Bayesian probability encoding in medical decision analysis

Abstract

The aim of this dissertation is to present two classes of Bayesian probability-encoding models useful for decision analysis within the context of evidence-based medicine (EBM) and in relation to the multi-facet nature of medical evidence. It allows both subjective and objective evidences be combined to provide a more complete approach for probability-encoding. The proposed Bayesian models are designed for routine use in clinical decision-making. One deals with patient-level data while the other handles aggregate-level reported evidences. Both conjugacy and Markov chain Monte Carlo are applied for the search of posteriors and several models do possess close-form solutions, thanks to the Jacobian technique. The practicability of the proposed models is illustrated with ten clinical decision problems commonly seen in Singapore. The dissertation also provides an overview of future EBM practices and potential areas for Bayesian methodological research.