Radiography (2010) 16, 136e142

ava i lab le a t www.sc iencedi rec t .com

journa l homepage : www.e lsev i er . com/ loca te / rad i

MRI reporting by radiographers: Findings of anaccredited postgraduate programme

Keith Piper a,*, Kaie Buscall a, Nigel Thomas b

a Allied Heath Professions Department, Canterbury Christ Church University, North Holmes Road, Canterbury,Kent CT1 1QU, UKb X-Ray Department, Trafford General Hospital, Manchester M41 5SL, UK

Received 28 October 2008; revised 28 October 2009; accepted 30 October 2009Available online 25 November 2009

KEYWORDSMagnetic resonanceimaging;Lumbar spine;Knee;Internal auditorymeatus training;Agreement;Radiographer;Radiologist;Interobserver variation

* Corresponding author. Tel.: þ44 1E-mail address: keith.piper@cante

1078-8174/$ - see front matter ª 200doi:10.1016/j.radi.2009.10.017

Abstract Aim: To analyse the objective structured examination (OSE) results of the firstthree cohorts of radiographers (n Z 39) who completed an accredited postgraduate certificate(PgC) programme in reporting of general magnetic resonance imaging (MRI) investigations andto compare the agreement rates with those demonstrated for a small group of consultant radi-ologists.Method: Forty MRI investigations were used in the OSE which included the following anatom-ical areas and abnormal appearances: knee; meniscal/ligament injuries, bone bruises, effu-sions and osteochondral defects; lumbar spine: intervertebral disc morphology, vertebralcollapse, tumours (bone and soft tissue), spinal stenosis and/or nerve root involvement;internal auditory meati (IAM): acoustic neuroma. Incidental findings included maxillary polyp,arachnoid cyst, renal cyst, hydroureter, pleural effusion and metastases (adrenal, lung, peri-renal and/or thoracic spine). Sensitivity, specificity and total percentage agreement rateswere calculated for all radiographers (n Z 39) using all reports (n Z 1560). A small represen-tative subgroup of reports (n Z 27) was compared to the three consultant radiologists’ reportswhich were produced when constructing the OSE. Kappa values were estimated to measureagreement in four groups: consultant radiologists only; radiographers and each of the consul-tant radiologists independently.Results: The sensitivity, specificity and agreement rates for the three cohorts (combined) ofradiographers were 99.0%, 99.0% and 89.2%, respectively. For the majority (5/9) of anatomicalareas and/or pathological categories no significant differences (p< 0.05) were found betweenthe mean Kappa scores (K Z 0.47e0.76) for different groups of observers, whether radiogra-phers were included in the group analysis or not. Where differences were apparent, thiswas in cases (4/9) where the variation was either not greater than found between radiologistsand/or of no clinical significance. These results suggest therefore that in an academic setting,

227 782425; fax: þ44 1227 451739.rbury.ac.uk (K. Piper).

9 The College of Radiographers. Published by Elsevier Ltd. All rights reserved.

MRI reporting by radiographers 137

these groups of radiographers have the ability to correctly identify normal investigations andare able to provide a report on the abnormal appearances to a high standard. Further work isrequired to confirm the clinical application of these findings.ª 2009 The College of Radiographers. Published by Elsevier Ltd. All rights reserved.

Introduction

The role of the radiographer in the reporting of magneticresonance imaging (MRI) investigations is still a relatively newconcept although in the United Kingdom (UK) this practice isnow receiving growing attention. It has been a number ofyears since the Audit Commission recommended that radi-ographers’ roles could develop into this area of practice1 andthe considerable challenges posed by recent guidance2

suggest that some clinical imaging departments may find itdifficult to meet such targets as the 18 week patient pathwaywithout wider introduction of effective skill mix. TheNational Diagnostics Imaging Board (NDIB) recognises this asa key challenge and recommends that there should bea ‘focus on roles and responsibilities, skills and competencesto facilitate better use of skill mix’ (2007, p. 5).3 Mostrecently, the NDIB has published more challenging guidancewhich recommends that ‘imaging services should aim toprovide reporting turn around times as follows: urgent casese Immediate (within 30 min); Inpatients and A&E e Sameworking day; All other cases e By next working day’.4

