- Email: [email protected]
An observational study to evaluate the performance of units using two radiographers to read screening mammograms
Clinical Radiology 67 (2012) 114e121
Contents lists available at ScienceDirect
Clinical Radiology journal homepage: www.clinicalradiologyonline.net
Original Paper
An observational study to evaluate the performance of units using two radiographers to read screening mammograms R.L. Bennett a, *, S.J. Sellars b, R.G. Blanks a, S.M. Moss a a b
Cancer Screening Evaluation Unit, Institute of Cancer Research, Sutton, Surrey, UK NHS Cancer Screening Programmes, Sheffield, UK
article in formation Article history: Received 13 April 2011 Received in revised form 2 June 2011 Accepted 16 June 2011
AIM: To examine the performance of screening units in which a proportion of mammograms were double read using “non-discordant radiographer only (double) reading” (NDROR). MATERIALS AND METHODS: NDROR was used by seven pilot units between 2006 and 2009, and six further units in 2009 only. There were 51 comparison units. Screening performance outcome measures were calculated, and logistic regression was used to compare performance between the pilot and comparison units. RESULTS: Phase 1 pilot units read between on average 15 and 48% of mammograms per year using NDROR between 2006 and 2009 (median, 33%) and in 2009, phase 2 pilot units used NDROR to read between 4 and 77% of mammograms (median, 34%). The results showed an increase in recall rates in the phase 1 pilot units relative to the comparison units at both prevalent and incident screens (adjusted OR 1.09, 95% CI 1.05, 1.14; and adjusted OR 1.10, 95% CI 1.07, 1.14, respectively). There were also increases in the phase 2 pilot units relative to the comparison units; adjusted OR 1.08 (95% 1.00, 1.17) at prevalent screens, and adjusted OR 1.07 (95% CI 1.02, 1.14) at incident screens. There was no evidence to suggest a difference in cancer-detection rates between the pilot units and the comparison units. CONCLUSIONS: Evidence from the present study suggests that recall rates may increase as a result of units using radiographers to double read a proportion of their mammograms. However, there is little evidence to suggest that NDROR, as practiced by the pilot units in the present study, is likely to have major impacts on performance in the UK National Health Service Breast Screening Programme (NHSBSP), particularly if it is fully supported and closely monitored (particularly recall rates). Ó 2011 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
Introduction The initial recommendations of the Forrest Report, on which the UK National Health Service Breast Screening Programme (NHSBSP) was established, were that women * Guarantor and correspondent: R.L. Bennett, Cancer Screening Evaluation Unit, Institute of Cancer Research, 15 Cotswold Road, Sutton, Surrey SM2 5NG, UK. Tel.: þ44 (0)20 8725 5894; fax: þ44 (0)20 8725 3328. E-mail address: [email protected] (R.L. Bennett).
aged 50e64 years should be invited to screening at 3 yearly intervals and that mammograms (mediolateral oblique view only) should be single read by a radiologist or other medically qualified individual.1 Further evidence, published after the Forrest report, led to the introduction of two-view mammography, and additionally the expansion of the age range to include women up to age 70 years. Single reading remains programme policy although some units chose to double read mammograms from the beginning, and a survey conducted in 1995 showed that only 12% of units
0009-9260/$ e see front matter Ó 2011 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.crad.2011.06.015
R.L. Bennett et al. / Clinical Radiology 67 (2012) 114e121
routinely used single reading.2 A further study published in 1998 showed that cancer detection rates were higher in units using double reading than in those using single reading, although there was no difference in recall rates, and as a result many more units began to use double reading.3 However, staffing and other resources sometimes do not permit the double reading of mammograms. A general shortage of radiologists has resulted in the expansion of the radiographer role4 and many NHSBSP screening units have trained radiographers to read mammograms.5 They act as the second reader alongside the radiologist, thus ensuring that mammograms are double read. Reading by radiographers in addition to radiologist (double) reading has also been proposed,6 but pre-reading by radiographers was shown not to be cost-effective in the NHSBSP.7 The expansion of the radiographer role was supported further by the announcement of the four-tier skill mix model in the NHS Cancer Plan; in which roles were to be based on skills and experience rather than profession.8 In the NHSBSP the expansion of the radiographer role was also influenced by an increase in workload, which was associated with units in England having to have introduced two views at incident screens by 2003 and begun inviting women up to age 70 by the end of 2004. Additional involvement of radiographers in film reading occurred when, in an attempt to manage their workloads, a small number of units introduced double reading by two radiographers. In these units only discordant cases were read (at arbitration) by a medically qualified individual (radiologist/breast clinician). Whilst experimental evidence suggests that, after allowing for years of experience, radiographers read as well as radiologists,9 there was no reallife evidence to support this change in practice. The aim of the present study was to establish the impact on overall unit performance of a proportion of mammograms being double read by only two radiographers in a real-life setting.
Materials and methods Study design An observational study was initiated in 2004 and a set of criteria was prepared that pilot units were expected to follow. The study protocol specified that units in the study would read a proportion of their mammograms using “non-discordant radiographer only (double) reading” (NDROR): two radiographers would independently double read. Concordant cases were automatically recalled for assessment and discordant cases were arbitrated by an experienced radiologist (or equivalent), either separately or at a consensus meeting (including the original film readers). All film readers in the NHSBSP must complete a recognized course of study, read a minimum of 5000 mammograms per year and participate annually in PERFORMS.10 Additionally in this study one of the radiographer pair was expected to have at least 2 years screen-reading experience. All pilot units were required to register with the NHS Cancer Screening Programmes (NHSCSP) National Office and to gain Trust Board approval before commencing non-discordant radiographer only (double) reading.
115
Film reading practice and workforce questionnaire A questionnaire was developed and circulated to all UK screening units in January 2006 to document both their film-reading practices and the number of years of experience of individual film readers in each of the 5 years from 1 April 2000 to 31 March 2005. Pilot units were surveyed again for the 4 years from 1 April 2005 to 31 March 2009. The questionnaire asked whether radiographer film readers were used, and for the total number of film readers in each year. Information, by year, on each film reader was requested including their profession, if they read fewer than the recommended number of mammograms per year (5000) and years of experience (where years of experience were recorded as “more than n years’, n years was taken as the number of years of experience). Reading protocols were categorized as single reading; double reading: recall if one suggests; double reading: consensus; double reading: arbitration by a third reader; and double reading: arbitration by a third reader at a consensus meeting. Units were also asked to record whether the second reader was blinded to the findings of the first reader (if using double reading); the percentage of mammograms read by each reading protocol in each year; the readers involved; and, if using double reading with arbitration, the percentage of mammograms sent to arbitration.
