Federally-mandated quality control (QC) in Papanicolaou (Pap) smear testing requires rescreening of 10% of negative smears, to include cases selected randomly as well as smears from patients that may have a higher risk for developing cervical cancer based on clinical information. FDA approval of NeoPath's AutoPap 300 QC system (NeoPath, Inc., Redmond, WA) allows practical QC rescreening of all negatives. We tested the ability of AutoPap to help increase identification of detection errors compared to random 10%/high-risk selection. From March 1-August 30, 1997, we utilized AutoPap/high-risk status to select cases for manual rescreen, and compared the rate of identification of primary screening errors to that for the preceding year using 10% random selection/high-risk status. Of 35,027 smears accessioned, 31,240 (89.1%) were screened as negative and 7,965 were selected for manual rescreen. Of these, 353 were determined to be abnormal. Most abnormals identified by this protocol were classified as atypical squamous or glandular cells of undetermined significance (ASCUS or AGUS). However, 59 low-grade squamous intraepithelial lesions (LSIL) and 13 high-grade squamous intraepithelial lesions (HSIL), many with few abnormal cells, were also identified. These results represented an increase in pickup rate of false negative due to detection errors of 2.3-, 2.8- and 5.6-fold for atypical squamous or glandular cells of undetermined significance, LSIL, and HSIL, respectively, when accounting for the volume differences over the time period measured. Our findings strongly support the conclusions drawn from clinical trials of the AutoPap that false negatives due to detection error can be significantly reduced when using AutoPap as part of a routine quality control program.