Online Program

Return to main conference page

Friday, January 12
Fri, Jan 12, 10:30 AM - 12:15 PM
Crystal Ballroom B
Instrumental Variables and Treatment Effect Heterogeneity

Using Conceptual Modeling to Plan and Execute Quasi-Experimental Studies (304214)

*Annie Simpson, Medical University of South Carolina 
Kit N. Simpson, Medical University of South Carolina 

Keywords: conceptual modeling, quasi-experimental study design, hearing loss

Observational studies are often judge of being lower quality in part because they are not randomized but also because they are not planned “a priori”. We examine what “a priori” means in terms of quasi-experimental study design. Health services researchers are trained to understand foundational theoretical frameworks such as Donebedian’s “Structure-Process-Outcomes” model for examining quality of care. These models are taught as a precursor to understanding how to develop conceptual models that help justify the reason for a study as well as lay out the factors that might influence results or add bias. An excellent way to support “a priori” planning of quasi-experimental studies is to generate a conceptual model before any data are touched. These models make it easier to ensure that selection is well thought out and adjustment is considered separately. We use a study of the effects of age-related hearing loss on health outcomes to illustrate when one might need multiple models, and how to bring these models into the practice of study design and analysis. We compare and contrast the effect of different choices on study results and discuss the idea of “designing away the effect”.