How radiation exposure histories influence physician imaging decisions: A multicenter radiologist survey study

OBJECTIVE. The purpose of this article is to evaluate the influence of patient radiation exposure histories on radiologists' imaging decisions. MATERIALS AND METHODS. We conducted a physician survey study in three academic medical centers. Radiologists were asked to make an imaging recommendation for a hypothetical patient with a history of multiple CT scans. We queried radiologists' decision making, evaluating whether they incorporated cancer risks from previous imaging, reported acceptance (or rejection) of the linear no-threshold model, and understood linear no-threshold model implications in this setting. Consistency between radiologists' decisions and their linear no-threshold model beliefs was evaluated; those acting in accordance with the linear no-threshold model were expected to disregard previously incurred cancer risks. A Fisher exact test was used to verify the generalizability of results across institutions and training levels (residents, fellows, and attending physicians). RESULTS. Fifty-six percent (322/578) of radiologists completed the survey. Most (92% [295/322]) incorporated risks from the patient's exposure history during decision making. Most (61% [196/322]) also reported acceptance of the linear no-threshold model. Fewer (25% [79/322]) rejected the linear no-threshold model; 15% (47/322) could not judge. Among radiologists reporting linear no-threshold model acceptance or rejection, the minority (36% [98/275]) made decisions that were consistent with their linear no-threshold model beliefs. This finding was not statistically different across institutions (p = 0.070) or training levels (p = 0.183). Few radiologists (4% [13/322]) had an accurate understanding of linear no-threshold model implications. CONCLUSION. Most radiologists, when faced with patient exposure histories, make decisions that contradict their self-reported acceptance of the linear no-threshold model and the linear no-threshold model itself. These findings underscore a need for educational initiatives.