Turtles all the way down: do biological mechanisms for epidemiological observations always matter?