Statistical analysis Three units had begun using radiographers to double read mammograms prior to the study commencing, and a further four units were initially recruited as pilot units. In total 13 pilot units were in the study group. This was split into two subgroups: phase 1 consisted of the seven units that commenced using non-discordant radiographer only reading before 1 April 2005 and the six units that began after 1 April 2006 formed the phase 2 subgroup. The performance of the pilot units was compared before (control period) and after (study period) they had introduced NDROR. For the analysis of the phase 1 pilot units the control period was the 4 year period between 1998 and 2001. (Although one unit began double reading with radiographers in 2001 comparable data were not available for the period 1997e2000. Data were therefore excluded for this unit for 2001.) The study period was the 4 year period between 1 April 2005 and 31 March 2009 during which all phase 1 pilot units had undertaken NDROR for 4 complete years. For the analysis of the phase 2 pilot units the control period was the single year 1 April 2005 to 31 March 2006 when none of the units had commenced NDROR, and the study period was 1 April 2008 to 31 March 2009. After joining the study in 2007, one pilot unit merged with another unit (not in the study) in April 2008. The newly formed unit adopted NDROR across its entire service. The data from this merged unit were analysed for the study period and from both units combined for the control period. A comparison group was used to adjust for temporal changes between the control and study periods not associated with the introduction of NDROR. This group was
116
R.L. Bennett et al. / Clinical Radiology 67 (2012) 114e121
formed of 51 units in England that did not register as pilot units, and had returned complete questionnaires for the 5 year period from 1 April 2000 to 31 March 2005. Logistic regression was used to compare the performance of both the pilot group and the comparison group before and after the introduction of radiographer only double reading, and to adjust the performance of the pilot units for time effects in the comparison units. All statistical tests were two-sided, and analyses were performed in Stata version 10.1.
Results Non-discordant radiographer only reading
Percentage of films double read by two radiographers
Fig 1 shows the proportion of mammograms read using NDROR in the phase 1 and phase 2 pilot units during the study periods. In the phase 1 pilot units the mean percentage per year of NDROR mammograms varied between 15 and 48% (median 33%). The percentage of mammograms read using NDROR in the phase 2 pilot units in 2009 ranged from 4 to 77%, (median 34%). The pilot unit in which 77% of mammograms were read using NDROR did not fully implement arbitration of all discordant films until November 2008, prior to this discordant cases were automatically recalled by the arbitrating radiologist. The 100 90 80 70 60 50
30 20 10 2006
2007
2008
2009
Year
Percentage of films double read by two radiographers
(a) 100 90 80 70 60 50 40 30 20 10 0
Screening performance outcome measures Phase 1 pilot units Table 1a shows screening performance measures at prevalent screens. There was an increase in rates of recall to assessment between the control period and the study period in both the pilot and comparison units. There was a 9% increase in recall rates in the pilot units between the control and study periods, after adjustment for that in the comparison units (adjusted OR 1.09, 95% CI 1.05, 1.14). Performance outcome measures for women being screened at incident screens are shown in Table 1b. There were significant decreases in rates of recall to assessment for women being screened at incident screens in both the pilot and comparison units (OR 0.91, 95% CI 0.88, 0.94; and OR 0.83, 95% CI 0.82, 0.84, respectively). However, in the pilot units this change is equated to a 10% increase in recall rates between the control and study periods relative to that in the comparison units (adjusted OR 1.10, 95% CI 1.07, 1.14). At incident screens the PPV in the study period was 20.92% in the comparison units and 20.48% in the phase 1 pilot units. There was a greater increase in PPV in the comparison units than in the pilot units (OR 1.66, 95% CI 1.62,1.71 versus OR 1.51, 95% CI 1.39,1.63), and this equated to a decrease in PPV in the pilot units relative to the comparison units (adjusted OR 0.91, 95% CI 0.84, 0.99). The observed increases in cancer-detection rates were similar in the comparison and pilot units.
Phase 2 pilot units
40
0
non-NDROR mammograms in the pilot units were mainly double read with arbitration by either a third reader or by a third reader at a consensus meeting.
2006
2007
2008
Results from the phase 2 pilot units at prevalent screens are shown in Table 2a. In the pilot units there was a 12% increase (OR 1.12, 95% CI 1.04, 1.20) in recall rates between the control and study periods and a 3% increase in the comparison units (OR 1.03, 95% CI 1.00, 1.06). This was equivalent to an 8% increase in recall rates in the pilot units between the control and study periods, relative to that in the comparison units (adjusted OR 1.08, 95% CI 1.00, 1.17). At incident screens there was a decrease in recall rates between the control and study periods in the comparison units (OR 0.91, 95% CI 0.89, 0.93; Table 2b). There was no significant difference in recall rates between the control and study periods in the pilot units (OR 0.98, 95% CI 0.93, 1.03); however, relative to the comparison units this equated to a 7% increase between the two periods after adjustment for that in the comparison units (adjusted OR 1.07, 95% CI 1.02, 1.14). There was no significant difference in PPV after adjustment, due to a non-significant increase in the cancerdetection rate.
2009
Year
(b) Figure 1 (a) Percentage of all mammograms double read by only two radiographers in the seven phase 1 pilot units during the study period. (b) Percentage of all mammograms double read by only two radiographers in the six phase 2 pilot units during the study period.
