Objective: To determine whether mammographic interpretations are biased by the patient's clinical history.
Design: On 2 occasions, separated by a 5-month wash-out period, 10 radiologists read mammograms for the same 100 women, randomly divided into 2 groups of 50. For 1 group, the clinical history was supplied for the first reading and omitted (except for age) for the second reading. This sequence was reversed in the other group. In addition, 5 cases were shown a third time with a deliberately leading sham history.
Patients: Selected with stratified random sampling from 3 categories of diagnostic findings (64 had mammographic abnormalities) and from the definitive designation of breast cancer or no breast cancer (18 had breast cancer).
Main outcome measures: Radiologists' diagnostic accuracy and directional changes in interpretations and recommendations between the 2 readings.
Results: The direction suggested by the history led to small but consistent changes in the interpretations. Overall diagnostic accuracy was not altered, but recommendations were affected for appropriate further diagnostic workup: an alerting history (eg, breast symptoms or family history of breast cancer) increased the number of workups recommended in patients without cancer (P=.01); and a nonalerting history led to fewer recommended workups in the cancer patients (P=.02). The direction of the sham histories led an average of 4 of the 10 radiologists to change previous diagnoses and an average of 1 radiologist to change a previous biopsy recommendation.
Conclusions: Knowledge of the clinical history may alter a radiologist's level of diagnostic suspicion without improving performance in either diagnosis or management recommendations.