Prognostic stratification of patients with known or suspected cardiovascular disease influences both the effectiveness and efficiency of their management. Physicians currently perform such analyses informally in a variety of settings when making triage decisions. This project will test the hypothesis that the accuracy of these analyses can be enhanced by providing information from predictive algorithms derived through multivariate analysis, leading to improvements in diagnostic performance and efficiency of resource utilization. Algorithms for the prediction of myocardial infarction and complications requiring intensive care will be derived through analysis of clinical and electrocardiographic data from 1382 previously-seen emergency room patients, including 259 with myocardial infarctions, and from an additional group of 985 patients with acute myocardial infarction seen in the Multicenter Investigation of the Limitation of Infarct Size (MILIS). Advantages and disadvantages of competing multivariate techniques will be explored. Algorithms will be tested prospectively and compared with the performance of physicians on patients from both teaching and community hospitals, with subgroup analysis examining the performance of the algorithms in different settings and the impact of baseline data on diagnostic accuracy. The impact of results of cardiac isoenzyme assays in the emergency room on the probability of myocardial infarction will be examined through techniques of multivariate analysis. Cost-effectiveness analysis will be used to determine probability ranges within which specific management courses are recommended. In an intervention phase, algorithm predictions and recommendations from cost-effectiveness analysis will be provided to physicians during the emergency ward evaluation. The impact of these interventions on diagnostic accuracy and resource utilization will be measured.