Film-reading practices and skill mix Table 3 shows film-reading protocols for the control and study periods in both the pilot and comparison units. In the comparison units, there was a decrease in the proportion of films read using single reading or double reading: recall if
Table 1a Screening performance measures in phase 1 pilot units: First invitation to routine screening (women aged 50e54 years). Phase 1 pilot units
Control period (scr ¼74,588) 7.58 women screened) 6.33 4.69 1.57 8.35
Phase 1 pilot units adjusted for time effects in comparison units
Study period (scr ¼ 66,955)
OR
95% CI
Control period (scr ¼ 570,881)
Study period (scr ¼ 515,143)
OR
95% CI
OR
95% CI
8.39
1.12
(1.07, 1.16)
8.32
8.48
1.02
(1.01, 1.03)
1.09
(1.05, 1.14)
7.41 5.12 2.29 8.83
1.17 1.09 1.46 1.06
(1.03, (0.94, (1.14, (0.93,
6.54 4.81 1.69 7.86
7.36 5.20 2.14 8.68
1.13 1.08 1.26 1.11
(1.08, (1.02, (1.16, (1.06,
1.04 1.01 1.15 0.95
(0.91, (0.86, (0.89, (0.83,
1.33) 1.27) 1.87) 1.22)
1.18) 1.14) 1.38) 1.17)
1.19) 1.18) 1.49) 1.10)
Table 1b Screening performance measures in phase 1 pilot units - Routine invitation to previous attenders (women aged 50-64 years). Phase 1 pilot units
Control period (scr ¼ 236319) Recalled to assessment (%) 3.57 Cancer detection rate (per 1000 women screened) Overall 5.21 Invasive 3.96 In-situ/non-invasive 1.21 Positive predictive 14.58 value of recall (%)
Comparison units
Study period (scr ¼ 301814)
OR
95% CI
Control period (scr ¼ 1852991)
3.26
0.91
(0.88,0.94)
3.89
6.67 5.12 1.53 20.48
1.28 1.29 1.26 1.51
(1.19,1.38) (1.19,1.41) (1.09,1.47) (1.39,1.63)
5.34 4.24 1.07 13.74
Phase 1 pilot units adjusted for time effects in comparison units Study period (scr ¼2225624)
OR
95% CI
OR
95% CI
3.23
0.83
(0.82,0.84)
1.10
(1.07,1.14)
6.77 5.37 1.38 20.92
1.27 1.27 1.29 1.66
(1.24,1.30) (1.23,1.31) (1.21,1.36) (1.62,1.71)
1.01 1.02 0.98 0.91
(0.94,1.09) (0.94,1.11) (0.84,1.15) (0.84,0.99)
R.L. Bennett et al. / Clinical Radiology 67 (2012) 114e121
Recalled to assessment (%) Cancer detection rate (per 1000 Overall Invasive In-situ/non-invasive Positive predictive value of recall (%)
Comparison units
117
118
Table 2a Screening performance measures in phase 2 pilot units - First invitation to routine screening (women aged 50-54 years). Phase 2 pilot units
Control period (scr ¼19657) 8.11 women screened) 7.07 5.29 1.78 8.72
Phase 2 pilot units adjusted for time effects in comparison units
Study period (scr ¼ 20148)
OR
95% CI
Control period (scr ¼ 124655)
Study period (scr ¼ 133895)
OR
95% CI
OR
95% CI
8.99
1.12
(1.04,1.20)
8.21
8.46
1.03
(1.00,1.06)
1.08
(1.00,1.17)
7.89 5.81 2.08 8.77
1.12 1.10 1.17 1.01
(0.88,1.41) (0.84,1.45) (0.73,1.89) (0.79,1.29)
7.12 4.84 2.23 8.68
7.23 5.22 1.99 8.54
1.01 1.08 0.89 0.98
(0.93,1.11) (0.97,1.21) (0.75,1.06) (0.89,1.08)
1.10 1.02 1.31 1.02
(0.86,1.41) (0.76,1.35) (0.81,2.12) (0.79,1.32)
Table 2b Screening performance measures in phase 2 pilot units - Routine invitation to previous attenders (women aged 50-64 years). Phase 2 pilot units
Control period (scr ¼ 88405) Recalled to assessment (%) 3.33 Cancer detection rate (per 1000 women screened) Overall 6.52 Invasive 5.18 In-situ/non-invasive 1.33 Positive predictive 19.55 value of recall (%)
Comparison units
Study period (scr ¼ 87634)
OR
95% CI
3.27
0.98
(0.93,1.03)
6.93 5.40 1.52 21.21
1.06 1.04 1.14 1.11
(0.95, (0.91, (0.88, (0.97,
1.19) 1.19) 1.47) 1.26)
Control period (scr ¼ 528902)
Phase 2 pilot units adjusted for time effects in comparison units Study period (scr ¼ 576641)
OR
95% CI
OR
95% CI
3.39
3.10
0.91
(0.89,0.93)
1.07
(1.02,1.14)
6.88 5.45 1.41 20.28
6.70 5.32 1.37 21.59
0.97 0.98 0.97 1.08
(0.93, (0.93, (0.88, (1.03,
1.09 1.07 1.17 1.02
(0.97, (0.93, (0.90, (0.89,
1.02) 1.03) 1.08) 1.14)
1.24) 1.23) 1.53) 1.17)
R.L. Bennett et al. / Clinical Radiology 67 (2012) 114e121
Recalled to assessment (%) Cancer detection rate (per 1000 Overall Invasive In-situ/non-invasive Positive predictive value of recall (%)
Comparison units
R.L. Bennett et al. / Clinical Radiology 67 (2012) 114e121 Table 3 Film reading practices. Main reading protocol
Single reading Double reading; recall if one suggests Double reading: consensus Double reading: arbitration of all cases Double reading: arbitration of discordant cases Triple reading Other
Comparison units
Phase 1 pilot units
Phase 2 pilot units
2001
2005
2001
2006
12% 18%
2% 12%
57% 28%
29%
119
of a “learning curve” during the 4 year study period with decreases in rates of recall to assessment, and in the percentage of films sent to arbitration. Furthermore, whilst experience improves film-reading skill, the radiographers in the pilot units had less film-reading experience than medically qualified film readers in the comparison units.
NDROR good practice 31%
20%
4%
4%
29%
51%
6%
6% 6%
14%
29%
43%
one suggests and an increase in units using double reading: arbitration of discordant cases between 2001 and 2005. In the units using double reading, 43% of the comparison units reported that the second reader was “blinded” to the results of the first reader. This figure was 29% in the phase 1 pilot units and 33% in the phase 2 pilot units. (All units were expected to have implemented direct entry of screening results by December 2006, and thus could undertake blind double reading more easily.) The proportion of comparison units with radiographer film readers increased during the period of the questionnaire, from 33% in 2001 to 94% in 2005. However, radiologists and breast clinicians still represented the majority of film readers (57%), and on average these readers were more experienced than radiographer film readers in the pilot units [9.7 years (range 1e21 years) compared to 3.5 years (range 1.7e6 years) in phase 1 pilot units].
Discussion Screening performance outcome measures Results from this study showed that relative to the comparison units there was an increase in recall rates at both prevalent and incident screens in units using NDROR to read a proportion of their mammograms, but this was not accompanied by an increase in rates of cancer detection. In the comparison units there was a 17% decrease in recall rates at the incident screens between the control and study periods. Between these two periods two-view mammography was introduced at incident screens, and this resulted in an 11% decrease in recall rates (from 4.2% to 3.9%)11 although in Wales, recall rates increased from 3.5% to 4.1% as a result of two-view mammography.12 Film-reading radiographers have been shown to have higher rates of recall to assessment than radiologists13 and the reading protocol of the study, i.e. double reading with arbitration by an experienced clinician was designed in part to minimize recall rates. In such a setting it is reasonable to expect rates to decrease and in the phase 1 pilot units there was evidence
Feedback from the pilot units highlighted a number of criteria that they considered to be of particular importance to the practice of NDROR. These included protected film-reading sessions; regular attendance by the radiographers at assessment clinics; and involvement in discussions on routine audits and in the radiological review of interval cancers. Also, the radiographers requested more formal training in handling the responsibility associated with the additional clinical decision-making of NDROR. Furthermore, units need to plan their workforce carefully so that NDROR can be practiced regularly, and to ensure that it continues even if there is an increase in the availability of medically qualified film readers.
Film readers in the NHSBSP As a result of this study the film reading in the NHSBSP may need to be reviewed. For many years radiographers have read alongside radiologists or breast clinicians, and have often enabled mammograms to be double read in situations in which otherwise they would not have been. However, consideration must be given both to the film-reading radiographer and to the practice of NDROR in units that have enough staff such that all mammograms can be/are double read by either two clinicians or by a clinician and radiographer. The professional development of all staff is important but units must balance this against their staffing needs so that film-reading skills are not wasted, film readers develop and maintain their reading skills in a screening environment, and all film readers meet the reading volume requirement of the NHSBSP.
