Discordant radiograph interpretation between emergency physicians and radiologists in a pediatric emergency department

Objectives: To describe the types of discrepancies in radiograph interpretation between emergency physicians and radiologists in a pediatric emergency department, and to determine the impact of discrepant interpretations on patient care. Methods: Prospective cohort study of discordant radiographs from the period beginning March 1, 1995 and ending March 31, 1.996. During this period, 2083 radiographs were coded by the radiologist as concordant or discordant. Three hundred forty-nine were coded as discordant, and 324 were eligible for the study. Charts were reviewed for relevant physical examination findings and emergency department management. Discrepancies that affected patient care were deemed clinically significant. Results: Twenty-three (1.1%) of 2083 radiographs were interpreted differently by the emergency physician and the radiologist in a way that might have changed patient management. This represents 7% (23/324) of the radiographs originally coded by a radiologist as discrepant, The most common discrepancy was a patient with a normal chest examination and a radiograph interpreted as having an infiltrate by the emergency physician, but subsequently read as having no infiltrate by a radiologist (12/324), These patients may have received antibiotics unnecessarily. Two discrepant interpretations had the potential to have serious consequences to the patient if not identified. One patient with cardiomegaly and another patient with free air on abdominal radiograph were not noted by the emergency physician. Conclusions: Emergency physicians would benefit from more rigorous interpretation of chest x-rays to avoid unnecessary treatment with antibiotics. Emergency physicians do a good job interpreting plain radiographs, but occasionally miss significant findings that could lead to adverse outcomes. The presence of radiologists to immediately read radiographs 24 hours a day could prevent missed findings, but, given the small number of significant misinterpretations, is unlikely to be cost effective.