Recalls and false-positives for breast imaging patients were significantly more likely when the results were read by less-experienced radiologists who had worked more hours that day, based on data from more than 97,000 screening mammograms.
Psychology literature has shown the impact of fatigue on performance in a range of settings, and previous studies have shown that radiologists’ performances are more accurate earlier in their shifts compared to later-shift performance, write Michael H. Bernstein, PhD, and colleagues at Brown University, Providence, Rhode Island, in a study published online January 11 in Radiology.
The effect of time of day on performance may be greater for more detailed imaging modalities that are more “cognitively taxing,” and the effect may be greater in less-experienced radiologists, but the impact of time and experience on overall patient recall and false-positive rates has not been well-studied, the researchers said.
In the retrospective review the researchers identified 97,671 screening mammograms read by 18 radiologists at one of 12 community sites between January 2018 and December 2019. The researchers analyzed the results by type of image, either standard digital mammography (DM) or the more complex digital breast tomosynthesis (DBT). The researchers separated radiologists into two groups: those with at least 5 post-training years of experience and those with less than 5 post-training years of experience. A total of nine radiologists fell into each category.
Overall, the recall rates were significantly different, and higher for DM vs DBT (10.2% vs 9.0%; P = .006). The false-positive (FP) rate also differed significantly, and was higher for DM vs DBT (9.8% vs 8.6%; P = .004).
The odds of recall increased by 11.5% with each hour of reading time for radiologists with less than 5 post-training years of experience for both DBT (odds ratio [OR], 1.12) and DM (OR, 1.09). For the more experienced radiologists, the odds of recall increased by 1.6% for each hour of reading time for DBT, but decreased by 0.1% for DM, with no significant difference.
Similarly, the odds of an FP result increased by 12.1% for DBT and 9% for DM per hour of reading time for radiologists with less experience. For more experienced radiologists, the odds of an FP increased by 1.6% for DBT but decreased by 1.1% for DM per hour of reading time.
Cancer detection (defined as true-positive, or TP) was not higher for DM across time, the researchers note. However, “DBT achieved a higher TP rate than DM regardless of the time of day; this shows that for DBT to maintain a constant and superior TP rate relative to DM, radiologists’ FP rates had to go up as the day went on,” they write. “That is, although DBT achieves a superior TP rate, more junior radiologists appeared to compensate for their fatigue later in the day when using DBT by recalling a broader range of mammograms, more of which were FP findings.”
The researchers caution that their findings were limited by several factors, including the study’s retrospective design and the lack of randomization of the imaging technology, patients, and time of day, which prohibit conclusions regarding causality. Other limitations included the consideration of time of day without the ability to use hours since the start of a clinical shift, and the use of a 5-year mark to indicate experience without accounting for work volume.
However, the stronger impact of a time-of-day effect for more junior radiologists agrees with findings from other studies, the researchers add. More empirical research is needed, and the researchers propose a longitudinal study of how time of day affects radiologists as they gain experience, as well as experimental studies to test strategies for mitigating the time-of-day effect observed in the current study.
Scheduled Breaks May Reduce Impact of Fatigue
“Digital breast tomosynthesis is increasingly used in clinical practice and takes significantly longer to interpret compared with digital mammography,” said corresponding author Ana P. Lourenco, MD, in an interview with Medscape Medical News. “Radiologists interpret hundreds of images for each screening digital breast tomosynthesis exam, compared with four images for each screening digital mammogram exam; this may certainly contribute to radiologist fatigue.
”I found it interesting that there was a difference based on years of experience of the radiologist, but I was not surprised that recall rate increased later in the day, as some of us had anecdotally noted this in our clinical practice,” Lourenco said. In fact, the idea to conduct the study was prompted by a conversation with her statistician colleagues “about how I subjectively felt like my own recall rate increased at the end of the day.”
Ways to counteract the impact of fatigue could include intermittent breaks to refocus attention, said Lourenco. “Potential barriers would include imaging volumes and attending to patients in the breast imaging center,” she said. “If we can show that decreasing fatigue improves mammography performance metrics, then this may encourage practices to support such interventions.”
However, “More research that includes a larger number of radiologists, wider range of imaging interpretation experience; perhaps even experimental studies comparing metrics for radiologists reading with scheduled breaks vs without such breaks would be of interest,” Lourenco said.
Fatigue in Healthcare Goes Beyond Radiology
“Due primarily to staffing shortages and increased volume and complexity of patients, burnout and fatigue of all medical personnel, not just physicians, have become hallmarks of modern healthcare delivery in the United States, and this has been exacerbated by COVID-19 and other societal factors,” said Jeffrey C. Weinreb, MD, professor of radiology and biomedical imaging at Yale University, New Haven, Connecticut, in an interview with Medscape.
Previous studies have documented the fact that radiologists are among the specialists most affected by burnout and fatigue, and it has an impact on their performance, Weinreb said. The current study is important because it tries to pinpoint the key variables that are responsible for fatigue, so resources can be directed to effect change, he said.
Weinreb said he was not particularly surprised by the study findings. “Diagnostic mammography is a high-volume repetitive enterprise, so it would have been surprising if radiologist experience and time of day had no effect on performance and recall rate,” he said. “As most radiologists will attest based on personal experience, human beings get tired and lose some level of cognition over the course of a long, intense workday,” he added.
“I am a bit surprised that less experienced radiologists were more likely to recommend additional imaging at a higher rate when interpreting DBT, but not for DM and only later in the day,” Weinreb noted. “The authors suggest that this could be due to the increased number of images that are viewed with DBT and the different ways experienced and less experienced radiologist process the information. However, there could be other explanations, such as differences in volumes or differences in ages.
“Reducing the study volumes per radiologist is one obvious solution to reducing fatigue, but it will not be practical in many practices,” said Weinreb. “The important work of interpreting diagnostic mammograms needs to continue and grow. Without an increase in radiologist mammographers in the labor pool, this is not going to happen any time soon.”
Instead, “more immediate obvious solutions to radiologist fatigue in clinical practice include more frequent breaks during the workday, which would include walking around and not looking at a computer or cell phone screen, fewer images per study, report templates, streamlined workflow, more variety in daily work, and AI assistance for interpretation and reporting,” said Weinreb. Using nonradiologists when possible to relieve some of the burden could be considered, “but this is a complex and politically charged issue,” he noted.
Radiology is a well-compensated specialty, but further increasing compensation would help to mitigate burnout, said Weinreb. However, “Perhaps even more important is making certain that the efforts of individual radiologists are appreciated and recognized,” he said.
As for additional research needs, “Mammographers are not the only radiologists experiencing fatigue, but the most critical contributing factors for other types of imaging exams and subspecialities may not be identical,” Weinreb emphasized. “Data for other radiologists, similar to that provided by this study for diagnostic mammography, could be useful.
“An additional area of research could address the issue of individual radiologist circadian rhythms,” said Weinreb. “Perhaps we could rigorously determine whom amongst us is a ‘morning person’ vs one who performs equally as well or better later in the day, and use this information for radiologist scheduling,” he said. “Finally, once we know the key factors affecting performance for each type of exam and subspecialty, studies of possible incremental and combined benefits of various interventions would be needed.”
The study received no outside funding. The researchers and Weinreb have disclosed no relevant financial relationships.
Radiology. Published online January 11, 2022. Full text