Resurrecting treatment histories of dead patients: a study design that should be laid to rest.

In this article we address whether studies of care rendered to patients prior to their death ("studies of decedents") produce an accurate portrait of care provided to patients who are dying. Studies of decedents typically analyze the care provided to patients over a defined interval antecedent to death. Studies of dying patients analyze care provided to patients subsequent to the time that their terminal status is perceived. We address whether 2 fundamental differences between studies of decedents and studies of the dying--the ways that subjects are identified and the time periods that are examined--lead to differences in interpretation of study results. Using examples from population-based cohorts of individuals with cancer, we show that both the differences in subject selection and time period introduce very substantial biases into studies of decedents. We conclude that studying care received prior to death can lead to invalid conclusions about the quality or type of care provided to dying patients.