Extension of the Life Table to Repeating and Changing Events

Abstract

Estimation of cumulative and adjusted occurrence of life events and measures over time is often important in settings where study subjects have incomplete or different follow-up periods. Well-known methods to do this, such as the life table and age adjustment, exist for binary nonrecurrent events (i.e., death). However, a general approach to evaluate recurrent life events (e.g., repeated infections), life events with different durations (e.g., hospitalization days), costs, or changing life measures (e.g., body weight) is not available. This paper develops the "Life-Event Table," an analog of the life table that can analyze occurrence of diverse types of events and measures when the observation periods of subjects are incomplete and different. This method, based on a central limit theorem for incomplete multivariate data, obtains point estimates and variances for cumulative incidence, age-adjusted expectation, and other quantities. Simple hypotheses tests and comparisons are possible with this approach. Three applications are presented. The Life-Event Table may facilitate use of currently available robust multivariate analysis methods. Further research into other multivariate methods that take advantage of this Life-Event Table's specific covariance/variance structure may be warranted.