Really glad you are asking these questions about establishing causality regarding social determinants of health. It is definitely tough - probably tougher than some domains of epi (e.g., clinical settings where you sometimes have good info about how the exposure is allocated and get to do RCTs relatively frequently) but not worse than others (e.g., nutritional epi - really tough to disentangle). My view is that quasi-experimental approaches leveraging events such as policy changes, lotteries due to insufficient resources, or other apparently arbitrary factors that influence social exposures will provide the most compelling evidence. Usually those studies have their own limitations so are best interpreted in combination with more conventional observational studies. Rita Hamad will give the social policy lecture in the class - she is a leader in using this type of approach so hopefully will give you a flavor for it.
Maria