Limitations of the study Although there was no significant difference in cancer-detection rates between the pilot units and the comparison units, the power of the study to detect such differences was limited, and it was not feasible to study rates of small invasive cancers. There was considerable variation in the proportion of mammograms read using two radiographers in each of the pilot units. The study was initially designed as a randomized, controlled trial in which pilot units would rotate their normal reading practices with NDROR on a monthly basis. However, such a design was deemed not to be feasible as workload pressures had heavily influenced the decision to introduce double reading by radiographers. Therefore, results from this study are only able to suggest what the possible effect on the performance of a unit would be in units reading a similar proportion of mammograms with
120
R.L. Bennett et al. / Clinical Radiology 67 (2012) 114e121
only two radiographers and not what the effect would be of wholly adopting this reading practice. The results for the merged unit were included after one pilot unit merged with another non-pilot unit during the study. It was assumed that the performance of this unit would not have been affected by the merger, and results showed a similar increase in recall rates even if these units were excluded. The questionnaire did not exactly reflect the exact time period for which the performance outcome data were analysed. For the analysis of the phase 1 pilot units, film-reading practices in 2001 were used to reflect the control period (1998e2001) and 2005 for the study period (2006e2009). For the analysis of the phase 2 pilot units, film-reading practices in 2005 were used to reflect the control period (2006). More recent data were not available for the comparison sites; however, results of the questionnaire showed that there was an increase in the proportion of units using “double reading: arbitration of discordant cases” between 2001 and 2005; and it was assumed this trend would have continued. This reading protocol was also used in conjunction with NDROR in the pilot units.
NDROR in future practice The primary aim of units adopting NDROR in the present study was to manage workloads rather than to reduce workforce costs. Any potential reduction in reading costs associated with NDROR would need to take into account the costs, both for the NHSBSP and the woman, of additional assessments resulting from increased recall rates. NDROR could be adopted to manage the additional workload that will arise from the NHSBSP taking on the surveillance of women at high risk of developing breast cancer and from the further expansion of the programme to the 47 to 73 year age group.14 It could be used particularly during times of holiday or sick leave when radiologists and breast clinicians may not be available, particularly as radiographers take no longer to read than radiologists.15 However, the practice needs support and still requires the input of, at present, an experienced clinician to arbitrate all discordant cases. As radiographers become more experienced in film reading, the need for the arbitrator to be medically qualified could be reviewed. The involvement of an experienced clinician as an arbitrator may be further increased if all recall cases were arbitrated in an effort to reduce higher recall rates. A shortage of radiologists to act as arbitrator could reduce the ability of a screening unit to meet NHSBSP standards,16 and also to reduce the availability of assessment clinics, thus limiting the number of women who can be recalled to assessment. Although the film-reading workload of radiologists could be reduced by the introduction of more radiographer film readers and NDROR, an increase in the availability of radiologists for assessment clinics could potentially increase recall rates further. In conclusion, the evidence from this study showed that recall rates increased in pilot units relative to the comparison units after the introduction of NDROR. However, if the
practice is fully supported, and the issues associated with NDROR are addressed, including those identified by the film readers who participated in this study, and also ensuring that recall rates are closely monitored, then NDROR as practiced in this study is unlikely to have major impacts on the performance of units in the NHSBSP.
Acknowledgements The authors are grateful to staff at all screening units for providing detailed information on reading protocols and readers’ experience, and particularly those in the following pilot units who have provided additional information during the course of this study: Barnsley Breast Screening Unit, Beds. and Herts. Breast Screening Centre, City, Sandwell and Walsall Breast Screening Service, Cornwall Breast Screening Service, Hereford and Worcester Breast Screening Service, Humberside Breast Screening Service, Kettering Breast Screening Service, North Staffordshire Breast Screening Service, Peterborough Breast Screening Unit, Sheffield Breast Screening Unit, South Essex Breast Screening Service, Warwickshire, Solihull and Coventry Breast Screening Service, West Sussex Breast Screening Service. The Cancer Screening Evaluation Unit receives funding from the Department of Health Policy Research Programme. The views expressed in this publication are those of the authors and not necessarily those of the Department of Health.
References 1. Forrest APM. Breast cancer screening: report to the health ministers of England, Wales, Scotland and Northern Ireland. London: HMSO; 1986. 2. Gerard K, Brown J, Johnston K. UK breast screening programme: how does it reflect the Forrest recommendations? J Med Screen 1997;4:10e5. 3. Blanks RG, Wallis MG, Moss SM. A comparison of cancer detection rates achieved by breast cancer screening programmes by number of readers, for one and two view mammography: results from the UK National Health Service breast screening programme. J Med Screening 1998; 5:195e201. 4. Price RC, Miller LR, Mellor F. Longitudinal changes in extended roles in radiography. Radiography 2002:223e34. 5. Pauli R, Hammond S, Cooke J, et al. Comparison of radiographer/radiologist double film reading with single reading in breast cancer screening. J Med Screening 1996;3:18e22. 6. Duijm LE, Groenewoud JH, Fracheboud J, et al. Introduction of additional double reading of mammograms by radiographers: effects on a biennial screening programme outcome. Eur J Cancer 2008;44:1223e8. 7. Haiart DC, Henderson J. A comparison of interpretation of screening mammograms by a radiographer, a doctor and a radiologist: results and implications. Br J Clin Pract 1991;45:43e5. 8. Department of Health. The NHS cancer plan. A plan for investment. A plan for reform. London: Department of Health; 2000. 9. Scott HJ, Gale AG, Woodman C. Breast screening technologists: does real-life case volume affect performance? In: Chakraborty DP, Eckstein MP, editors. Medical Imaging 2004: Image Perception, Observer Performance, and Technology Assessment. Bellingham, WA: SPIE Press; 2004. 10. NHS Breast Screening Programme. Consolidated guidance on standards for the NHS Breast Screening Programme. Sheffield: NHS Cancer Screening Programmes; 2005. 11. Blanks RG, Bennett RL, Patnick J, et al. The effect of changing from one to two views at incident (subsequent) screens in the NHS breast screening
R.L. Bennett et al. / Clinical Radiology 67 (2012) 114e121 programme in England: impact on cancer detection and recall rates. Clin Radiol 2005;60:674e80. 12. Osborn GD, Beer H, Wade R, et al. Two-view mammography at the incident round has improved the rate of screen-detected breast cancer in Wales. Clin Radiol 2006;61:478e82. 13. van den Biggelaar FJ, Nelemans PJ, Flobbe K. Performance of radiographers in mammogram interpretation: a systematic review. Breast 2008;17:85e90.
121
14. Department of Health. Cancer reform strategy. London: Department of Health; 2007. 15. Wivell G, Denton ER, Eve CB, et al. Can radiographers read screening mammograms? Clin Radiol 2003;58:63e7. 16. NHS Breast Screening Programme. Quality assurance guidelines for adminstrative and clerical staff. Sheffield: NHS Cancer Screening Programmes; 2000.