Some departments have responded to these recommen-dations and have already introduced initiatives which includethe reporting of certain MRI investigations by radiographerswho have received appropriate education and training.5,6 Apostgraduate programme (PgC Clinical Reporting e MRIGeneral Investigations) accredited and approved by theCollege of Radiographers (UK)/Canterbury Christ ChurchUniversity,7 described previously,8 has recruited nationally,annual cohorts of radiographers wishing to develop their skillsin reporting of MRI investigations of the knee, lumbar spineand internal auditory meati (IAM) since 2003. The 12-monthworkplace based programme consists of short, two day,briefing blocks held at the university approximately everytwo months. Experienced MRI consultant radiologists areinvolved in the design, management, teaching and assess-ment aspects of the programme. The assessment scheduleincludes a case-study, an assignment which requires studentsto critically reflect on their developing competence in MRIreporting and 500 practice reports, 125 of which must bechecked by a consultant radiologist mentor in the students’workplace. One of the final summative assessments for thePgC is an Objective Structured Examination (OSE) whichconsists of 40 MRI investigations.

The aim of this study was to analyse the OSE resultsachieved by the first three cohorts consisting of a total of 39radiographers who completed the PgC.

Objectives

(1) To present and evaluate the results of the first threecohorts of radiographers who completed the PgC pro-gramme, for the different anatomical areas (knee,

lumbar spine and internal auditory meatus) included inthe OSE.

(2) To compare the agreement rates (by weighted andunweighted Kappa values), for a small representativesubgroup of radiographers with the rates of a smallgroup of radiologists.

Method

As part of the summative assessment at the end of the PgCprogramme the radiographers were required to report 40general MRI investigations in the form of an OSE, theconstruction of which is described in detail previously.8

During construction of the OSE, 72 MRI investigations (25knee; 29 lumbar spine; 18 [IAM] internal auditory meati)were reported independently by three groups of experi-enced consultant radiologists employed within the UK, andwho routinely report MRI investigations as part of theirclinical role. As reported earlier, the extent to which allthree reports agreed on the appearances demonstratedvaried, as follows: knee; 68e96%; lumbar spine; 78e99%:IAM; 100%, dependent on the anatomical area and/orpathological category.8 Due to the poor agreement, 16investigations (knee Z 9; lumbar spine Z 7) were excludedfrom possible use in the OSE banks. Five banks of eight MRIinvestigations were randomly selected from the 56 poten-tial cases ensuring that the prevalence of abnormal tonormal cases approximated 50% and that the anatomicalareas were represented similarly. The specific casesselected were randomised and rotated for each of the threecohorts but typically the actual number of investigations,for each anatomical area, in each of the OSEs, was asfollows: knee; 11e13, lumbar spine; 14e16 and IAM; 12e14.

Expected answers (based on the three previous inde-pendent radiological reports),8 for each of the 40 investi-gations selected for the OSE, were then agreed by theprogramme team (KP and KB) and one of the externalexaminers (a consultant radiologist [NT] experienced in MRIreporting), who also confirmed that an appropriate selec-tion of discriminatory cases were included.9 A range of caseswere included to adequately test the depth and breadth ofthe candidates’ knowledge and to demonstrate competenceand excellence at postgraduate level. Typical abnormalappearances included: knee; meniscal/ligament injuries,bone bruises, effusions and osteochondral defects, lumbarspine; intervertebral disc morphology (bulge, protrusion,extrusion, sequestration, annular tear), vertebral changes(Modic, collapse), tumours (bone and soft tissue, includingmetastases) e with and without cord compression, spinalstenosis and/or nerve root involvement, IAM; acousticneuroma and polyps. Other incidental findings, particularlyin the lumbar spine category, included arachnoid cyst, renalcyst and hydroureter; pleural effusion and metastases(adrenal, lung, perirenal and/or thoracic spine).

