Death and the sleepless resident
It turns out that work-hour regulations for residents do nothing one way or the other for patient mortality. (Or almost nothing.) I'm not sure what this tells us. Were work-hour regulations really instituted on the theory that this would directly benefit patient mortality? There are other indices, I think, which are more likely affected (patient-provider relationships, morbidity, error rate), even though they are harder to study.
Take two different patients who both die in their second week of hospitalization. Patient A is cared for by well-rested residents; patient B, by zombies. I think patient A received the better care by a number of measures, even if mortality-wise they're even.