Contents lists available at ScienceDirect
Clinical Radiology journal homepage: www.clinicalradiologyonline.net
Original Paper
An observational study to evaluate the performance of units using two radiographers to read screening mammograms R.L. Bennett a, *, S.J. Sellars b, R.G. Blanks a, S.M. Moss a a b
Cancer Screening Evaluation Unit, Institute of Cancer Research, Sutton, Surrey, UK NHS Cancer Screening Programmes, Sheffield, UK
article in formation Article history: Received 13 April 2011 Received in revised form 2 June 2011 Accepted 16 June 2011
AIM: To examine the performance of screening units in which a proportion of mammograms were double read using “non-discordant radiographer only (double) reading” (NDROR). MATERIALS AND METHODS: NDROR was used by seven pilot units between 2006 and 2009, and six further units in 2009 only. There were 51 comparison units. Screening performance outcome measures were calculated, and logistic regression was used to compare performance between the pilot and comparison units. RESULTS: Phase 1 pilot units read between on average 15 and 48% of mammograms per year using NDROR between 2006 and 2009 (median, 33%) and in 2009, phase 2 pilot units used NDROR to read between 4 and 77% of mammograms (median, 34%). The results showed an increase in recall rates in the phase 1 pilot units relative to the comparison units at both prevalent and incident screens (adjusted OR 1.09, 95% CI 1.05, 1.14; and adjusted OR 1.10, 95% CI 1.07, 1.14, respectively). There were also increases in the phase 2 pilot units relative to the comparison units; adjusted OR 1.08 (95% 1.00, 1.17) at prevalent screens, and adjusted OR 1.07 (95% CI 1.02, 1.14) at incident screens. There was no evidence to suggest a difference in cancer-detection rates between the pilot units and the comparison units. CONCLUSIONS: Evidence from the present study suggests that recall rates may increase as a result of units using radiographers to double read a proportion of their mammograms. However, there is little evidence to suggest that NDROR, as practiced by the pilot units in the present study, is likely to have major impacts on performance in the UK National Health Service Breast Screening Programme (NHSBSP), particularly if it is fully supported and closely monitored (particularly recall rates). Ó 2011 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
Introduction The initial recommendations of the Forrest Report, on which the UK National Health Service Breast Screening Programme (NHSBSP) was established, were that women * Guarantor and correspondent: R.L. Bennett, Cancer Screening Evaluation Unit, Institute of Cancer Research, 15 Cotswold Road, Sutton, Surrey SM2 5NG, UK. Tel.: þ44 (0)20 8725 5894; fax: þ44 (0)20 8725 3328. E-mail address: [email protected] (R.L. Bennett).
aged 50e64 years should be invited to screening at 3 yearly intervals and that mammograms (mediolateral oblique view only) should be single read by a radiologist or other medically qualified individual.1 Further evidence, published after the Forrest report, led to the introduction of two-view mammography, and additionally the expansion of the age range to include women up to age 70 years. Single reading remains programme policy although some units chose to double read mammograms from the beginning, and a survey conducted in 1995 showed that only 12% of units
0009-9260/$ e see front matter Ó 2011 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.crad.2011.06.015
R.L. Bennett et al. / Clinical Radiology 67 (2012) 114e121
routinely used single reading.2 A further study published in 1998 showed that cancer detection rates were higher in units using double reading than in those using single reading, although there was no difference in recall rates, and as a result many more units began to use double reading.3 However, staffing and other resources sometimes do not permit the double reading of mammograms. A general shortage of radiologists has resulted in the expansion of the radiographer role4 and many NHSBSP screening units have trained radiographers to read mammograms.5 They act as the second reader alongside the radiologist, thus ensuring that mammograms are double read. Reading by radiographers in addition to radiologist (double) reading has also been proposed,6 but pre-reading by radiographers was shown not to be cost-effective in the NHSBSP.7 The expansion of the radiographer role was supported further by the announcement of the four-tier skill mix model in the NHS Cancer Plan; in which roles were to be based on skills and experience rather than profession.8 In the NHSBSP the expansion of the radiographer role was also influenced by an increase in workload, which was associated with units in England having to have introduced two views at incident screens by 2003 and begun inviting women up to age 70 by the end of 2004. Additional involvement of radiographers in film reading occurred when, in an attempt to manage their workloads, a small number of units introduced double reading by two radiographers. In these units only discordant cases were read (at arbitration) by a medically qualified individual (radiologist/breast clinician). Whilst experimental evidence suggests that, after allowing for years of experience, radiographers read as well as radiologists,9 there was no reallife evidence to support this change in practice. The aim of the present study was to establish the impact on overall unit performance of a proportion of mammograms being double read by only two radiographers in a real-life setting.
Materials and methods Study design An observational study was initiated in 2004 and a set of criteria was prepared that pilot units were expected to follow. The study protocol specified that units in the study would read a proportion of their mammograms using “non-discordant radiographer only (double) reading” (NDROR): two radiographers would independently double read. Concordant cases were automatically recalled for assessment and discordant cases were arbitrated by an experienced radiologist (or equivalent), either separately or at a consensus meeting (including the original film readers). All film readers in the NHSBSP must complete a recognized course of study, read a minimum of 5000 mammograms per year and participate annually in PERFORMS.10 Additionally in this study one of the radiographer pair was expected to have at least 2 years screen-reading experience. All pilot units were required to register with the NHS Cancer Screening Programmes (NHSCSP) National Office and to gain Trust Board approval before commencing non-discordant radiographer only (double) reading.
115
Film reading practice and workforce questionnaire A questionnaire was developed and circulated to all UK screening units in January 2006 to document both their film-reading practices and the number of years of experience of individual film readers in each of the 5 years from 1 April 2000 to 31 March 2005. Pilot units were surveyed again for the 4 years from 1 April 2005 to 31 March 2009. The questionnaire asked whether radiographer film readers were used, and for the total number of film readers in each year. Information, by year, on each film reader was requested including their profession, if they read fewer than the recommended number of mammograms per year (5000) and years of experience (where years of experience were recorded as “more than n years’, n years was taken as the number of years of experience). Reading protocols were categorized as single reading; double reading: recall if one suggests; double reading: consensus; double reading: arbitration by a third reader; and double reading: arbitration by a third reader at a consensus meeting. Units were also asked to record whether the second reader was blinded to the findings of the first reader (if using double reading); the percentage of mammograms read by each reading protocol in each year; the readers involved; and, if using double reading with arbitration, the percentage of mammograms sent to arbitration.
