In a new study conducted by Mehmet U.S. Ayvaci from the University of Texas at Dallas (UT Dallas) studying whether risk profiling would influence the interpretation of mammograms, researchers found that considering a woman’s health risk profile while reviewing those exams would not only reduce the number of missed cancer incidents, but also diminish the number of false positives.
The paper was recently presented at the Advances in Decision Analysis, a conference sponsored by the Institute for Operations Research and the Management Sciences (INFORMS) Decision Analysis Society (DAS), held at Georgetown University in Washington, DC, earlier this week.
Risk profile information may include a variety of factors, such as family history, reproductive history, age and ethnicity.
The research was carefully conducted in order to provide a definitive answer to whether radiologists could provide a biased diagnosis, when in possession of risk profile information, and, there being bias, if this would help them produce more accurate readings of the exams.
Researchers used linear opinion polling, a decision science technique in which weights are assigned in order to better aggregate probability estimates to explore profile information and potential bias in mammography interpretation.
The decision performance of three groups was analyzed. First, a mammogram-only reading, in which no risk profile information about the patient was provided, second an unbiased reading, with radiologists consulting the risk profile after examining the mammogram, and a third biased or “influenced” reading, where radiologists consider a woman’s risk profile while they are examining the mammogram. Researchers then examined when and how providing profile information could come as an improvement in biopsy decisions.
Until now, no conclusive clinical evidence has been produced on interpreting mammograms using profile information. Although some scientists seem to believe that profile information does help radiologists make better decisions, and should therefore be used when reading mammograms, others hold a contradictory opinion, arguing that profile information may bias the radiologist’s decision.
Numerical analysis using a clinical dataset provided by the Breast Cancer Surveillance Consortium revealed that, compared to cases where profile information was not examined, careful use of this kind of information could reduce not only the false positives but also the number of missed cancers.