The consequences of medical decisions are inherently uncertain at the decisive moment. Using clinical examples related to the diagnosis and management of low back pain, the authors review some principles that can help physicians deal with this uncertainty. This article addresses the following: the use of probability as a useful representation of uncertainty, the use of Bayes' theorem to update probability estimates when new information is obtained, the measurement of a diagnostic test's accuracy, the use of the threshold model for choosing a diagnostic test, the principles of expected-value decision making, the use of utility assessment as a way of attaching value to outcomes, and the use of quality-adjusted life years as a measure of value. These principles can aid physicians in approaching complex and uncertain decisions with their patients. As the use of computers becomes more integrated into the process of care, the opportunity exists to move formal decision models from the policy level to the patient care level.