East Coast Ballroom
A comparison of synthetic control approaches to control for unobserved confounding using observational data: Evaluating the impact of redesigning urgent and emergency care in Northumberland, England (307838)
*Geraldine M Clarke, The Health FoundationO'Neill Stephen, National University of Ireland Galway
Adam Steventon, The Health Foundation
Keywords: synthetic controls, difference-in-difference, routine data, public health, statistical methods
Unobserved confounding is a key concern when evaluating the effects of health system reforms using observational data. Researchers face challenges deciding which method is most appropriate for robust causal evaluation. This paper contrasts the widely-used difference-in-differences method and the original synthetic control approach with a variety of novel approaches (Generalized, Micro and Bayesian structural time series), all of which use synthetic control style methods to construct counterfactuals. We make comparisons across a range of simulation scenarios, and apply the methods to evaluate the impact of a new specialist urgent and emergency care hospital in Northumberland, England, which opened in 2015 under the National Health Service (NHS) England New Care Models programme.