Date on Master's Thesis/Doctoral Dissertation


Document Type

Doctoral Dissertation

Degree Name

Ph. D.


Bioinformatics and Biostatistics

Committee Chair

Evans, Gerald W.

Author's Keywords

Medical decisions; Decision-making; Stochastic tree


Medical care--Decision making; Decision making--Simulation methods


A variety of methodologies have been employed for decision making related to the treatment of diseases/injury. Decision trees are a functional way in which to examine problems under uncertainty by providing a method to analyze decisions under risk (Detsky, 1996, 97). However, conventional decision trees do not completely represent "the real world" since they cannot investigate problems that are cyclic in nature (Jaafari, 2003). The stochastic tree that developed Hazen during 1992-to-1996 is one of the most relevant methods and techniques related to decision analyses that append more incorporation for medical intervention related to recurring diseases/injuries. "The approach combines features of continuous-time Markov chains with those of decision trees and that enable time to be modeled as a range where health state transitions can occur at any instant" (Hazen 1992-to-96). It can also accommodate patients' preferences regarding risk and quality of life. In this research we enhance Hazen's stochastic tree by developing an analytical model, and we extend its capabilities more by developing multi-objective simulation based methodologists for medical decision making. First, with our enhancement on the Hazen's stochastic tree, the model is improved by utilizing the Weibull Accelerated Failure Time model. This new technique will fill the gap between the experimental circumstances and the corresponding circumstances or conditions of standard/current treatment. Second, as simulation can be a final alternative for problems that are mathematically intractable for other techniques (Banks 1996), our multi-objective simulation based model for medical decision making extends the capabilities of Hazen stochastic tree. It adds more flexibility with the use of survival distributions for health states sojourn, and combines two sound theories: multi attribute utility (MAU) theory, and Ranking-Selection procedures. Indeed, our simulation model (considering patient's profile/preferences and health states survival/quality/cost, QALY) presents an investigation of the use of simulation on the stochastic tree, with associated techniques related to ranking and selection, and multi-objectives decision analysis.