In the absence of randomized trials, analyses of observational (non-experimental) data can provide timely results to fill the pressing need for information on comparative effectiveness. But unlike well-conducted randomized trials, observational studies may be biased due to differences between those that receive one treatment rather than another. Doubly robust (DR) estimation is a novel statistical method which combines two regression models to provide some additional protection against bias due to confounding. The proposed research will systematically evaluate six key aspects of DR estimators: 1) quantify the degree to which DR estimation leads to bias reduction relative to standard methods, 2) evaluate estimated standard errors under a broad range of conditions, 3) determine the optimal covariate types for inclusion in the component models, 4) validate a new approach to variable and model-form selection, 5) develop and test strategies specific to DR estimation for the identification of non-uniform treatment effects, and 6) identify conditions under which DR methods perform worse than standard methods. Dr Jonsson Funk will use more realistic and complex data as the basis for the Monte Carlo simulations by creating a hybrid of covariate data from de-identified records of actual patients combined with simulated exposures and outcomes. Dr Jonsson Funk will employ the DR method in three comparative effectiveness analyses of large healthcare databases to evaluate its use in real-world applications, and ultimately disseminate 'best practices'for using DR estimation in comparative effectiveness research. In addition to the comprehensive evaluation of DR methods, this award would provide protected time for the candidate to undertake formal and informal training in the use of large healthcare databases to complete the skill set needed for a productive career in comparative effectiveness research. Drawing on rich local resources including Decide and EPC sites, experts in advanced methods, clinical colleagues, and key healthcare databases, the proposed portfolio of research will make important advances in comparative effectiveness research and firmly establish the candidate as an independent investigator. Making an informed choice among treatment options depends on the availability of reliable data about the comparative effectiveness and safety of the alternatives. Dr Jonsson Funk will develop and disseminate 'best practices'for using a new statistical method in an effort to deliver more robust and relevant information in a timely manner to support patients, providers, payers and policy-makers making critical decisions about healthcare.
Showing the most recent 10 out of 23 publications