Statistical analysis Three units had begun using radiographers to double read mammograms prior to the study commencing, and a further four units were initially recruited as pilot units. In total 13 pilot units were in the study group. This was split into two subgroups: phase 1 consisted of the seven units that commenced using non-discordant radiographer only reading before 1 April 2005 and the six units that began after 1 April 2006 formed the phase 2 subgroup. The performance of the pilot units was compared before (control period) and after (study period) they had introduced NDROR. For the analysis of the phase 1 pilot units the control period was the 4 year period between 1998 and 2001. (Although one unit began double reading with radiographers in 2001 comparable data were not available for the period 1997e2000. Data were therefore excluded for this unit for 2001.) The study period was the 4 year period between 1 April 2005 and 31 March 2009 during which all phase 1 pilot units had undertaken NDROR for 4 complete years. For the analysis of the phase 2 pilot units the control period was the single year 1 April 2005 to 31 March 2006 when none of the units had commenced NDROR, and the study period was 1 April 2008 to 31 March 2009. After joining the study in 2007, one pilot unit merged with another unit (not in the study) in April 2008. The newly formed unit adopted NDROR across its entire service. The data from this merged unit were analysed for the study period and from both units combined for the control period. A comparison group was used to adjust for temporal changes between the control and study periods not associated with the introduction of NDROR. This group was
116
R.L. Bennett et al. / Clinical Radiology 67 (2012) 114e121
formed of 51 units in England that did not register as pilot units, and had returned complete questionnaires for the 5 year period from 1 April 2000 to 31 March 2005. Logistic regression was used to compare the performance of both the pilot group and the comparison group before and after the introduction of radiographer only double reading, and to adjust the performance of the pilot units for time effects in the comparison units. All statistical tests were two-sided, and analyses were performed in Stata version 10.1.
Results Non-discordant radiographer only reading
Percentage of films double read by two radiographers
Fig 1 shows the proportion of mammograms read using NDROR in the phase 1 and phase 2 pilot units during the study periods. In the phase 1 pilot units the mean percentage per year of NDROR mammograms varied between 15 and 48% (median 33%). The percentage of mammograms read using NDROR in the phase 2 pilot units in 2009 ranged from 4 to 77%, (median 34%). The pilot unit in which 77% of mammograms were read using NDROR did not fully implement arbitration of all discordant films until November 2008, prior to this discordant cases were automatically recalled by the arbitrating radiologist. The 100 90 80 70 60 50
30 20 10 2006
2007
2008
2009
Year
Percentage of films double read by two radiographers
(a) 100 90 80 70 60 50 40 30 20 10 0
Screening performance outcome measures Phase 1 pilot units Table 1a shows screening performance measures at prevalent screens. There was an increase in rates of recall to assessment between the control period and the study period in both the pilot and comparison units. There was a 9% increase in recall rates in the pilot units between the control and study periods, after adjustment for that in the comparison units (adjusted OR 1.09, 95% CI 1.05, 1.14). Performance outcome measures for women being screened at incident screens are shown in Table 1b. There were significant decreases in rates of recall to assessment for women being screened at incident screens in both the pilot and comparison units (OR 0.91, 95% CI 0.88, 0.94; and OR 0.83, 95% CI 0.82, 0.84, respectively). However, in the pilot units this change is equated to a 10% increase in recall rates between the control and study periods relative to that in the comparison units (adjusted OR 1.10, 95% CI 1.07, 1.14). At incident screens the PPV in the study period was 20.92% in the comparison units and 20.48% in the phase 1 pilot units. There was a greater increase in PPV in the comparison units than in the pilot units (OR 1.66, 95% CI 1.62,1.71 versus OR 1.51, 95% CI 1.39,1.63), and this equated to a decrease in PPV in the pilot units relative to the comparison units (adjusted OR 0.91, 95% CI 0.84, 0.99). The observed increases in cancer-detection rates were similar in the comparison and pilot units.
Phase 2 pilot units
40
0
non-NDROR mammograms in the pilot units were mainly double read with arbitration by either a third reader or by a third reader at a consensus meeting.
2006
2007
2008
Results from the phase 2 pilot units at prevalent screens are shown in Table 2a. In the pilot units there was a 12% increase (OR 1.12, 95% CI 1.04, 1.20) in recall rates between the control and study periods and a 3% increase in the comparison units (OR 1.03, 95% CI 1.00, 1.06). This was equivalent to an 8% increase in recall rates in the pilot units between the control and study periods, relative to that in the comparison units (adjusted OR 1.08, 95% CI 1.00, 1.17). At incident screens there was a decrease in recall rates between the control and study periods in the comparison units (OR 0.91, 95% CI 0.89, 0.93; Table 2b). There was no significant difference in recall rates between the control and study periods in the pilot units (OR 0.98, 95% CI 0.93, 1.03); however, relative to the comparison units this equated to a 7% increase between the two periods after adjustment for that in the comparison units (adjusted OR 1.07, 95% CI 1.02, 1.14). There was no significant difference in PPV after adjustment, due to a non-significant increase in the cancerdetection rate.
2009
Year
(b) Figure 1 (a) Percentage of all mammograms double read by only two radiographers in the seven phase 1 pilot units during the study period. (b) Percentage of all mammograms double read by only two radiographers in the six phase 2 pilot units during the study period.
