Thank Nick - I would add effect modifier to the list of how SES can be treated. Also, SES as a "confounder" is complicated in the ways you describe.
The C-section rate is a great example of a health services context in which multilevel analysis can be useful, or even essential. One of the complex factors of course is that these different levels likely interact with each other (i.e. "baby friendly hospitals" might have different effects depending on demographics of the patients) which can be very challenging to tease out.