| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
The Royal Wolverhampton Hospitals NHS Trust Wolverhampton WV10 0QP, United Kingdom
Donnelly and Strife [1] are to be commended on the work in progress that they report on assessing radiologists' performance. The peer-review database is a potent instrument whereby all radiologists in a department can learn from mistakes and discrepancies made by themselves and their colleagues. It also can serve as a prompt whereby concern can be raised about the performance of an individual radiologist in respect to a particular type of activity.
At first sight, it would seem entirely reasonable that anyone appearing more than 2 SDs above the normalized mean should be notified and reported to the radiologist in chief. Yet in a British hospital where 25 radiology discrepancy meetings are held per year, only 81 cases involving error were reviewed in a 12-month period during which 105,000 examinations were performed [2]. The number of cases entered into the Cincinnati Children's Hospital peer review database and the individual workload figures of its radiologists are not disclosed [1]. The number of potential errors discussed at radiology discrepancy meetings typically is such that there is not an adequate sample size to derive error rates of individual radiologists. Studies from the development of the U.K. General Medical Council Performance Procedures found that "generalizable results would be obtained with not less than 3 judges all rating the same 60 reports from a radiologist" and "any assessment of performance of technical abilities in this field will need to use multiple assessors basing judgements on an adequate sample of reports" [3]. Radiology peer review databases and discrepancy meetings are subject to many pitfalls: hindsight bias, information bias, presentation or format bias, outcome bias, selection bias, faulty data collection, confounders, and, very important, effect of sample size. The difference between common cause and special cause variation needs to be recognized [4].
When comparing individual radiologist performance with reported performance in the literature, bear in mind that the quality of reporting in diagnostic accuracy articles in radiology journals is "less than optimal" [5]. Radiologic accuracy is often much less than we would expect. Two experienced university neuroradiologists, when evaluating lumbar disk herniation on MRI of patients for whom they had adequate clinical information, reported that full agreement is only 77% [6].
With the exception of breast imaging, Soffa et al. [7] are quite right that "radiologists do not currently have an objective benchmark for an acceptable level of missed diagnoses." Their department has shown that a blinded second interpretation quality assurance system can work in daily practice [7]. It does not have the problems of selection bias and recruitment of the ACR RADPEER system, which has been described as a "less than perfect measuring system"[8].
Setting a valid radiologic standard will take time. A radiologist whose performance is below average may still be performing adequately. A reporting discrepancy does not imply error [9]. The clinical information given to a reporting radiologist may be inadequate, incomplete, or incorrect. Moreover, radiology reports are only part of the diagnostic jigsaw. Even if a radiology report has the potential to adversely affect patient management, it often does not. The genesis of radiologic error and discrepancy is frequently systemic [10]. Although monitoring systems such as the peer review database are important, they must be used skillfully and sensitively. There needs to be equal emphasis on better teamwork, more knowledge sharing, and the value of targeted instruction of those radiologists whose performance is perceived to be less than adequate [11].
 Top References |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
