There is an extensive literature on causal inference that goes back decades but is experiencing a recent boom. One of the recent foci is on when regression coefficients can be considered estimates of causal coefficients and when that is inappropriate.
For background, you might look to work by Rubin and collaborators. Pearl is doing some leading philosophical/conceptual work on the topic. Sobel and many others are bringing the work to more applied audiences. I don't know epidemiology literature to give you names, but there's a strong emphasis on appropriate causal inference in that area as well.
There's far more depth than can easily be presented in a forum like this, so I'm afraid you have a lot of reading ahead of you!