The effect of accessing medical records by date of death on estimates of survival.

An inefficient, but quickly and easily calculated, estimate of a survival-time distribution is described. Conditions for the estimate to be unbiased specify bounds on the length of time a study has been in progress and the way in which people must enter a study. These appear to be frequently met when a large medical registry is examined with a view to establishing base line survival experience for a selected class of patients.