FACTORED STOCHASTIC TREES - A TOOL FOR SOLVING COMPLEX TEMPORAL MEDICAL DECISION-MODELS

The stochastic tree is a continuous-time version of a Markov-cycle tre e, useful for constructing and solving medical decision models in whic h risks of mortality and morbidity may extend over time. Stochastic tr ees have advantages over Markov-cycle trees in graphic display and com putational solution. Like the decision tree or Markov-cycle tree, stoc hastic tree models of complex medical decision problems can be too lar ge for convenient graphic formulation and display. This paper introduc es the notion of factoring a large stochastic tree into simpler compon ents, each of which may be easily displayed. It also shows how the rol lback solution procedure for unfactored stochastic trees may be conven iently adapted to solve factored trees. These concepts are illustrated using published examples from the medical literature.