Prior studies have evaluated discordance rates among radiology residents in interpretation of head computed tomograms (CTs). To our knowledge, there has been no study to compare performance among first-year fellows and more experienced general staff radiologists. This study will compare performances of these groups and evaluate the effect of a redundant system as part of a quality control program. Retrospective review of 3,886 consecutive head CTs in the Emergency Department from 7/01/04 to 6/30/2005 was performed. Fellows interpreted 2,150 and general staff radiologists 1,736 cases. Staff radiologist mean experience was 4 years (2-10 years). All primary interpretations were over-read by staff neuroradiologists (>10 years experience) as quality control. Discrepancies were divided into "major discordance" and "minor discordance." Major discordance is defined as a misread occurred that potentially delayed clinical management and thus may have incurred in mortality or disability. Minor discordance is defined as if there was no change in clinical management or impact on the patient's outcome. The patient electronic medical records were obtained and retrospectively reviewed to identify if there was an acute change in clinical management. Overall discordance rate of both groups was 2.7% (103/3,886), 0.3% major false negative (10/3,886), 1.7% minor false negative (65/3,886), 0.4% false positive (15/3,886). Fellows overall discordance rate was 2.6% (55/2,150) with major false negatives 4/2,150 (0.2%) and 2.8% (48/1,736) for general staff radiologists with 6/1736 (0.3%) major false negatives, p values 0.69 and 0.14, respectively. Three out of ten major false negatives were confirmed with the quality assurance interpretation on follow-up studies; four cases were in agreement with initial interpretation. Performance among first-year fellows and general staff radiologists in interpretation of head CTs was highly accurate (97.3%) without statistically significant difference between the groups. The overall relatively low discrepant rate between fellowship trainees and generalist staffs, as well as the negligible change in clinical management, suggests little utility in over-reads of head CT scans by the neuroradiology service as part of a year-round quality control program. However, because of a relative high discrepant rate in the early months of fellowship training (>5%) in our study, it may be wise to implement a quality assurance program in the first few months to improve patient care. Increasing over-reading rate may reduce false negative rate, as the overall false positive rate is relatively low (<0.5%).