Table 1 Pairs of observers.

Group A

Consultant radiologist 1 (CRI) and consultant radiologist 2(CR2)Consultant radiologist 1 (CRI) and consultant radiologist 3(CR3)Consultant radiologist 2 (CR2) and consultant radiologist 3(CR3)

Group B

Consultant radiologist 1 (CRI) and reporting radiographer 1(RRI)Consultant radiologist 1 (CRI) and reporting radiographer 2(RR2)Consultant radiologist 1 (CRI) and reporting radiographer 3(RR3)

Group C

Consultant radiologist 2 (CR2) and reporting radiographer 1(RRI)Consultant radiologist 2 (CR2) and reporting radiographer 2(RR2)Consultant radiologist 2 (CR2) and reporting radiographer 3(RR3)

Group D

Consultant radiologist 3 (CR3) and reporting radiographer 1(RRI)Consultant radiologist 3 (CR3) and reporting radiographer 2(RR2)Consultant radiologist 3 (CR3) and reporting radiographer 3(RR3)

138 K. Piper et al.

The candidates were provided with the patient’s details(age, gender and referral source) and the clinical historyprovided at the time of the original MRI investigation, andwere asked to make a decision as to whether the appear-ances were normal or abnormal, recording this decision onthe examination sheet proforma by ticking a box. Studentswere advised to classify normal variants as normal. If theappearances were considered to be abnormal the studentwas requested to provide brief key details on the abnormalappearances and to include the suggested pathology(ies),where applicable, in the form of a free text hand-writtenreport. As part of the general guidance provided during thePgC programme students were also advised to be specificwhen stating the precise anatomical area, to includeabnormal signal descriptions and to measure and state thesize of any mass or cyst.

Marking criteria for the OSE

The radiographers’ responses were compared to theexpected answer by two members of the programme team(KP and KB). If the examination was correctly identified asnormal or abnormal a true negative or true positive scorewas allocated, respectively. If the case was marked incor-rectly as normal, or abnormal, a false negative or falsepositive score was recorded. The overall sensitivity andspecificity was then calculated.10.

One mark for each normal and a maximum of five marks foreach abnormal case was allocated and fractionated10 wherenecessary to reflect the different key aspects that wererequired in each report. For example, the mark allocation foran abnormal IAM case was as follows: right sided 3e4 mmmasswith low signal intensity (2 marks), intracannicular andenhancement following the introduction of Gadolinium(2 marks), acoustic neuroma (1 mark). For an abnormallumbar spine case for example, the following key points wereidentified: lesion at level T11 [and appropriate signaldescription] (1 mark), extending posteriorly causing cardcompression (1 mark) and extending into T10 (1 mark),consistent with metastatic deposit (1½ mark). Pleural effu-sion (1⁄4 mark) and perirenal metastases (1⁄4 mark) alsonoted. An abnormal knee case had the following key pointsidentified in the marking scheme: horizontal tear (1 mark),posterior third (1⁄4 mark), medial meniscus (1⁄4 mark),complete tear of posterior cruciate ligament (1½ marks),partial tear of medial collateral ligament (1½ marks) andappropriate signal descriptions (½ mark).

The candidates score for each case was reduced accord-ingly if all expected elements were not included and/or iffeatures were described as abnormal, which had beenconfirmed normal by the three radiologists’ reports, theprogramme team and the external examiner. All scores weresummed and the overall agreement percentage calculated.

The mean percentage agreement, standard deviationand 95% confidence intervals were calculated for each setof cohort scores and for each anatomical area (1560 reportsin total), as follows:

(a) knee;(b) lumbar spine;(c) IAM; and(d) all investigations.