Film-reading practices and skill mix Table 3 shows film-reading protocols for the control and study periods in both the pilot and comparison units. In the comparison units, there was a decrease in the proportion of films read using single reading or double reading: recall if
Table 1a Screening performance measures in phase 1 pilot units: First invitation to routine screening (women aged 50e54 years). Phase 1 pilot units
Control period (scr ¼74,588) 7.58 women screened) 6.33 4.69 1.57 8.35
Phase 1 pilot units adjusted for time effects in comparison units
Study period (scr ¼ 66,955)
OR
95% CI
Control period (scr ¼ 570,881)
Study period (scr ¼ 515,143)
OR
95% CI
OR
95% CI
8.39
1.12
(1.07, 1.16)
8.32
8.48
1.02
(1.01, 1.03)
1.09
(1.05, 1.14)
7.41 5.12 2.29 8.83
1.17 1.09 1.46 1.06
(1.03, (0.94, (1.14, (0.93,
6.54 4.81 1.69 7.86
7.36 5.20 2.14 8.68
1.13 1.08 1.26 1.11
(1.08, (1.02, (1.16, (1.06,
1.04 1.01 1.15 0.95
(0.91, (0.86, (0.89, (0.83,
1.33) 1.27) 1.87) 1.22)
1.18) 1.14) 1.38) 1.17)
1.19) 1.18) 1.49) 1.10)
Table 1b Screening performance measures in phase 1 pilot units - Routine invitation to previous attenders (women aged 50-64 years). Phase 1 pilot units
Control period (scr ¼ 236319) Recalled to assessment (%) 3.57 Cancer detection rate (per 1000 women screened) Overall 5.21 Invasive 3.96 In-situ/non-invasive 1.21 Positive predictive 14.58 value of recall (%)
Comparison units
Study period (scr ¼ 301814)
OR
95% CI
Control period (scr ¼ 1852991)
3.26
0.91
(0.88,0.94)
3.89
6.67 5.12 1.53 20.48
1.28 1.29 1.26 1.51
(1.19,1.38) (1.19,1.41) (1.09,1.47) (1.39,1.63)
5.34 4.24 1.07 13.74
Phase 1 pilot units adjusted for time effects in comparison units Study period (scr ¼2225624)
OR
95% CI
OR
95% CI
3.23
0.83
(0.82,0.84)
1.10
(1.07,1.14)
6.77 5.37 1.38 20.92
1.27 1.27 1.29 1.66
(1.24,1.30) (1.23,1.31) (1.21,1.36) (1.62,1.71)
1.01 1.02 0.98 0.91
(0.94,1.09) (0.94,1.11) (0.84,1.15) (0.84,0.99)
R.L. Bennett et al. / Clinical Radiology 67 (2012) 114e121
Recalled to assessment (%) Cancer detection rate (per 1000 Overall Invasive In-situ/non-invasive Positive predictive value of recall (%)
Comparison units
117
118
Table 2a Screening performance measures in phase 2 pilot units - First invitation to routine screening (women aged 50-54 years). Phase 2 pilot units
Control period (scr ¼19657) 8.11 women screened) 7.07 5.29 1.78 8.72
Phase 2 pilot units adjusted for time effects in comparison units
Study period (scr ¼ 20148)
OR
95% CI
Control period (scr ¼ 124655)
Study period (scr ¼ 133895)
OR
95% CI
OR
95% CI
8.99
1.12
(1.04,1.20)
8.21
8.46
1.03
(1.00,1.06)
1.08
(1.00,1.17)
7.89 5.81 2.08 8.77
1.12 1.10 1.17 1.01
(0.88,1.41) (0.84,1.45) (0.73,1.89) (0.79,1.29)
7.12 4.84 2.23 8.68
7.23 5.22 1.99 8.54
1.01 1.08 0.89 0.98
(0.93,1.11) (0.97,1.21) (0.75,1.06) (0.89,1.08)
1.10 1.02 1.31 1.02
(0.86,1.41) (0.76,1.35) (0.81,2.12) (0.79,1.32)
Table 2b Screening performance measures in phase 2 pilot units - Routine invitation to previous attenders (women aged 50-64 years). Phase 2 pilot units
Control period (scr ¼ 88405) Recalled to assessment (%) 3.33 Cancer detection rate (per 1000 women screened) Overall 6.52 Invasive 5.18 In-situ/non-invasive 1.33 Positive predictive 19.55 value of recall (%)
Comparison units
Study period (scr ¼ 87634)
OR
95% CI
3.27
0.98
(0.93,1.03)
6.93 5.40 1.52 21.21
1.06 1.04 1.14 1.11
(0.95, (0.91, (0.88, (0.97,
1.19) 1.19) 1.47) 1.26)
Control period (scr ¼ 528902)
Phase 2 pilot units adjusted for time effects in comparison units Study period (scr ¼ 576641)
OR
95% CI
OR
95% CI
3.39
3.10
0.91
(0.89,0.93)
1.07
(1.02,1.14)
6.88 5.45 1.41 20.28
6.70 5.32 1.37 21.59
0.97 0.98 0.97 1.08
(0.93, (0.93, (0.88, (1.03,
1.09 1.07 1.17 1.02
(0.97, (0.93, (0.90, (0.89,
1.02) 1.03) 1.08) 1.14)
1.24) 1.23) 1.53) 1.17)
R.L. Bennett et al. / Clinical Radiology 67 (2012) 114e121
Recalled to assessment (%) Cancer detection rate (per 1000 Overall Invasive In-situ/non-invasive Positive predictive value of recall (%)
Comparison units
R.L. Bennett et al. / Clinical Radiology 67 (2012) 114e121 Table 3 Film reading practices. Main reading protocol
Single reading Double reading; recall if one suggests Double reading: consensus Double reading: arbitration of all cases Double reading: arbitration of discordant cases Triple reading Other
Comparison units
Phase 1 pilot units
Phase 2 pilot units
2001
2005
2001
2006
12% 18%
2% 12%
57% 28%
29%
119
of a “learning curve” during the 4 year study period with decreases in rates of recall to assessment, and in the percentage of films sent to arbitration. Furthermore, whilst experience improves film-reading skill, the radiographers in the pilot units had less film-reading experience than medically qualified film readers in the comparison units.
NDROR good practice 31%
20%
4%
4%
29%
51%
6%
6% 6%
14%
29%
43%
one suggests and an increase in units using double reading: arbitration of discordant cases between 2001 and 2005. In the units using double reading, 43% of the comparison units reported that the second reader was “blinded” to the results of the first reader. This figure was 29% in the phase 1 pilot units and 33% in the phase 2 pilot units. (All units were expected to have implemented direct entry of screening results by December 2006, and thus could undertake blind double reading more easily.) The proportion of comparison units with radiographer film readers increased during the period of the questionnaire, from 33% in 2001 to 94% in 2005. However, radiologists and breast clinicians still represented the majority of film readers (57%), and on average these readers were more experienced than radiographer film readers in the pilot units [9.7 years (range 1e21 years) compared to 3.5 years (range 1.7e6 years) in phase 1 pilot units].
Discussion Screening performance outcome measures Results from this study showed that relative to the comparison units there was an increase in recall rates at both prevalent and incident screens in units using NDROR to read a proportion of their mammograms, but this was not accompanied by an increase in rates of cancer detection. In the comparison units there was a 17% decrease in recall rates at the incident screens between the control and study periods. Between these two periods two-view mammography was introduced at incident screens, and this resulted in an 11% decrease in recall rates (from 4.2% to 3.9%)11 although in Wales, recall rates increased from 3.5% to 4.1% as a result of two-view mammography.12 Film-reading radiographers have been shown to have higher rates of recall to assessment than radiologists13 and the reading protocol of the study, i.e. double reading with arbitration by an experienced clinician was designed in part to minimize recall rates. In such a setting it is reasonable to expect rates to decrease and in the phase 1 pilot units there was evidence
Feedback from the pilot units highlighted a number of criteria that they considered to be of particular importance to the practice of NDROR. These included protected film-reading sessions; regular attendance by the radiographers at assessment clinics; and involvement in discussions on routine audits and in the radiological review of interval cancers. Also, the radiographers requested more formal training in handling the responsibility associated with the additional clinical decision-making of NDROR. Furthermore, units need to plan their workforce carefully so that NDROR can be practiced regularly, and to ensure that it continues even if there is an increase in the availability of medically qualified film readers.
Film readers in the NHSBSP As a result of this study the film reading in the NHSBSP may need to be reviewed. For many years radiographers have read alongside radiologists or breast clinicians, and have often enabled mammograms to be double read in situations in which otherwise they would not have been. However, consideration must be given both to the film-reading radiographer and to the practice of NDROR in units that have enough staff such that all mammograms can be/are double read by either two clinicians or by a clinician and radiographer. The professional development of all staff is important but units must balance this against their staffing needs so that film-reading skills are not wasted, film readers develop and maintain their reading skills in a screening environment, and all film readers meet the reading volume requirement of the NHSBSP.
