Is Generalizability a Misnomer? Not Just Semantics

• Activity Number: 216
• Type: Invited
• Date/Time: Monday, August 2, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Health Policy Statistics Section
• Author(s): Herbert Weisberg+ and Victor P. Pontes* and Stacey A. Missmer
• Companies: Correlation Research, Inc. and Inatec, Inc. and Brigham and Women's Hospital/Harvard Medical School
• Keywords: causal model ; individual effects ; generalizability ; sampling bias

Abstract

Generalizability connotes the idea that a "true" effect found in a study holds in a broader target population. From this perspective, differences among results of several studies are often attributed to methodological problems. However, if the causal effect can vary across individuals, the study effect depends intrinsically on the nature of the study population. The varying study results may all be valid, but relevant to different populations. Theoretical implications of this alternative perspective will be demonstrated with a causal model. For the members of a target population, the model specifies potential outcomes that describe individual-level causal effects. The implications of this perspective can be surprising, as illustrated by application to a major current public health dilemma: weighing the benefits and risks of aspirin therapy for prevention of heart disease.