health services research, health research methods, health policy research, statistical methods for observational data,
cardiovascular outcomes studies, randomized trials, accountable care systems
Further Information at Dartmouth Medical School
Senior Scientist, Institute for Clinical Evaluative Sciences (ICES)
Senior Scientist, Evaluative Clinical Sciences, Sunnybrook Research Institute
Professor, Dartmouth Institute for Health Policy and Clinical Practice (TDI), Geisel School of Medicine at Dartmouth, Hanover, NH, USA
1) We created virtual Multispecialty Physician Networks consisting of primary care physicians, specialists, interdisciplinary health providers and hospitals using health administrative data. These self-organizing, virtual networks could form the basis of “systems of care” for chronic disease patients as they link providers who are already working together to care for a defined population. Strengthening these existing links may be an efficient way to build networks of providers to provide efficient care for chronic disease patients. The Ontario Ministry of Health used the conceptual ideas behind these networks as the basis of Ontario’s Health Links. http://www.huffingtonpost.ca/therese-stukel/canada-health-care-system-physician-networks_b_3327655.html
2) The association between higher spending and better outcomes is one of the more important health policy questions. We demonstrated that U.S. regions with higher spending had neither better quality of care nor increased survival. This study received international media attention and was selected by NIH for presentation to the U.S. Congress as one of the top 10 research articles in 2003 having significant impact on science. We replicated this study in Canada, a system with universal access to healthcare but lower spending, fewer specialists, and selective access to medical technology. Canadian results showed that higher spending hospitals had better outcomes and higher quality of care, the reverse of what was found in the U.S.
3) We demonstrated that standard statistical methods (traditional regression and propensity score matching) may fail to remove unmeasured confounding and incur survival bias in observational studies in a methodological paper published in a leading medical journal, and demonstrate the utility of instrumental variables methods. This paper was accompanied by an editorial and was nominated for Lancet’s 2007 Paper of the Year.