Decision analysis with cumulative prospect theory

Background. Individuals sometimes express preferences that do not follow ex pected utility theory. Cumulative prospect theory adjusts for some phenomen a by using decision weights rather than probabilities when analyzing a deci sion tree. Methods. The authors examined how probability transformations fr om cumulative prospect theory might alter a decision analysis of a prophyla ctic therapy in AIDS, eliciting utilities from patients with HIV infection (n = 75) and calculating expected outcomes using an established Markov mode l. They next focused on transformations of three sets of probabilities: 1) the probabilities used in calculating standard-gamble utility scores; 2) th e probabilities of being in discrete Markov states; 3) the probabilities of transitioning between Markov states. Results. The same prophylaxis strateg y yielded the highest quality-adjusted survival under all transformations. For the average patient, prophylaxis appeared relatively less advantageous when standard-gamble utilities were transformed. Prophylaxis appeared relat ively more advantageous when state probabilities were transformed and relat ively less advantageous when transition probabilities were transformed. Tra nsforming standard-gamble and transition probabilities simultaneously decre ased the gain from prophylaxis by almost half. Sensitivity analysis indicat ed that even near-linear probability weighting transformations could substa ntially alter quality-adjusted survival estimates. Conclusion. The magnitud e of benefit estimated in a decision-analytic model can change significantl y after using cumulative prospect theory. Incorporating cumulative prospect theory into decision analysis can provide a form of sensitivity analysis a nd may help describe when people deviate from expected utility theory.