The Role of Statistics in Comparative Effectiveness Research

• Activity Number: 11
• Type: Invited
• Date/Time: Sunday, July 31, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Section on Health Policy Statistics
• Author(s): Lilly Yue*+ and S. Stanley Young*+ and Allen Heller*+ and Donald Rubin*+
• Companies: National Institute of Statistical Sciences and U.S. Food and Drug Administration and Bayer HealthCare Pharmaceuticals Inc. and Harvard University
• Keywords: causal inference ; comparative effectiveness research ; administrative data ; observational data

Abstract

Comparative effectiveness research is taking on a very important role in US health care reform. The basic idea of this approach is to take advantage of existing data sets, which are primarily observational in nature, to infer the relative effectiveness of medical interventions. These data sets, however, typically have severe limitations for drawing reliable causal inferences about medical interventions. For example, administrative hospital insurance claims data only include billable charges for patients who are in hospital and exclude patients who are not in hospital -- how can such data be used to compare outcomes for patients taking one drug versus another, even if the data for in-hospital patients are perfectly accurate? For another example, is it reasonable to assume that if a medical condition is not noted, it is absent? And of course, observational data, even when accurately record