MD@: a physician-friendly decision analysis tool.

Abstract

The cost-effective use of medical resources is increasingly important in justifying strategies for medical diagnosis and management. Although some software is available to help with decision analysis, it can be difficult to use these tools for medical applications. We have developed a prototype package for modeling various medical decision strategies, which can be used with a Macintosh or Windows-based personal computer. The system is graphically based, intuitive, and user-friendly. The user constructs decision trees for comparing alternative strategies for diagnosis and management. Selecting blocks from a library, the user sets mean values for variables such as prevalence, sensitivity, specificity, cost, morbidity, and mortality. The system then generates the probabilities of various pathways, using Bayesian analysis, without requiring the user to enter equations. It displays the best strategy in terms of a particular criterion and, when appropriate, performs sensitivity analysis.