Uncovering bias in the cytologic evaluation of cervical squamous lesions.

Abstract

It is well recognized that biases exist in medical decision making. We sought evidence for such bias in diagnostic testing. We investigated whether a cytotechnologist's Papanicolaou (Pap) test interpretation of a squamous cell abnormality influenced the likelihood of making the same interpretation again that day using analysis based on the β distribution. For squamous intraepithelial lesion (SIL) interpretations, significant deviation away from the mean daily diagnostic rate was seen within all three cytotechnologists and, for atypical squamous cells of undetermined significance interpretations, within two of three. Cytotechnologists are not influenced by an expected number of abnormal Pap cases per day since this would result in deviation toward the mean daily rate of diagnosis. Possible explanations for the unanticipated clustering of SIL interpretations include clinical clustering effects or, alternatively, the influence a SIL interpretation might have on lowering the threshold for this interpretation again in subsequent specimens on the same day. The analysis presented here could serve as a model to detect bias in other aspects of medical decision making.