Evaluating cost-effectiveness of interventions using observational studies
Observational data from studies in which allocation of individuals to different interventions is not random are increasingly being used for the purpose of cost-effectiveness analysis. The increased availability of such data electronically (e.g. disease registries or routinely collected data) and their presumed superior external validity have been the main contributors to this trend. These studies are open to selection bias, when factors determining allocation of individuals to different interventions also determine health and other outcomes. Current work in HERC focuses on studying statistical methods to adjust for selection bias, including propensity scores matching and instrumental variables, and their performance.
Mihaylova Borislava, Pitman Richard, Tincello Douglas, van der Vaart Huub, Tunn Ralf, Timlin Louise, Quail Deborah, Johns Adam, and Sculpher Mark (2010) Cost-effectiveness of duloxetine: the Stress Urinary Incontinence Treatment (SUIT) study. Value Health, 13(5):565-72