Decision Analysis: A Toy or a Tool for Clinical Practice?

Decision analysis: A toy or a tool for clinical practice? I n their paper evaluating cefepime monotherapy, Halpern et al (pages 19-2 7) use a decision analysis modelling method to determine costs and outcomes associated with this therapy. Decision analysis has its origins in the disciplines of economics and management studies (1) and, over the past two decades, has been applied to many aspects of medical practice (2,3). The main benefit of clinical decision analysis (CDA) is that it forces one to lay out and analyze critically approaches to therapy for a given disease state. One group has gone so far as to raise the methodology to the level of a clinical consultation service (5). Pharmacoeconomic evaluation is a more recent variation of CDA that has developed as we try to improve our decision making mechanisms. By definition , decision analysis is a process; that is , a systematic approach to decision making under conditions of uncertainty (eg, where there is imperfect or incomplete information available) (5). It allows one to link choices , actions and outcomes in an effort to improve the decision made. The process, more commonly referred to as CDA modelling, is described by a series of fundamental stages: one identifies that there is a need to make a choice from among a number of options; each option leads to multiple consequences; the probability or uncertainty associated with each consequence is estimated in some manner; and one assesses the value attributable to each consequence (via costs, utilities or some measure of effectiveness). CDA modelling involves pulling together related information from a variety of data sources to create a scenario reflecting the choice(s) which must be made and its (their) consequences. Where information is not specifically available, assumptions are made based on clinical experience, assumptions that are then subject to challenge during sensitivity analysis. Sensitivity analysis is an integral part ofCDA, in that it tests the model for robustness (a reliability issue) and, some would argue, for validity. It is imperative that the estimates and/or assumptions used for key parameters in the model (especially those based on 'opinion ' or where there is a high degree of uncertainty or variability) be subjected to sensitivity analysis. One of the major advantages of CDA methodology is its ability to change decision making from a process that is subjective and nebulous into a concrete entity that is capable of identifying the decision option that should be chosen based on the information put into the model. The fundamental premise underlying decision analysis is the belief that the tradition-