Quantitative Framework

Decision Tree


1. Description

A decision tree is a horizontal tree diagram, with decisions as roots, uncertain events with their outcomes, and further decisions and outcomes as branches, ending with consequences. The expected utility rule is applied repeatedly and added at each node ("averaging out"), and then "folded back" for the highest utility decision.

2. Evaluation

2.1 Principle
  • Decision trees are easily-visualised graphical representations of the expected utility rule.
  • It is possible with decision tree software to show the expected resources and rewards at each branch of the tree, with the final net change displayed as the overall consequences at the end of each path through the tree.
  • Computer algorithm is used in "averaging out" and "folding back" the decision tree.
  • The relevant expected utility results are shown at each node for transparency
  • This is a fully transparent representation of the inputs and the results.
  • It shows which initial decision should be most favoured, and it indicates what subsequent decisions are the best policies depending on the outcomes of the future uncertain events.

2.2 Features
  • Decision trees can accommodate very complex problems involving many options, numerous uncertain events and multiple objectives.
  • Deterministic sensitivity analyse are performed to identify uncertain quantities that most affect the final expected utilities.

2.3 Visualisation
AA decision tree uses the tree diagram to represent the decision problems. Other visualisations that are associated with the application of decision tree are:
  • Tornado diagram
  • Sensitivity graph (area line graph)
  • 2.4 Assessability and accessibility
  • Familiarity with decision theory and decision analytic software is required for complex problems.
  • Developing the structure of the problem can be challenging because most people are not accustomed to thinking in terms of clearly-defined decision, events and their outcomes, and the possible consequences.
  • It is theoretically possible to be applied to drugs in development, at the approval stage, and post-approval.

3. References

[1] Hunink M, Glasziou P, Siegel J, Weeks J, Pliskin J, Elstein A, et al. Decision making in health and medicine: Integrating evidence and values. Cambridge: Cambridge University Press; 2001.
[2] Raiffa H. Decision analysis: Introductory lectures on choices under uncertainty. McGraw Hill; 1968.
[3] Spiegelhalter D, Abrams K, Myles JP. Bayesian Approaches to Clinical Trials and Health-Care Evaluation. Chicester, West Sussex: John Wiley & Sons Ltd.; 2004.