When it comes to providing overnight image-interpretation services in academic medical centers, should our specialty shelve its traditional model—contemporaneous preliminary reads performed by radiology residents with later faculty oversight—and replace it with 24-hour coverage by attending radiologists? Some say yes. I say no. 

Those in favor of making the switch contend that attending radiologists provide a higher level of after-hours care than residents while simultaneously minimizing turnaround times and promoting goodwill with emergency physicians. This argument seems logical on its face, but it’s actually unsupported by the evidence as published in peer-reviewed journals over the past several years. The research shows that overnight attending coverage is, in fact, associated with a lower standard of after-hours care. The evidence further fails to demonstrate any care-quality or patient-safety benefits from moving to overnight attending coverage that would justify the high cost involved. And, perhaps worst of all, the change could do real harm to our profession’s educational mission. 

I know how counter-intuitive these evidence-based conclusions are, so let’s go through each one in a bit more detail. 

First, the extremely high accuracy of radiology trainees in the interpretation of overnight studies, especially in the emergency setting, has been copiously documented over the years. As the table below shows, study after study has found that residents generally perform at least as well as—and commonly outperform—attending radiologists on accuracy of overnight reads. 

Since the time of Leo Henry Garland in 1949, and in subsequent research conducted by many others, the discrepancy rates for attending radiologists for all types of studies have never been measured below three to four percent. In more than a few studies, attendings’ discrepant readings came in as high as 30 percent. By contrast, residents’ discrepancy rates for significant missed overnight findings are consistently very low. For example, in 2007, Cooper et al. retrospectively reviewed 141,381 exams and found residents to have significant discrepancies with the final attending reading in only one percent of the cases. 

Here’s a chart briefly summarizing some of the published research that turned up similar findings on radiology residents’ accuracy. 

[[{"fid":"24036","view_mode":"media_original","type":"media","attributes":{"height":466,"width":1172,"style":"width: 500px; height: 199px;","alt":" - ER coverage","class":"media-element file-media-original"}}]]

We can only speculate as to why this might be the case. Experimental work in other fields has revealed that expertise is domain-specific and does not translate to other, even similar, fields. Expert musicians, for example, are not necessarily effective public speakers, nor are master plumbers adequate electricians. And radiologists who can readily identify abnormalities on x-ray images in half a second or less are, on average, no better than the average person at finding Waldo in Where’s Waldo puzzles.

Further buttressing the cause of holding steady with the traditional model is its uncompromising approach to quality. Consider: The most common type of interpretive error in radiology is perceptual error, in which findings deemed obvious by a second reader are missed by the first. Double-reads are of course inherent to the traditional model, as residents’ overnight reads are routinely overread by attendings the next morning. Only then is a final report issued. This double-reading process—which traditionally brings in attendings with subspecialty expertise as needed for the second reads—is often lost with the move to an overnight attending model. 

Turning to the matter of return on investment, it’s important to recognize that no additional revenue stands to be generated by having attending radiologists rather than radiology residents preliminarily read overnight exams. But the expenses associated with providing a workable “nocturnalist” service are quite high, beginning with the costs to recruit, hire and retain—with generous compensation and paid time-off packages—enough “overnight-willing” radiology generalists to make a go of it. The academic radiology business model, wherein academic centers distinguish themselves by providing subspecialist service, is also partly undermined in the overnight attending model by creating a generalist-level standard after-hours.

The small handful of academic institutions that are able to pull this off successfully, such as Yale Health, Partners Healthcare and Michigan Medicine, are staffed with fully qualified academic subspecialist ED faculty providing both day and night coverage. These faculty members are given ample time for scholarly work and teaching. Few academic institutions are in a position to duplicate their success, which is financially subsidized by their parent institutions. 

In stark contrast to the outstanding and well-developed ER radiology sections at Michigan and Yale (and also at MGH), the majority of other academic medical centers across the nation provide overnight attending coverage in a much less satisfactory way, partly by excusing the overnight attendings from scholarly work and teaching. In a good many of these departments, there is no real pathway to academic career advancement for the overnight faculty, who are typically viewed as “hired guns.” These radiologists are highly compensated financially and given ample paid leave but are not really made part of the academic faculty community. 

My final argument against the 24-hour attending model has to do with its negative impact on the training and education of the next generation of radiologists. Resident autonomy is a longstanding feature of our training model, having served our specialty well for generations.  Clearly, moving to an overnight attending model can only hurt the education component of academic medicine’s multi-part mission. Admittedly, there is little published data to support this viewpoint. However, Collins et al., writing in the Journal of the American College of Radiology in 2014, published a survey of six academic medical centers showing that the majority of faculty and residents felt the loss of resident autonomy had strongly negative effects on resident training and on their subsequent confidence on entering the field as independent practitioners. 

To conclude, I would like to offer a few succinct words of advice for my fellow academic radiologists and, especially, for department chairpersons who are contemplating switching from the traditional academic model of overnight resident coverage to a nocturnalist overnight attending coverage model. To quote the late first lady Nancy Reagan: Just say no!  

Michael A. Bruno, MD, is a professor of radiology & medicine at Penn State College of Medicine and chief of emergency radiology at the Milton S. Hershey Medical Center.