Quantifying the Impact of Variability and Noise on Patient Outcomes in Breast Cancer Decision Making

There are many factors that can affect the breast cancer decision-making process. This article addresses issues of uncertainty. Specifically, we seek to answer two questions: (1) What are the major contributors to false positive test results for patients? (2) How does variability between different radiologists affect outcomes for patients? We develop a simulation-based model that combines statistical process control (SPC) with a partially observable Markov decision process (POMDP) to incorporate uncertainty, the inherent variability between radiologists, and system noise (i.e., screening characteristics such as different densities of breast tissue, inherent variability between different women, and imperfections in the mammogram and technology) to determine the impact on the breast cancer monitoring decision. When compared to population-based noise, we find that the variability among different radiologists in the ability to correctly interpret a mammogram has the most significant impact on whether a woman will receive incorrect results. Variability within the population of radiologists significantly increases in the number of false-positive mammogram results a woman receives. This suggests that reducing the variability between radiologists should be a primary concern to improve health care for women.