Premium
CLINICAL INFERENCES AND DECISIONS—III. UTILITY ASSESSMENT AND THE BAYESIAN DECISION MODEL
Author(s) -
Aspinall P. A.,
Hill A. R.
Publication year - 1984
Publication title -
ophthalmic and physiological optics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.147
H-Index - 66
eISSN - 1475-1313
pISSN - 0275-5408
DOI - 10.1111/j.1475-1313.1984.tb00363.x
Subject(s) - weighting , computer science , bayesian probability , decision theory , normative , process (computing) , outcome (game theory) , normative model of decision making , principal (computer security) , econometrics , artificial intelligence , mathematical economics , statistics , mathematics , medicine , philosophy , epistemology , radiology , operating system
It is accepted that errors of misclassifications, however small, can occur in clinical decisions but it cannot be assumed that the importance associated with false positive errors is the same as that for false negatives. The relative importance of these two types of error is frequently implied by a decision maker in the different weighting factors or utilities he assigns to the alternative consequences of the decisions. Formal procedures are available by which it is possible lo make explicit in numerical form the value or worth of the outcome of a decision process. The two principal methods are described for generating utilities as associated with clinical decisions. The concept and application of utility is then expanded from a unidimensional to a multidimensional problem where, for example, one variable may be state of health and another monetary assets. When combined with the principles of subjective probability and test criterion selection outlined in Parts I and II of this series, the consequent use of utilities completes the framework upon which the general Bayesian model of clinical decision making is based. The five main stages in this general decision making model arc described and applications of the model are illustrated with clinical examples from the field of ophthalmology. These include examples for unidimensional and multidimensional problems which are worked through in detail to illustrate both the principles and methodology involved in a rationalized normative model of clinical decision making behaviour.