: This proposal explores a new method for describing and understanding racial disparities through the use of """"""""Tapered Multivariate Matching"""""""" (TMM). TMM represents a new theory and conceptual framework for examining racial disparities that will aid in better measuring and more clearly describing racial disparities as well as better isolating and identifying specific features of the healthcare system that lead to differences in quality of care across racial groups. Presently, the size of a disparity is typically measured by reporting a regression model coefficient on race. However: (1) Minorities are, by definition, underrepresented in these models because whites generally outweigh blacks in most analyses, which can tend to weight model coefficients from a white perspective. At the same time, minorities often present with very different covariate distributions;therefore, (2) the interactions needed to properly define the relationships between race and explanatory covariates are usually not analyzed, or if analyzed, not reported. It therefore becomes very difficult for policy analysts to clearly convey to the public the true extent of the racial disparity. TMM starts with the minority population (say the black or Hispanic population) and, using multivariate matching, matches white patients to the entire minority group. Matching is done sequentially: for example, starting matching at diagnosis, then matching at both diagnosis and treatment. In so doing, TMM can, for example, directly compare survival for whites who present at diagnosis similarly to their matched black pairs (using again all blacks), and then compare whites who were matched on both diagnosis and treatment variables, again matched to the same black patients (all blacks). For reporting, TMM can decompose Total Racial Disparity at, say, 5 years from diagnosis (DTOT) into the sum of the disparity in survival due to diagnosis differences between whites and blacks (DDx) plus differences in survival due to initial treatment (DTx) plus a residual disparity (DR), or DTOT = DDx + DTx + DR. These components are more clearly understood than reporting a complex, fully interacted model, and they provide direct information on aspects of the treatment of patients that can be changed to improve quality.
The aims of the study are to apply the TMM framework to breast and colorectal cancer (though TMM can be used on any medical problem beyond cancer) using the national Medicare-SEER database and demonstrate the usefulness of TMM in both identifying specific aspects of patient care that lead to quality differences by race, and to demonstrate how TMM results can be conveyed in a clear and understandable report that can be used to help reduce racial disparities. Furthermore, the project will create a website to facilitate the understanding and implementation of TMM. Through TMM, we aim to establish a new and better method to measure and describe disparities in patient treatment and outcomes for all areas of medicine and in so doing, shed light on etiologies for these quality differences and on how to reduce them.
This proposal explores Tapered Multivariate Matching, a new theory and conceptual framework for examining racial disparities and quality of care. This approach will aid in determining etiologic mechanisms and will guide policy recommendations. Through Tapered Matching, we believe we can propose a new method to better measure racial disparities in the treatment and outcomes of cancer patients, and all other patients, and in so doing, shed light on etiologies for those disparities and how to reduce them.
| Silber, Jeffrey H; Rosenbaum, Paul R; Ross, Richard N et al. (2014) Racial disparities in colon cancer survival: a matched cohort study. Ann Intern Med 161:845-54 |
| Silber, Jeffrey H; Rosenbaum, Paul R; Ross, Richard N et al. (2013) Racial disparities in operative procedure time: the influence of obesity. Anesthesiology 119:43-51 |
| Silber, Jeffrey H; Rosenbaum, Paul R; Clark, Amy S et al. (2013) Characteristics associated with differences in survival among black and white women with breast cancer. JAMA 310:389-97 |
