Jin, this is an interesting example: most of the research on education and health focuses on relatively crude aspects of education, such as how many years or degrees someone completes. The question of the health impacts of affirmative action presents extra challenges, including that it is difficult to identify the "exposed" and "unexposed" but also that affirmative action may have important spillover effects, i.e., benefit people who were not direct beneficiaries because of role models or social norms achieved via improved diversity in higher education. On the other hand, affirmative action is a social policy that directly adversely affects some individuals, so it is tricky to decide how to evaluate it.
I like your idea of using information on the admissions committee decision. Some universities or states may have hard criteria for eligibility that could be used. I bet there has been some good research on this topic looking at labor market outcomes, such as earnings, so we may be able to find a good design in another field that could be applied to health.
-maria