Medical Decision Making Using the Analytic Hierarchy Process: Choice of Initial Antimicrobial Therapy for Acute

The analytic hierarchy process (AHP) was used to determine which of seven recommended antibiotic regimens represented optimal initial therapy for a young woman hospitalized for treat-ment of acute pyelonephritis. The model included the following criteria: maximize cure, minimize adverse effects (broken down into very serious, serious, and limited), minimize antibiotic resis-tance, and minimize cost (divided into total cost and patient cost). The criteria were weighted according to judgments made by 61 practicing clinicians. Alternatives were compared relative to the criteria using published information on the expected frequencies of urinary pathogens and drug toxicity, local antibiotic sensitivities and antibiotic charges, and expert opinion regarding their propensities for inducing antimicrobial resistance. The analysis identified ampicillin com-bined with gentamicin as the optimal regimen. This study illustrates several features of the AHP that make it promising for use in medical decision making: its ability to incorporate multiple criteria into a formal decision model, its procedural simplicity, and its similarity to current patient management guidelines. Further studies to establish the role of the AHP in medical decision making are warranted. Key words: analytic hierarchy process; decision making; pyelonephritis; decision analysis. (Med Decis Making 1989;9:51-56) The analytic hierarchy process (AHP), a decision making method developed during the 1970s by Thomas Saaty, has at least three features that make it appealing for use in medical decision making: it can accom-modate the multiple factors, subjective and objective, that typically influence clinical decisions; the basic evaluative technique used in the AHP, sequential pair-wise comparisons, is simple and easy to understand, and the hierarchical structure of the AHP model re-sembles familiar clinical management guidelines. To illustrate the use of the AHP in medical decision making, I used it to determine optimal initial antibiotic therapy for a young woman hospitalized for treatment of acute uncomplicated pyelonephritis