = £94
Other (OA, fracture, SLAP)	3	Clinical examination and X-ray
Total		£12,154

^aOf the 23 with a diagnosis of pain or no diagnosis 100 % (23/23) had MRI diagnoses that would have been better demonstrated by X-ray, clinical examination or USS

Reducing inappropriate investigations

Innattoi and Willams found 41 % inappropriate MRI shoulder requests from non-shoulder specialist physicians [6]. In our study the overall rate of inappropriate MRI shoulder requests was 56 %. We found shoulder specialists to be significantly more appropriate in requesting MRI scans than other groups (70 % vs. 38 %, p = 0.04). Our study supports the limiting of MRI requests to specialist shoulder surgeons. If all of the clinical diagnoses had been investigated appropriately the savings for our department would have been £36,599 (€43,187) per annum.