Prospective decision analysis modeling indicates that clinical decisions in vascular surgery often fail to maximize patient expected utility

BACKGROUND: Applied prospectively to patients with peripheral arterial disease, individualized decision analysis has the potential to improve the surgeon's ability to optimize patient outcome. METHODS: A prospective, randomized trial comparing Markov surgical decision analysis to standard decision-making was performed in 206 patients with symptomatic lower extremity arterial disease. Utility assessment and quality of life were determined from individual patients prior to treatment. Vascular surgeons provided estimates of probability of treatment outcome, intended and actual treatment plans, and assessment of comfort with the decision (PDPI). Treatment plans and PDPI evaluations were repeated after each surgeon was made aware of model predictions for half of the patients in a randomized manner. RESULTS: Optimal treatments predicted by decision analysis differed significantly from the surgeon's initial plan and consisted of bypass for 30 versus 29%, respectively, angioplasty for 28 versus 11%, amputation for 31 versus 6%, and medical management for 34 versus 54% (agreement 50%, kappa 0.28). Surgeon awareness of the decision model results did not alter the verbalized final plan, but did trend toward less frequent use of bypass. Patients for whom the model agreed with the surgeon's initial plan were less likely to undergo bypass (13 versus 30%, P < 0.01). Greater surgeon comfort was present when the initial plan and model agreed (PDPI score 47.5 versus 45.6, P < 0.005). CONCLUSIONS: Individualized application of a decision model to patients with peripheral arterial disease suggests that arterial bypass is frequently recommended even when it may not maximize patient expected utility.