Interpretations and Research Needs
This discussion of the interpretation of the findings in Chapter 5 starts with problems inherent in the use of ecologic studies, along with various "errors of measurement" ( MacMahon and Pugh, 1970). A consideration of potential explanations involving the concentration or neighborhood effects of segregation follows. Because the ecologic studies presented in this book serve mainly to generate hypotheses, the discussion is speculative. It involves mainly a review of the literature on the effects of discrimination in such areas as medical (health) care, quality of life issues for blacks (especially in inner cities), exposures to stresses, and certain psychosocial and physiologic responses to stress. This review deals with the epidemiology of discrimination and racism. It is intended to stimulate further studies of the association between segregation and the health of blacks. The final discussion concerns the need for further ecologic studies, with more detailed data than were available for Chapter 5, along with nonecologic (or "analytic") epidemiologic studies that obtain information on individuals.
The ecologic analyses of geographic variation in death rates in Chapter 5 also raised issues that go beyond the issue of the association between mortality and segregation or discrimination and racism. This is especially true for time trends in black infant mortality rates in different regions of the country, which are not explained by trends in segregation. Hypotheses involving temporal and regional changes in health care delivery, albeit speculative, are discussed in this chapter.
Issues of biases in ecologic studies have been considered mainly with regard to multiple linear regression models ( Greenland and Robins, 1994a,b) such as those used in Chapter 5. A major problem concerns the inability to control