Limitations of the study Although there was no significant difference in cancer-detection rates between the pilot units and the comparison units, the power of the study to detect such differences was limited, and it was not feasible to study rates of small invasive cancers. There was considerable variation in the proportion of mammograms read using two radiographers in each of the pilot units. The study was initially designed as a randomized, controlled trial in which pilot units would rotate their normal reading practices with NDROR on a monthly basis. However, such a design was deemed not to be feasible as workload pressures had heavily influenced the decision to introduce double reading by radiographers. Therefore, results from this study are only able to suggest what the possible effect on the performance of a unit would be in units reading a similar proportion of mammograms with
120
R.L. Bennett et al. / Clinical Radiology 67 (2012) 114e121
only two radiographers and not what the effect would be of wholly adopting this reading practice. The results for the merged unit were included after one pilot unit merged with another non-pilot unit during the study. It was assumed that the performance of this unit would not have been affected by the merger, and results showed a similar increase in recall rates even if these units were excluded. The questionnaire did not exactly reflect the exact time period for which the performance outcome data were analysed. For the analysis of the phase 1 pilot units, film-reading practices in 2001 were used to reflect the control period (1998e2001) and 2005 for the study period (2006e2009). For the analysis of the phase 2 pilot units, film-reading practices in 2005 were used to reflect the control period (2006). More recent data were not available for the comparison sites; however, results of the questionnaire showed that there was an increase in the proportion of units using “double reading: arbitration of discordant cases” between 2001 and 2005; and it was assumed this trend would have continued. This reading protocol was also used in conjunction with NDROR in the pilot units.
NDROR in future practice The primary aim of units adopting NDROR in the present study was to manage workloads rather than to reduce workforce costs. Any potential reduction in reading costs associated with NDROR would need to take into account the costs, both for the NHSBSP and the woman, of additional assessments resulting from increased recall rates. NDROR could be adopted to manage the additional workload that will arise from the NHSBSP taking on the surveillance of women at high risk of developing breast cancer and from the further expansion of the programme to the 47 to 73 year age group.14 It could be used particularly during times of holiday or sick leave when radiologists and breast clinicians may not be available, particularly as radiographers take no longer to read than radiologists.15 However, the practice needs support and still requires the input of, at present, an experienced clinician to arbitrate all discordant cases. As radiographers become more experienced in film reading, the need for the arbitrator to be medically qualified could be reviewed. The involvement of an experienced clinician as an arbitrator may be further increased if all recall cases were arbitrated in an effort to reduce higher recall rates. A shortage of radiologists to act as arbitrator could reduce the ability of a screening unit to meet NHSBSP standards,16 and also to reduce the availability of assessment clinics, thus limiting the number of women who can be recalled to assessment. Although the film-reading workload of radiologists could be reduced by the introduction of more radiographer film readers and NDROR, an increase in the availability of radiologists for assessment clinics could potentially increase recall rates further. In conclusion, the evidence from this study showed that recall rates increased in pilot units relative to the comparison units after the introduction of NDROR. However, if the
practice is fully supported, and the issues associated with NDROR are addressed, including those identified by the film readers who participated in this study, and also ensuring that recall rates are closely monitored, then NDROR as practiced in this study is unlikely to have major impacts on the performance of units in the NHSBSP.
Acknowledgements The authors are grateful to staff at all screening units for providing detailed information on reading protocols and readers’ experience, and particularly those in the following pilot units who have provided additional information during the course of this study: Barnsley Breast Screening Unit, Beds. and Herts. Breast Screening Centre, City, Sandwell and Walsall Breast Screening Service, Cornwall Breast Screening Service, Hereford and Worcester Breast Screening Service, Humberside Breast Screening Service, Kettering Breast Screening Service, North Staffordshire Breast Screening Service, Peterborough Breast Screening Unit, Sheffield Breast Screening Unit, South Essex Breast Screening Service, Warwickshire, Solihull and Coventry Breast Screening Service, West Sussex Breast Screening Service. The Cancer Screening Evaluation Unit receives funding from the Department of Health Policy Research Programme. The views expressed in this publication are those of the authors and not necessarily those of the Department of Health.
References 1. Forrest APM. Breast cancer screening: report to the health ministers of England, Wales, Scotland and Northern Ireland. London: HMSO; 1986. 2. Gerard K, Brown J, Johnston K. UK breast screening programme: how does it reflect the Forrest recommendations? J Med Screen 1997;4:10e5. 3. Blanks RG, Wallis MG, Moss SM. A comparison of cancer detection rates achieved by breast cancer screening programmes by number of readers, for one and two view mammography: results from the UK National Health Service breast screening programme. J Med Screening 1998; 5:195e201. 4. Price RC, Miller LR, Mellor F. Longitudinal changes in extended roles in radiography. Radiography 2002:223e34. 5. Pauli R, Hammond S, Cooke J, et al. Comparison of radiographer/radiologist double film reading with single reading in breast cancer screening. J Med Screening 1996;3:18e22. 6. Duijm LE, Groenewoud JH, Fracheboud J, et al. Introduction of additional double reading of mammograms by radiographers: effects on a biennial screening programme outcome. Eur J Cancer 2008;44:1223e8. 7. Haiart DC, Henderson J. A comparison of interpretation of screening mammograms by a radiographer, a doctor and a radiologist: results and implications. Br J Clin Pract 1991;45:43e5. 8. Department of Health. The NHS cancer plan. A plan for investment. A plan for reform. London: Department of Health; 2000. 9. Scott HJ, Gale AG, Woodman C. Breast screening technologists: does real-life case volume affect performance? In: Chakraborty DP, Eckstein MP, editors. Medical Imaging 2004: Image Perception, Observer Performance, and Technology Assessment. Bellingham, WA: SPIE Press; 2004. 10. NHS Breast Screening Programme. Consolidated guidance on standards for the NHS Breast Screening Programme. Sheffield: NHS Cancer Screening Programmes; 2005. 11. Blanks RG, Bennett RL, Patnick J, et al. The effect of changing from one to two views at incident (subsequent) screens in the NHS breast screening
R.L. Bennett et al. / Clinical Radiology 67 (2012) 114e121 programme in England: impact on cancer detection and recall rates. Clin Radiol 2005;60:674e80. 12. Osborn GD, Beer H, Wade R, et al. Two-view mammography at the incident round has improved the rate of screen-detected breast cancer in Wales. Clin Radiol 2006;61:478e82. 13. van den Biggelaar FJ, Nelemans PJ, Flobbe K. Performance of radiographers in mammogram interpretation: a systematic review. Breast 2008;17:85e90.
121
14. Department of Health. Cancer reform strategy. London: Department of Health; 2007. 15. Wivell G, Denton ER, Eve CB, et al. Can radiographers read screening mammograms? Clin Radiol 2003;58:63e7. 16. NHS Breast Screening Programme. Quality assurance guidelines for adminstrative and clerical staff. Sheffield: NHS Cancer Screening Programmes; 2000.