Topic

Decision trees

A decision tree sets out the choices and chance events that follow a decision, with each pathway carrying a probability, a cost and an outcome. Folding back the tree from its branches to the decision node gives the expected value of each option. Trees suit short-term problems such as acute care.

Concepts in this topic

• Chance NodeA chance node is a point in a decision tree where an uncertain event, not a choice, decides the pathway, and its branch probabilities sum to one.
• Decision NodeA decision node is a point in a decision tree, drawn as a square, where the decision maker chooses between options and no probabilities are attached.
• Decision TreeA decision tree is a branching model used in economic evaluation: strategies, chance events and end points give each option's expected costs and QALYs.
• Decision Tree ModelA decision-analytic model built on the decision tree structure, with all possible patient pathways explicitly represented as branches to calculate expected cost.
• Expected ValueExpected value is the probability-weighted mean of an uncertain outcome, such as a cost or QALY total, across all its possible values.
• Expected Value AnalysisExpected value analysis compares health care options under uncertainty by their probability-weighted average outcomes, such as expected costs and QALYs.
• Folding BackFolding back is the backward evaluation of a decision tree that takes probability-weighted values at chance nodes and selects the preferred feasible action at decision nodes under a stated objective.