Point estimates and 95% confidence intervals were alsoproduced for the sensitivity and specificity rates followingmultilevel logistic regression (MLWin 2.10).11e13

A small subgroup of 81 reports (produced by threeradiographers during the OSE); representing the minimum(85.2%), maximum (94.2%) and mean (88.5%) total agree-ment scores achieved by the most recent cohort (2006) ofradiographers, were compared to the three consultantradiologists’ reports used when constructing the OSE (knee;11, and lumbar spine; 16).8

Weighted and unweighted Kappa values were calcu-lated, as appropriate, and as described previously,8 foreach pair of observers’ reports shown in Table 1

The following anatomical areas/pathological categorieswere included in the Kappa analysis and the total numberof sites for each area/category are included in parentheses:11 knee investigations; ligament injury (44: anteriorcruciate, posterior cruciate, medial collateral and lateralcollateral ligaments), meniscal tear (88: medial and lateral;anterior; middle, posterior and bucket handle type/complex), bone bruise (22:femur and tibia), and effusion(11); 16 lumbar spine investigations (80 sites: five verte-bral/intervertebral disc levels) for all areas/categories);disc (bulge/protrusion/extrusion], nerve root involvement,degenerative disc disease, tumour and other incidental

Table

2M

ean

agr

eem

ent,

sensi

tivi

tyand

speci

fici

tyra

tes.

Anato

mic

al

are

a20

04Cohort

(nZ

10)

2005

Cohort

(nZ

11)

2006

Cohort

(nZ

18)

Tota

l(n

Z39

)

Mean

(%)

95%

CI

SDM

ean

(%)

95%

CI

SDM

ean

(%)

95%

CI

SDM

ean

(%)

95%

CI

SD

Knee

88.0

79.8

e95

.25.

287

.377

.7e

97.6

5.7

84.8

73.9

e92

4.8

86.3

73.9

e97

.65.

3Lu

mbar

spin

e85

.477

.9e

92.2

4.3

90.1

84.9

e96

.94.

086

.577

.1e

96.4

5.9

87.2

77.1

e96

.95.

2IA

M98

.794

.0e

100

1.9

98.8

96.0

e10

01.

297

.994

.0e

100

2.0

98.4

94.0

e10

01.

8Tota

lagr

eem

ent

%89

.284

.5e

92.0

2.4

90.8

86.5

e95

.93.

488

.284

.4e

94.2

2.7

89.2

84.4

e95

.93.

0Se

nsi

tivi

ty%

98.5

*2.

499

.5*

1.5

98.9

*2.

199

.0*

97.9

7e

99.4

9*2.

0Sp

eci

fici

ty%

98.5

*2.

499

.5*

1.5

98.9

*2.

199

.0*

97.9

7e

99.4

9*2.

0

Ast

eri

ksin

dic

ate

no

sign

ifica

nt

dif

fere

nce

sin

sensi

tivi

tyor

speci

fici

tyra

tes

(pZ

0.05

)betw

een

dif

fere

nt

cohort

s.

MRI reporting by radiographers 139

findings. The Kappa values (CR1/CR2/CR3; CRI and RRs; CR2and RRs; and CR3 and RRs) were then compared using theAnalysis of Variance (ANOVA) statistic and the Tukeymultiple comparison test14 where any significant differ-ences were identified between groups A, B, C and/or D.

Results

All radiographers included in this study had a minimum oftwo years MRI experience and no previous reporting expe-rience on other modalities.

The mean agreement, sensitivity and specificity rates foreach cohort, and for each anatomical area, are shown inTable 2. The highest mean % agreement rates demonstratedby all cohorts were for IAM investigations: all cohorts: 98.4%(94.0e100%). The mean rates for the knee and lumbar spineinvestigations were 86.3% (73.9e97.6%) and 87.2% (77.1e96.9%), respectively. For 2 of the 3 cohorts the lumbar spineagreement rates are slightly higher than those seen for theknee investigations. The mean % agreement rate for allanatomical areas and for all cohorts combined was approxi-mately 90% (89.2%: 88.4e95.9%). Mean sensitivity and spec-ificity % rates for all cohorts were both 99.0% (97.97e99.49%).The lowest mean scores for each anatomical area, and for allcohorts combined, are listed in Table 3 indicating the keyabnormal appearances outlined in the expected answer.

