A Hierarchical Decision Support System for Farming Clinical Evidence

• Activity Number: 7
• Type: Invited
• Date/Time: Sunday, August 3, 2003 : 2:00 PM to 3:50 PM
• Sponsor: WNAR
• Author(s): Xiaowei Yang*+ and Hongquan Xu and Naihua Duan
• Companies: University of California, Los Angeles and University of California, Los Angeles and University of California, Los Angeles
• Keywords: decision support system ; clinical evidence ; propensity score ; data farming

Abstract

A hierarchical decision system that incorporates local knowledge from individual clinical sites and global knowledge from external clinical trials is proposed to assist clinicians' real-time treatment decision-making. At the local level, a propensity score (Rosenbaum and Rubin 1983) approach is proposed to deal with the endogenous selection bias that occurs in observational studies. A propensity score is defined as the conditional probability of a patient assigning to a particular treatment given the pretreatment covariates. Propensity scores are used to match patients with similar covariates so that treatment effects for behavioral or pharmaceutical therapies can be compared. At the global level, an empirical Bayes method is proposed to combine information from geographically distributed local sites and external clinical trials so that both local and global evidence are used properly in the treatment decisions. The proposed system is applied to data collected from Project IMPACT's Clinical Information System (Unutzer, et al. 2002) which has about 900 patients and 10,000 encounters across eight clinical sites.