A new approach to causal inference in mortality studies with a sustained exposure period–application to control of the healthy worker survivor effect
Regression models for cohort mortality studies,