Tables 4 and 5 include the Kappa values (11 kneeinvestigations, 16 lumbar spine investigations) for thegroups of observers shown in Table 1.

Analysis of Variance and Tukey multiple comparisontests are also included in Tables 3 and 4.

Discussion

The sensitivity and specificity rates for the three cohorts ofradiographers were all high (mean rate e all cohorts; 99.0%).The majority of radiographers (cohort 2004, 7/10; cohort2005, 10/11; and cohort 2006, 14/18) achieved sensitivityand specificity rates of 100% demonstrating they couldcorrectly identify all the abnormal investigations included inthe OSE. The remaining radiographers, 9 in total, achievedrates of 95% and successfully passed that element of theassessment for which the pass mark had been agreed at 90%.8

This is an encouraging finding as it is important for anypractitioner interpreting clinical images as part of their role,to be able to differentiate between normal and abnormalexaminations to a high level.

The mean % agreement rates (86.3%; 73.9e97.6%) ach-ieved by the radiographers, for the knee investigations inthis study compares reasonably with the rates found in thePhase Two analysis of an earlier study,8 which reported onthe construction of the OSE, and found that, for the kneecases, the range of overall agreement rates between threeconsultant radiologists was 68e96%.

Table 3 outlines the cases where, in this study, candidateslost the majority of their marks. It is interesting to note,however, that reporting a knee effusion produced onlymoderate agreement between the radiologists in Phase Oneand Two of the earlier study (K Z 0.30; K Z 0.43, respec-tively) and only fair agreement previously.15 This suggeststhat radiologists and radiographers may have a tendency to

Table 3 Common errors in interpretation.

Anatomicalarea

Mean score(Max. 5 marks)

Outline expected answer

Knee1 3.54 Multiloculated cyst

(synovial cyst/ganglion),[head of gastrocnemius].

2 3.60 Tear posteriorcruciate ligament, [effusion].

3 4.07 Tear postthird medial meniscus,tear posterior cruciateligament, [partial tearmedial collateralligament, effusion].

Lumbar spine1 3.58 Multiple metastases,

multilevel disc degeneration,[scoliosis, hydronephrosisleft kidney].

2 3.68 Partial [osteoporotic]collapse (LV1), discprotrusion (L4/L5 and L5/S1),with nerve root involvement,[hydroureter].

3 4.27 Metastatic deposit (T11)with cord compression,[lung and adrenalmetastases].

IAM1 4.8 3 mm intracannicular

acoustic neuroma,[maxillary polyp].

Text in italics, in brackets indicates where greatest proportionof marks lost per case.

140 K. Piper et al.

over or under report the presence of a knee effusion,however, this inconsistency is probably of little, or no clinicalsignificance. In addition, at least one of the consultantradiologists failed to specify the site (head of gastrocnemius)of the multiloculated cyst in the knee case which alsoproduced the lowest score when reported by the radiogra-phers in the OSE.

The partial tear of the medial collateral ligament wasnot identified by all the radiographers, however, in previousstudies the agreement between radiologists when inter-preting this area has been moderate at best16 (K Z 0.50e0.58) and sometimes poor.15

With reference to the lumbar spine case identified inTable 3 it can be seen that most marks were lost for failingto note additional incidental findings. The reasons for thisare unclear and there appear to be no studies which haveinvestigated this phenomenon in detail. The effect,however, is not limited to reporting by radiographers. In theprevious study, the Phase One mean Kappa agreement rateof 0.30 increased only marginally in Phase Two (K Z 0.36;0.23e0.58) for incidental findings, when the reports of threeconsultant radiologists were compared.8 In the reportsincluded in this study only one of the three radiologists hadreported the hydronephrosis and only two had noted the

hydroureter and metastases (lung and adrenal) in cases 2and 3 outlined in Table 3. The maxillary polyp, demon-strated on one of the IAM cases, was also only noted by twoof the radiologists, and again this abnormality, which was‘missed’ by some radiographers, is of no real clinicalsignificance. Magnetic resonance imaging is very sensitivefor sinus disease meaning again it has a tendency to be overor under reported depending on the threshold of the indi-vidual radiologist, or in this case, the radiographer.

