2.2.1Decision-trees
Decision-trees are simple directed graphs consisting of branches showing a clinical problem, decision nodes presenting a decision maker’s choice between competing strategies, chance nodes determining the consequences of treatment probabilistically and terminal nodes denoting the final outcome state associated with each possible pathway[67]. The object that is usually modelled is a hypothetical homogeneous cohort. The parameters of the tree have either the average or probability distributions, which are generally constant over time.
If the given SDDP is simple enough to be presented diagrammatically as a successive decision-tree, a decision-tree can be used to solve the SDDP[36, 68-70]. The successive decision-tree consists of 1) branches showing the disease pathway or the treatment pathway, 2) chance nodes determining the probability of treatment success or failure and 3) decision nodes allowing sequential treatment decisions. The main assumption is that a drug is continuously used if it is effective for the previous period and switched if it is not effective (see Figure 2.). However, the decision-tree would quickly become unmanageable where there are more time periods, more health states or more drug alternatives. Moreover, decision-trees may be of limited value to represent real world SDDPs, which inevitably involve the functionalities to memorise current and previous information and to carry information over time, and to perform logical evaluations based on this information.
Figure 2.. An example successive decision tree model for SDDPs
Dostları ilə paylaş: |
