HIERARCHICAL MODELING FOR ASSESSING BREAST CANCER CARE

点击翻译按钮获取中文摘要

The research proposed in this application is designed to address several key methodological issues in the epidemiology of breast cancer care using a new class of statistical models called hierarchical models. We will illustrate their application to two data sets of importance in the epidemiology of breast cancer care. Although breast cancer care is the primary area for development of these new methodologies, we show that hierarchical models permit a wide range of epidemiological problems to be approached from a common conceptual foundation. the specific problems addressed by these new methods will include an examination of the degree of compliance with clinical guidelines for breast cancer care and quantification of the magnitude of variation in compliance rates across hospitals and providers. Hierarchical models permit investigators to incorporate multiple sources of variability in the data, and to pool across units (e.g., hospitals, physicians), thus capitalizing on similarities within units while accounting for differences across units. thus, they may be used to provide better inferences for individual units by borrowing strength from the ensemble. We will compare the methods proposed in this grant with conventional methods for assessing breast cancer care to facilitate an understanding of the potential advantages of hierarchical models. The research described in this proposal will bring together a network of statistical, epidemiological and medical researchers, including an Advisory Board of leaders in each of these areas, in an effort to ensure validation and integration of these new methods in epidemiology research. We will disseminate our methods and results through a series of papers and meetings with the Advisory Board, and through presentations at scientific meetings.