2. Epidemiologic adjustments for confounding
when contrasting racial/ethnic groups are
necessarily inadequate and ultimately
uninterpretable because it is impossible to define
the conditions under which confounding would
be adequately “controlled”
• notion of a good “substitute population”
• exchangeability condition
NB: Unadjusted contrasts are generally
unproblematic (i.e., surveillance)
Adjusted contrasts can be interpreted as merely
forms of stratified surveillance, but do not
generally provide any valid etiologic inference...
| Previous slide | Next slide | Back to first slide | View graphic version |