F. J. Díez. Medical decision analysis with probabilistic graphical models. Institute of Technology Assessment. Boston, October 14, 2015.
73 slides (37 pages). PDF (1.2 MB).
Probabilistic graphical models offer important advantages over traditional methods for medical decision analysis: 1. Bayesian networks are much more accurate than the naive Bayes method, which consists of multiplying the likelihood ratios of the findings. 2. Influence diagrams and decision analysis networks can perform unicriterion analyses and cost-effectiveness analyses with models equivalent to decision trees containing embedded decision nodes and thousands of branches. 3. Markov influence diagrams can solve problems much more easily than when using spreadsheets, Markov decision trees, or a programming language such as R or C++. They can model various features of the patient without multiplying the number of states; in particular, they can represent the patient history without using tunnel states or microsimulation. 4. Factored Markov decision processes are especially suited for follow-up problems; for example, to determine the optimal time for an intervention. This seminar will offer an overview of probabilistic graphical model and show several examples built at the Center for Intelligent Decision-Support Systems (CISIAD) of the National University for Distance Education (UNED) in Madrid, Spain, as well as an open-source software package for decision analysis, OpenMarkov, built at the same center.