Kappa agreement rates e knee

The Kappa agreement rates (knee investigations) for thedifferent groups of observers (Table 4) were broadly similarfor the pathological categories of ligament injury andmeniscal tear and for these two categories ANOVA testsrevealed no significant differences between the differentgroups (CRs only; CR1 and RRs;CR2and RRs; and CR3 and RRs);p Z 0.67 (ligament injury) and p Z 0.34 (meniscal tear). Themean agreement for one group (CR2 and RRs) was slightlyhigher but overall the introduction of the three radiographersinto the group analysis had no significant effect. The trend formeniscal tears is similar with one group (CR3 and RRs)demonstrating a slightly higher agreement rate.

Significant differences were found, however, for thecategories of bone bruise (p Z 0.0067) and effusion(p< 0.001). The Tukey analysis for the bone bruise cate-gories revealed that the agreement (K Z 0.2) found ingroup D (CR3 and RRs) was significantly lower than in allother groups (p Z 0.027, p Z 0.006, p Z 0.042) and thatthe rate for group B (CR1 and RRs) was significantly higherthan group A (CRs only; p Z 0.012). For the effusion cate-gory a similar trend emerges, however, in this case theagreement rates for group D (CR3 and RRs) were signifi-cantly higher (p< 0.01) when compared by the Tukey test.

Results for the fracture/osteochondral defect categorywere not included in this part of the analysis as only onecase was included where an osteochondral defect was partof the expected answer. For the particular investigationincluded in the OSE all three radiologists had previouslyagreed the appearances and the three representativeradiographers included here had all also correctly identifiedand diagnosed the osteochondral defect.

The wide variation in agreement rates found here for thepresence of bone bruise is unsurprising as other studieshave also only found moderate agreement (K Z 0.63)16 andmoderate to good agreement (bone surfaces)15 betweenconsultant radiologists.

Similarly, the range of Kappa values for effusion iscomparable for the majority of groups in this study to thatfound by Bryan who found poor to fair agreement.15 Theexception is group D (CR3 and RRs), where the agreement wasvery good.17 Six of the knee investigations had effusionincluded as part of the expected answer and CR3 consistentlynoted the presence of an effusion to a greater extent thanCR1 and CR2 who tended to comment less on this appearance.

Kappa agreement rates e lumbar spine

The Kappa agreement rates (lumbar spine investigations) forthe different groups of observers were broadly similar for

Table 4 Agreement between observers (CRs only; CR1/RRs; CR2/RRs; CR3/RRs). Knee (11 investigations)a.

Category Groups of observers ANOVAP value

Tukey multiplecomparison testP value (if applicable)

Group A Group B Group C Group D

CRs CR1 andRRs

CR2 andRRs

CR3and RRs

Ligamentb Injury(12/48)c

0.77 0.73 0.69 0.72 0.67 e

0.71 0.74 0.86 0.740.67 0.69 0.74 0.69

Mean 0.72 0.72 0.76 0.72

Meniscal Tear (6/88)c 0.75 0.71 0.64 0.71 0.34 e

0.52 0.69 0.64 0.750.64 0.54 0.64 0.76

Mean 0.64 0.65 0.64 0.74

Bone Bruise (4/22) 0.42 0.77 0.42 0.13 0.0067 AvB, 0.012 AvD, 0.027 BvD,0.006 CvD, 0.0420.42 0.59 0.78 0.33

0.45 0.64 0.51 0.12Mean 0.40 0.70 0.6 0.2

Effusion (6/11) 0.38 0.29 0.48 0.82 <0.001 AvD, 0.009 BvC, BvD CvD;<0.0010.62 0.29 0.48 0.82

0.44 0.29 0.48 0.820.48 0.29 0.48 0.82

a Total no. of sites Z 44, 88, 22 and 11 for ligament injury, meniscal tear, bone bruise and fracture/osteochondral defect, respectively.b Weighted kappa. All other categories: unweighted kappa.c Prevalence of abnormal sites included in the OSE expected answer.

