Nearly one million American cancer patients per year are treated with radiation therapy. Recently attention has been drawn to the unacceptably high error rates in this discipline, which stem partly from a recent explosion in the complexity of technology and care delivery. Although there is a general consensus in the field that patient safety needs to be improved, the principles of quality improvement are not well understood and practical tools are lacking. Our recent risk analysis research has identified two key quality control measures: inspection of the computer treatment plan prior to delivery and a measurement of the radiation beam during treatment. These two steps have the potential to make a major impact on safety improvement but are currently not well standardized and rely heavily on human inspection. Here we propose to automate these checks with a two-tiered system that employs 1) an intelligent database model to identify plans to that do not match patterns of previous treatments, and 2) a technology to automatically measure radiation doses during treatment. Our preliminary data shows that these two checks are complementary and will provide a particularly effective 'defense in depth'. We will measure the combined effectiveness of this system, and implement a pilot distribution plan with a partner clinic. By reducing radiotherapy errors in a measurable way, we expect to have a major impact on the care of cancer patients.
We propose to develop a distributable toolkit that will automatically detect and prevent errors in the radiation oncology. This toolkit will address a major unmet need and, based on our preliminary studies, will have a major impact for the nearly one million cancer patients treated each year with radiation therapy.
Kalet, Alan M; Gennari, John H; Ford, Eric C et al. (2015) Bayesian network models for error detection in radiotherapy plans. Phys Med Biol 60:2735-49 |