This presentation is part of: Poster Session - Public Health; Methodological Advances
PROBABILISTIC MODELING OF MEDICAL ERRORS IN RADIOTHERAPY
Robert C. Lee1, Edidiong Ekaette1, Karie-Lynn Kelly2, Chris Newcomb3, Peter Craighead2, and Peter Dunscombe3. (1) University of Calgary, Faculty of Medicine, Calgary, AB, Canada, (2) Tom Baker Cancer Centre, Radiation Oncology Department, Calgary, AB, Canada, (3) Tom Baker Cancer Centre, Medical Physics Department, Calgary, AB, Canada
Purpose: The process of treating cancer with ionizing radiation (radiotherapy) is complex and subject to medical errors; potentially resulting in morbidity and mortality to multiple patients, and litigation or criminal charges in some cases. Existing mechanisms to prevent such errors tend to be based on caregiver experience as opposed to using a systematic approach to identify and manage risks. Our objective is to develop and implement a quantitative risk and decision analysis model to elucidate risks and inform evidence based risk management and resource allocation decisions in radiotherapy. Methods: A team of oncologists, medical physicists, and risk and decision analysts first qualitatively mapped the radiotherapy system, and defined categories of events. Based on this, a probabilistic risk and decision analysis (PRADA) model was developed. This model employs linked influence diagram and Bayesian network calculations in a user friendly environment. The model allows estimation of risks and reduction of those risks using different quality assurance/quality control (QA/QC) patient safety interventions. Multiattribute utility functions are used as outcomes, and risk aversion is explicitly addressed. The value of different means of informing cancer staging and subsequent radiotherapy decisions is estimated using Bayesian methods. Model variables have been defined using a combination of literature values and expert judgement. Results: The qualitative mapping process identified four major categories of activities: Assessment, Preparation, Treatment, and Follow-up, and hierarchical levels of physician, physicist, dosimetrist, and technician activities. Influence diagrams have been defined for Assessment and Follow-up (as these activities involve decisions), and Bayesian networks for Preparation and Treatment (as these activities are simple sequences of events). Preliminary results indicate that the potential for catastrophic errors is high in the Assessment and Preparation stages, and relatively low in Treatment due to existing QA/QC procedures at that stage. Risk management options that increase resources to the early stages appear to have a positive net benefit. Conclusions: We provide a quantitative multiattribute model describing the potential sources of medical errors in radiotherapy and their consequences. The model suggests alternative strategies for risk management that may not be routinely implemented in cancer centres. This model represents a novel application of risk and decision analysis methods in the healthcare field. The methods should be generalizable to many forms of technologically intensive forms of healthcare such as surgery.
