This dissertation proposal is concerned with the degree to which the clinical spectrum of diabetic patients seen in referral practice is distorted compared to that occurring in the community. The results of this research will serve to identify and quantify the distortions which arise from the effects of referral bias so that valid inferences can be drawn from the results of referral-practice based research and so that these findings can be appropriately related to the spectrum of patients seen in the community. In addition, this study will provide the first community population-based study of survival in Type II diabetes in the United States. While referral bias is likely to greatly distort the clinical spectrum of Type II diabetes, the health services research issues raised in this dissertation have direct relevance to the validity and generalizability of referral center patient-based research for many other medical conditions. Specifically, survival characteristics in the Rochester, Minnesota, 1/1/1980 prevalence cohort of Type II diabetics will be compared with Type II diabetic paients from the Mayo Clinic referral practice who were seen between 7/1/79-6/30/80.
The specific aim will be to identify the extent to which the differences in survival in the two groups of diabetic patients are attributable to differences in the distribution and/or differences in the effect of prognostic factors for survival in Type II diabetes. Consequently, this dissertation will serve to assess the extent to which studies concerned with survival characteristics and prognostic factor estimation for Type II diabetics seen in referral centers are generalizable to diabetic patients who receive medical care in community practice settings. The unique data resource of the Mayo Clinic and the Rochester Epidemiology Project provides the only research environment in the United States in which these crucial studies can be efficiently executed in order to provide essential definitive answers concerning the effect of referral bias on the clinical spectrum in Type II diabetes. The methodology that will be developed, in collaboration with a Dissertation Committee which is tremendously rich in the expertise necessary for this type of health services research, will serve as a model for the evaluation of referral bias for many other diseases.
