What characterises a useful concept of causation in epidemiology?

What characterises a useful concept of causation in epidemiology?