Time to Event Models
In many health sciences applications, binary outcomes are incompletely observed. For example, if we are studying whether cancer patients experience a relapse after a initial remission, we may may not be able to follow patients to the end of their lives; instead, we may only know whether each patient has relapsed before the end of the study. If a patient has not relapsed by that point, we might not know if they will relapse at some other date or if they will stay cancer-free for the rest of their lives. 1 Their recurrence status at end-of-life is missing data. If some study participants withdraw from a study before the end date in the study design, there will be even more missing data. All of this missing data will make logistic regression difficult for this type of data.
However, these outcome observations are not entirely missing. We know that those patients stayed relapse free at least until the time point when we last saw them. If we also know the time-to-event for the participants who did experience events while under study, we can analyze time-to-event-or-study-exit, combined with the indicator of which of these two cases occurred, using survival analysis. The survival analysis framework is the subject of the rest of these course notes.
Binary outcomes are typically defined for a specific time-point. It is important to clearly define whether we are interested in outcome status at end of study, at end of life, or at some other time.↩︎