Abstract
Since their inception, small area studies intended to measure health system performance have been challenged by concerns that regional variation in health care may be primarily explained by differences in patient health risk. Controlling for regional population differences depends on appropriate risk adjustment, but the adequacy of the methods used in early analyses was contested. A novel response to these concerns was the development of end-of-life cohorts by Dartmouth Atlas investigators. These were used initially to control for differences in population health status in studies investigating relative efficiency across regions. Later, they became useful for studying hospital-level variation in chronic illness care, and for measuring utilization and patient experiences at the very end of life. Altogether, end-of-life cohorts have been invaluable for clarifying the contribution of health system and provider factors to health care variation and outcomes.