Bayesian Estimation of the Costs of Optimal Allocation of a Surgical Procedure

• Activity Number: 397
• Type: Topic Contributed
• Date/Time: Thursday, August 12, 2004 : 8:30 AM to 10:20 AM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Marshall M. Joffe*+ and Michael R. Elliott
• Companies: University of Pennsylvania and University of Pennsylvania School of Medicine
• Keywords: Bayesian statistics ; causality ; classification

Abstract

Subjects with end-stage renal disease requiring hemodialysis need a way to access their vascular system. When successful, the construction of a naïve arteriovenous fistula (AVF) is the best way to do this; unfortunately, the procedure is often unsuccessful. It is thus important to determine when the procedure is likely to be successful. One can use information on costs to choose an optimal rule for deciding when to perform the procedure. We consider estimation of the costs of different decision rules and contrast Bayesian with frequentist approaches when the same data are used for choosing the rule and estimating the costs. Naïve application of frequentist methods results in overoptimism about the costs of the selected rule; corrections are required. These corrections are not generally required in Bayesian methods. We illustrate these points using data from a study of surgery to create an AVF.