Abstract
Medical errors are one of the leading causes of death in America. In diagnostic radiology, an estimated 30% of errors are perceptual (Berlin, 2007), and thus preventable. The radiologists' task is made more difficult by increasingly common telephone interruptions. Correlational studies have shown that more interruptions correlate with more disagreements over diagnosis (Balint et al., 2014), but it is not clear what causes these effects. More generally, it is not clear what effects telephone interruptions have on diagnostic accuracy or duration of reading time on a case-by-case basis. Our study aimed to address this gap by asking advanced radiology fellows and attending physicians to read a heavy case-load of medical images in a reading room environment during off-hours while wearing mobile eye-tracking glasses. Each session included two phone calls, which interrupted the radiologists while they were reading complex volumetric cases with multiple views. A pre-corded message instructed the radiologist to find and examine a different patient's case. Telephone interruptions were counter-balanced across cases, allowing us to compare the cost of interruption on identical cases across radiologists. We found that cases that were interrupted were examined ~60s longer than uninterrupted cases. Surprisingly, each expert observer missed a retrospectively visible sternal fracture. Interestingly, the radiologists who were not interrupted during this case spent ~6x longer looking in the vicinity of the fracture than those who were interrupted. This includes one radiologist who, in the face of a distracting telephone interruption, neglected to examine any of the images where the sternum could be visualized. Together, this preliminary examination provides strong evidence that interruptions contribute substantially to the already challenging task that faces diagnostic radiologists. We hope that our eye-tracking data may provide means by which to mitigate some of these costs by enhancing the education of future radiology residents.
Meeting abstract presented at VSS 2016