Quantifying uncertainty in medical decisions

AbstractEffective handling of uncertainty is one of the central problems in medical decision making. The sources and effects of uncertainty in medical decision making are examined and some new quantitative approaches for solving the associated problems are outlined. To handle uncertainty in the branching probabilities and node utilities for probability trees representing alternative treatment strategies, a public domain software package that can be used for the construction, analysis and comparison of probability trees with random parameters was developed. To facilitate specification of the random variables that arise in medical decision making problems, public domain software packages for both data-driven and subjective estimation of probability densities from the Johnson translation system of distributions have also been developed. For the analysis of complex problems that cannot be adequately represented by probability trees or by simple stochastic processes such as Markov chains, network simulation approaches that are oriented toward the sequence of activities seen by individual patients in the course of treatment are described