Table 5 Agreement between observers (CRs only; CR1/RRs; CR2/RRs; CR3/RRs). Lumbar spine (16 investigations)a.

Category Groups of observers ANOVAP value

Tukey multiplecomparison test,P value

CRs CR1 andRRs

CR2 andRRs

CR3 andRRs

A B C D

Disc (normal, bulge, protrusion,extrusion)b [7/80]c

0.45 0.56 0.57 0.58 0.17 e

0.64 0.64 0.48 0.790.31 0.71 0.31 0.59

Mean 0.47 0.64 0.45 0.65

Nerve root involvement [6/80]c 0.57 0.79 0.74 0.71 0.72 e

0.79 0.79 0.47 1.00.47 0.65 0.85 0.55

Mean 0.61 0.74 0.69 0.76

Degenerative disc disease [15/80]c 0.31 0.63 0.54 0.39 0.27 e

0.71 0.61 0.63 0.450.44 0.51 0.51 0.36

Mean 0.49 0.58 0.56 0.40

Tumour/metastases [9/80]c 0.56 0.48 0.9 0.65 0.02 AvB, 0.032 BvC,0.0030.79 0.48 0.9 0.65

0.74 0.38 0.74 0.31Mean 0.70 0.45 0.85 0.54

Other incidental findings [7/80]c 0.34 0.69 0.26 0.64 0.04 BvC, <0.001 CvD,<0.0010.86 0.69 0.26 0.64

0.42 0.59 0.31 0.71Mean 0.54 0.66 0.28 0.66

a Total no. of sites Z 80 for all categories.b Weighted kappa. All other categories: unweighted kappa.c Prevalence of abnormal sites included in the OSE expected answer.

MRI reporting by radiographers 141

142 K. Piper et al.

the categories of disc morphology, nerve root involvementand degenerative disc disease, where no significant differ-ences were found between the rates for the differentgroups of observers (p Z 0.17, p Z 0.72 and p Z 0.27,respectively). Some differences were apparent for theremaining categories: tumour/metastases (p Z 0.02) andother incidental findings (p Z 0.04). Tukey analysis for thetumour category revealed that agreement in groups A(CRs only) and C (CR2 and RRs) were significantly higher(p Z 0.032 and p Z 0.003) than in group B (CR1 and RRs).

No previous studies have been identified which havereported the agreement rate between radiologists wheninterpreting tumours on MRI investigations and/or thereporting of incidental findings. In this study, the meanKappa value of 0.70 for group A appears to be reasonablyconsistent, however, with the range of values found forother categories. It is interesting to note that the agree-ment between one of the radiologists (CR2) and the radi-ographers (group C) is higher (K Z 0.85) than the rate foundfor the radiologists’ only group.

Similarly the rate found for the other incidental findingswas significantly higher for two groups; group B (CR1 andRRs) and group D (CR3 and RRs) than either group A (CRsonly) or group C (CR2 and RRs). The Kappa rates outlined inTables 4 and 5 suggest that the agreement between radi-ologists and radiographers is not reduced when radiogra-phers form part of the group of observers, in fact if thereare significant differences it is because the group of radi-ographers, at times, have tended to agree with one of theradiologists’ reports to a greater extent than radiologistsagree as a group (CRs only).

