Analysis of longitudinal studies with death and drop-out: a case study

Stat Med. 2004 Jul 30;23(14):2215-26. doi: 10.1002/sim.1821.

Abstract

The analysis of longitudinal data has recently been an active area of biostatistical research. Two main approaches to analysis have emerged, the first concentrating on modelling evolution of marginal distributions of the main response variable of interest and the other on subject-specific trajectories. In epidemiology the analysis is usually complicated by missing data and by death of study participants. Motivated by a study of cognitive decline in the elderly, this paper argues that these two types of incomplete follow-up may need to be treated differently in the analysis, and proposes an extension to the marginal modelling approach. The problem of informative drop-out is also discussed. The methods are implemented in the 'Stata' statistical package.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Age Factors
  • Aged
  • Aged, 80 and over
  • Cognition Disorders / etiology
  • Cohort Studies
  • Female
  • Humans
  • Longitudinal Studies*
  • Male
  • Models, Biological
  • Models, Statistical*
  • Patient Dropouts*
  • Sex Factors