PhD

dissertationThe testing and treating of patients with suspected or confirmed coronary artery disease were analyzed by modelling the complex outcomes associated with each initial treatment choice and measuring the effect of model parameter variance on the optimal strategy. The analysis included the diagnostic decisions of using a stress test and selective coronary angiography, and treatment choices of bypass graft surgery, medical therapy and the relatively new procedure, coronary angioplasty. Decision analysis was used to structure the problem. Patient specific probabilities were taken from most recent medical literature, including data for mildly symptomatic patients with minimal extent of disease, or expanded from other sources by Bayes probability revision. The expected outcome of each treatment was mathematically modelled by considering the potential outcomes including subsequent therapeutic options arising from the initial treatment choice. The benefits, relief of angina, and the costs, procedural discomfort and monetary cost, of each therapy were combined with the equations relating the probability of the patient being in each therapeutic state with time and numerically integrated over the remainder of the patient's life to estimate the utility of the initial treatment choice. Error analysis was used in addition to sensitivity analysis to test the validity of the results. Surgery was preferred to medical treatment for nearly all patients, and angioplasty provided the greatest utility for single vessel disease patients. Surgery was preferred to medical therapy even though the option of a subsequent surgery was provided in the medical therapy model. Preference was a direct consequence of the lower annual death rates after bypass surgery. Only when the annual mortality rate of the therapies differed by no more than one death in a thousand per year did consideration of quality of life and monetary cost of treatment become important. Because of increased benefits from invasive therapy, angiography was generally recommended, and little difference in value was observed from either performing or not performing a stress test