Utilizing External Theory and Historical Data to Make Estimates of Comparative Effects More Justified

• Activity Number: 308
• Type: Invited
• Date/Time: Tuesday, July 31, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Health Policy Statistics Section
• Author(s): James O'Malley*+ and Jaeun Choi and Mary Beth Landrum and Bruce E Landon
• Companies: Harvard Medical School and Harvard Medical School and Harvard Medical School and Harvard Medical School
• Keywords: Comparative effectiveness ; instrumental variable ; sensitivity analysis ; structural equation model ; unmeasured confounder ; volume-outcomes

Abstract

The purpose of this talk is to illustrate how external information can be used to support comparisons of treatments in observational settings. Two cases will be considered. The first involves estimation of heterogeneous treatment effects under a structural equations model in which one equation relates predictors (including treatment) to the outcome and a second equation describes the treatment selection mechanism. A theoretical argument based on expert clinical knowledge of the relationship of unobserved variables influencing selection to the outcome informs model choice. The second case involves an instrumental variables analysis in which there are doubts over the legitimacy of the instruments. The process of determining values for the sensitivity parameters is described, including how historical data can be used, and the implications for the region of uncertainty about the treatment effect is depicted. Methods are illustrated using real data from vascular surgery and mental health respectively.