1 Graphic representation: the tree provides a model of the decision problem, with the advantages of providing an explicit basis for sharing knowledge and understanding of what is going on.
-
However, the trees can quickly become very big
and apparently complicated; large sub-trees may be duplicated many times.
2 Incorporates estimates of risk: this is the key characteristic of the method, which is an advantage if there are reasonable estimates of risk available (as in
a well-researched clinical problem) but arguably a disadvantage if there are not.
3 Incorporates utilities: the separation of risk and utility is one of the strong points of the method from a theoretical point of view.
-this has practical implications in clinical decision making when, in principle at least, the health care professional can supply the tree and the probabilities and the patient can supply the utilities.
-However, different methods of eliciting utilities can give different results and there may be questions about the validity and reliability of
the values used. Also there are questions about what to do when there are many stakeholders with different utilities.
4 Sensitivity analysis: the fact that decision trees lend themselves readily to sensitivity analyses is a strength and in some circumstances can overcome concerns
about the precision of the probability estimates and the validity of utilities.
=However, while one-dimensional sensitivity analyses are straightforward to conduct and interpret, the numbers of potential multi-dimensional analyses
(which address the effect of varying more than one factor at a time) can become very large and the results difficult to present and understand.
5 A single criterion approach: in most examples of Decision Analysis in the literature the utility of each possible outcome is expressed as a single figure. The task is to choose the option that maximizes the expected value of this utility, so DA is mainly used as a single-criterion decision method.
-but many problems naturally involve more than one outcome criterion or attribute
|