Can decision analysis adequately represent clinical problems?

Abstract

A major weakness of medical decision analysis has been the inability of the commonly used single attribute utility models to adequately represent clinical decision making situations. To illustrate this problem, I reanalyzed a well known decision analysis that is widely interpreted as proof that two decision alternatives are equivalent in all clinically meaningful respects. The reanalysis was based on a more representative decision model made possible by the use of the analytic hierarchy process (AHP), a multiobjective decision making technique. The use of this model resulted in the identification of a clearly preferred alternative, indicating that the results of the original analysis have been widely misinterpreted. The degree to which a decision model represents clinical reality influences the correct interpretation of a decision analysis. Limited decision models can yield only limited conclusions. The use of more representative multiobjective decision models would improve the clinical usefulness of medical decision analyses.