Friday, July 23, 2004
This presentation is part of : Multisite Research: Building Community Among Diverse Members of a Research Team
Application of Survival Analysis to Examine Outcomes in a Multi-Site Rural Intervention Research Study
Susan Wilhelm, RN, PhD, UNMC College of Nursing, UNMC College of Nursing, University of Nebraska, Scottsbluff, NE, USA

Purpose: The purpose of this presentation is to illustrate the use of survival analysis as a strategy to address problems associated with missing (censored) data.

Definition of Concept: Survival analysis originated from demographic and medical research that investigated time from treatment until death. However, it is applicable to areas other than mortality. For example, survival analysis has been used in studies to determine time taken to exercise to maximum tolerance.

Theory: Markovís theory, which formed the basis of survival analysis, is concerned with events that are conditional upon those that precede them and affect those that come afterwards. This concept was applied to the analysis of data in the social sciences. Graphical representations of survival (or failure) rate as a function of time provide richer descriptive information, such as identification of times of highest risk, than do simple analyses of mean or median times to occurrence. Differences between groups in terms of their survival functions can also be tested. Additional techniques are available for building and testing models to predict survival from other variables. All methods accommodate censored data.

Application: Several approaches to survivor analysis were illustrated using data from an intervention study designed to increase the number of days of breastfeeding for first-time mothers. Approximately 17% of the datasetís 73 cases were censored. Life tables, graphical presentations of survival functions with nonparametric group comparisons, and Cox regression were employed.

Conclusions: Maximum data were available for analysis using this statistical analysis technique. Modest-size studies in which the dependent variable is time to occurrence of events may benefit from the use of this technique.

Back to Multisite Research: Building Community Among Diverse Members of a Research Team
Back to 15th International Nursing Research Congress
Sigma Theta Tau International
July 22-24, 2004