How do we know that shorter sleep causes lower mortality, instead of, say, that better health or some genes cause both?

Even if we assume that correlation is causation, I haven’t seen much of that evidence.