The Clinically Relevant Breast Imaging Audit

Abstract

An audit of a breast imaging practice must be based on data with accepted definitions and rules so that the comparisons between breast imaging facilities and interpretive staff are comparable. The four basic data points for calculating these metrics are true positive (TP), true negative (TN), false positive (FP), and false negative (FN). For mammography, the definition of "true" is the presence of a proven malignancy within a year of the exam. The presence or absence of breast cancer within a year of the exam and an increase in patient mobility between different facilities may render the calculation of sensitivity and specificity difficult for most facilities unless a regional cancer registry is available.Thus, the metrics that can be easily calculated within a facility are recall rate (all the positive interpretations divided by all the exams read), positive predictive value (PPV) 1 = percentage of abnormal screening exams that result in a diagnosis of cancer within a year, PPV2 = percentage of all diagnostic exams recommended for biopsy and cancer discovered within a year, PPV3 = benign tissue diagnosis and no cancer within a year, and the cancer detection rate (the true positive exams per one thousand exams). Intuitively, one may assume that accuracy (TP + TN/TP + FP + TN + FN) is the best metric for an interpreter. However, this can produce spurious results. The most accurate method to determine a reader's skills is the use of the receiver operating characteristic (ROC) curve, which clearly presents, in graphic form, the relationship between the four basic data points.

Keywords: audit; mammography; statistics.