Introduction to survival analysis

Survival analysis can be defined as the statistical processing of survival data. It allows one to examine the time interval from a given starting point (e.g., diagnosis of disease) to a discrete outcome (e.g., death) for a specified group. If this group is a random sample of the population, then the survival experience of the group will reflect that of the general population. It is the goal of this chapter to provide an overview of survival analysis as it relates to the major disorders within neurology. Death will be the predominant outcome measure in discussions. The field of survival analysis has had many recent advances and for a more technical discussion, the reader is referred to more exhaustive sources (1–4).

The ideal survival study would follow all members of a cohort from a specified time to a predetermined outcome. Unfortunately, the reality of assembling a survival cohort must allow for different outcomes. This is illustrated in Figure 8.1 where seven subjects were followed for variable lengths of time after an acute ischemic stroke. The outcome in this example was death. The subjects were recruited during a 2-year period and followed for 4 years. Subjects four, five, six, and seven reached the endpoint in the study prior to the end of the observation period. Subjects one, two, and three were censored. Censoring occurred in subjects two and three because they reached the end of the study before dying. This is called right censoring.