Decision Tree

A decision tree is an analytical model used to represent a sequence of possible outcomes for a patient or patient cohort, depicted through distinct branches. Each branch originates from a series of nodes, where decisions or events occur. These nodes are categorized into ‘choice’ nodes, representing decisions between alternative interventions, and ‘probability’ nodes, representing events that occur by chance. The probabilities at any given node must sum to 1.

Costs and outcomes are assigned to each segment of a branch, including the terminal ends or ‘leaves’ of the tree. By calculating the outcomes and costs for each branch and their associated probabilities, the decision tree can be ‘rolled back’ to the initial decision node. This rollback process involves combining the outcomes and costs across all branches to determine the expected outcomes and costs for each treatment option, enabling a comparison of the alternatives.

Decision trees are particularly useful for modeling interventions with distinct, measurable outcomes at specific time points. They are less suited for evaluations where the timing of outcomes significantly impacts the results, as decision trees do not typically account for the timing of events as precisely as other models like Markov models. This makes them ideal for straightforward scenarios where outcomes are clear and discrete, simplifying complex decision-making processes in healthcare.