Overall, the results presented here in Tables 2e5 suggestthat this group of radiographers, at the end of an accreditedpostgraduate programme, can report this range of selected MRIinvestigations (knee, lumbar spine and IAM) with satisfactoryaccuracy under examination conditions in an academic setting.There also appears to be little or no difference between themean accuracy and/or agreement rates when compared toa small group of consultant radiologists.

The overall aim of the study presented here, however,was to investigate the performance of a group of radiog-raphers at the end of an accredited postgraduate pro-gramme in clinical reporting of general MRI investigations(knee, lumbar spine and IAM). A large number of reports(n Z 1560) were included in the analysis and at that time 39radiographers had been successful in achieving the passmark (90% sensitivity and specificity; 85% agreement) forthe OSE. The majority of the radiographers included in thisstudy (74%; 29/39) are reporting MRI investigations inpractice,18 some are leading multidisciplinary team meet-ings,4 and over 60 radiographers have completed, or arecompleting the PgC programme. In the future, it will beimportant to investigate the accuracy of MRI reporting byradiographers more widely and particularly during/afterimplementation in clinical practice.

Acknowledgments

The authors are grateful to the radiologists and radiographerswho reported the images referred to in this study, withoutwhose assistance this study would not have been possible.

References

1. Audit Commission. Radiology. Audit Commission; 2002. Acutehospital portfolio.

2. Department of Health. Commissioning an 18 week patientpathway. Proposed principles and definitions: a discussiondocument. COI; 2005.

3. National Diagnostics Imaging Board. Delivery of the 18 weekpatient pathway and beyond: a strategy for imaging workforce,http://www.18weeks.nhs.uk; March 2007 [accessed17/11/2009].

4. National Diagnostics Imaging Board. Delivering the 18 weekpatient pathway. Radiology reporting times guidance, http://www.18weeks.nhs.uk/content.aspx?path=/achieve-and-sustain/Diagnostics/Imaging [accessed 17/11/2009].

5. Griffin K. Visiting mobile scanners: An alternative solution?Synergy; Nov 2005:10e2.

6. Radiology Service Improvement Team. Case studies: Hin-chingbrooke and Medway Maritime NHS trusts, http://www.radiologyimprovement.nhs.uk/View.aspx?page=/default.html[accessed 17/11/2009].

7. Canterbury Christ Church University. Programme validationdocument: postgraduate certificate in clinical reporting (MRI e

general investigations). Canterbury: CCCU; December 2002.8. Piper KJ, Buscall KL. MRI reporting by radiographers: the

construction of an objective structured examination. Radio-graphy 2008;14:78e89.

9. Robinson PJA, Wilson D, Coral A, Murphy A, Verow A. Variationbetween experienced observers in the interpretation of acci-dent and emergency radiographs. Br J Radiol 1999;72:323e30.

10. Piper KJ, Paterson AM, Godfrey RC. Accuracy of radiographers’reports in the interpretation of radiographic examinations ofthe skeletal system: a review of 6796 cases. Radiography 2005;11:27e34.

11. Snijders T. Multilevel analysis: an introduction to basic andadvanced multilevel modelling. California: Sage; 1999.

12. Campbell MJ. Statistics at square two. Understanding modernstatistical applications in medicine. London: BMJ Books; 2001.

13. McCullagh P. Generalized linear models. Chapman and Hall;1989.

14. Rorden C.ezANOVAv.97, www.mricro.com [accessedJuly 2007].15. Bryan S, Weatherburn G, Bungay H, Hatrick C, Salas C, Parry D,

et al. The cost-effectiveness of magnetic resonance imagingfor investigation of the knee joint. Health Technology Assess-ment 2001;5(27).

16. Schweitzer ME, Tran D, Deely DM, Hume EL. Medial collateralligament injuries: evaluation of multiple signs, prevalence andlocation of associated bone bruises, and assessment with MR.Imaging. Radiology 1995;194:825e9.

17. Altman DG. Practical statistics for medical research. London:Chapman and Hall; 1991.

18. Canterbury Christ Church University. MSc clinical reporting:programme quality monitoring and evaluation report; 2006.