Reading Mammograms? Focus on Early Morning and Mid-Afternoon

September 22, 2020

Time of day can affect how well a radiologist can identify and diagnose normal lesions.

If you are reading mammograms, pay attention to the time of day. If it is late morning or late afternoon, consider giving yourself extra time to make an accurate diagnosis – new research shows that your ability to correctly pinpoint normal lesions drops during those times.

The industry already knows a good deal about how radiologist experience levels, fatigue, and lesion conspicuity affect interpretation errors, but there has been little research conducted into how the time of day impacts diagnostic performance. A team of investigators from the University of Sydney in Australia took a specific look at the effect of the time on the clock and published their findings Sept. 21 in Clinical Radiology.

“The results indicate that, as shown in other domains, the time of day when mammographic image readings are made may influence radiologists’ performance, specifically their ability to identify normal images correctly,” said the team led by A.S. Alshabibi, faculty of medicine and health.

Existing literature clearly outlines that the time of day does impact cognitive function, including reading competence, as well as visual selective attention, visual working memory, and decision-making. So, it is possible, Alshabibi’s team contends, that if you are reading a mammogram right after an early-morning breakfast or right before your shift is over, your diagnostic accuracy will likely be different.

To find out just how much of a difference time makes, the team conducted a retrospective mammographic reading assessment with 197 radiologists who attended workshops held during the Royal Australian and New Zealand College of Radiologists annual conferences from 2013 to 2017. They had providers read mammograms in a specialized reading room with ambient lighting for two-hour blocks between 8 a.m. and 8 p.m.

Radiologists were asked to analyze 60 de-identified mammograms – 20 abnormal and 40 normal – from the BREAST initiative, a program based on digital screen-reading test sets that can assess radiologists’ ability to distinguish between malignant and cancer-free images. In this study, the team asked providers to categorize any identified lesions using a four-point confidence scale with 2 being benign and 5 being an unquestioned malignancy.

The study authors recognized and noted that this work does not mimic a typical 8-hour workday, so they ruled out fatigue as a possible contributor to any performance differences during the two-hour sessions.

Looking at performance across times of day, the team determined that reader specificity was best during the early morning (8 a.m. to 10 a.m.) and mid-afternoon (2 p.m. to 4 p.m.) – 82.74 percent and 81.39 percent, respectively. It was much lower during late morning (10 a.m. to 12 p.m.) and late-afternoon (4 p.m. to 6 p.m.) – 71.7 percent and 73.95 percent, respectively.

In addition, the team found there was no significant variation in performance based on provider age, number of years reading mammograms, or proportions of radiologists who were more experienced. However, there was significant variation for those who had completed a mammography fellowship (p=0.011), they said.

“Fellowship is a potential confounding variable,” they said, “so, it was used as a covariate in the primary analysis.”

The time of day did not significantly impact lesion sensitivity, but when considered in tandem with changes in specificity, the team postulated that decision thresholds change during the day. For example, specificity decreased in the late morning, but lesion sensitivity non-significantly increased, and both specificity and lesion sensitivity non-significantly decreased in the late afternoon.

These findings indicate you might be less willing to take a risk on a diagnosis as the day progresses, they said.

“With regard to the threshold effect in particular, decreasing the threshold for considering potential abnormalities as cancerous between 10 a.m. and 12 p.m. is potentially representative of radiologists relying upon a more risk-averse decision strategy,” they explained. “During this period radiologists would make a positive decision if they harbor any doubt about a particular image appearance being cancerous.”

Although they did not specifically look at the effects of fatigue, the team did acknowledge that the longer you stay awake the more negative impact sleep pressure can have. Further studies are needed to explore the full impact of time of day.

“Radiologists need to be aware that time-of-day variability in reporting, particularly because optimal performance is crucial to satisfactory patient outcomes regardless of the hour when an interpretation is made,” they concluded. “These findings present significant implications for radiological clinicians.”