PDF-ikonPrintikon

Kim Mark Knudsen


Knudsen 1994
Knudsen, Kim Mark. 1994. Analysis of Event History Data in Measles and Diarrhea Epidemiology. Bandim Health Project, Statens Serum Institut. Biostatistical Department, University of Copenhagen. 93 p.
  This dissertation reflects research within the field of statistical analysis of event history data from several studies on measles and diarrhea in West Africa. The studies that are presented are all long-term cohort studies, and developments within the past 20 years of statistical methods for analysis of data concerning sequences of life events have permitted a greater utilization of -such data. The focus is on the use of statistical methods in infectious disease epidemiology on evaluation of risk factors and interventions related to the occurrence of measles related deaths and diarrhea. The study was performed at the Statistical Research Unit, University of Copenhagen. The dissertation contains three chapters, each being self-contained.

The first chapter on the analysis of measles cohort data first presents three smaller studies. The studies were observational, not randomized trials, and the hypotheses examined were not predefined in the study protocols. Emphasis is on the use of nonstandard survival analysis techniques. The intention is to show how different types of measles cohort data can be analyzed in an efficient manner to examine hypotheses concerning the effect of cross-sex transmission and early exposure of measles on survival.

A description and evaluation of a number of large randomized trials of high titer measles vaccines in Guinea-Bissau and Senegal follow. High titer vaccines were seen as a potential improvement in areas where measles is a major health problem. A combined analysis of the results from the West African trials on the introduction of high titer measles vaccines is performed to use the total amount of data collected. Combing trials of high titer measles vaccines showed higher mortality among the high titer group compared to the standard low titer group. Mortality of recipients of medium titer vaccine was not different from that of the standard group. The trials were not specifically designed to study long term mortality, but adjustments for several possible sources of bias did not alter the results. The conclusions of the combined analysis of the West African data had a significant influence on the World Health Organization’s decision to withdraw their recommendation of the use of high titer measles vaccines.

The concept of vaccine efficacy, the beneficial effect of a vaccination, is discussed in the second chapter. Simple transmission models with one rate of disease infection among the unvaccinated and two rates of infection among the vaccinated are discussed. A simulated example illustrates the size of uncertainty related with the different approaches. A random effect model is proposed that includes the vaccine efficacy parameter explicitly as an alternative to published models. Analyses on the simulated data indicate that exceptionally large studies are required to obtain sufficient power to detect heterogeneity in vaccine response.

The analysis of recurrent disease events is the topic of the last chapter and the methods are applied on a study of diarrhea from Guinea-Bissau. Diarrhea is a major cause of malnutrition and early childhood death in developing countries. In most areas of Guinea-Bissau, diarrhea is probably the single largest cause of childhood death. To propose measures leading to a decline in mortality and morbidity, a large-scale registration of childhood diarrhea in the Bandim district of Bissau has been carried out since 1987. The objective of the study was to determine diarrheal incidence among children below three years of age, to identify risk factors and to propose measures leading to a decline in diarrheal diseases, and thus also in child mortality. Furthermore, the study aimed at assessing the association between diarrhea and feeding practices.

For non-fatal events such as diarrhea, subjects may experience several episodes and thus contribute with more than one period of observation. To carry out ordinary survival analysis independence of periods must be assumed, including periods from the same child. Random effect models applied in the community study of childhood diarrhea to "extend" survival analysis to cope with dependency between a child’s periods. Results showed that the random effects model fitted the data significantly better than a standard Poisson model and identified child’s sex, family’s keeping of domestic animals, mother’s education and age of weaning as risk factors. A manifest dependence of time since the last episode of diarrhea was also found with an increased risk of renewed illness for several weeks following recovery.

 

 

 

Last revised 6 October 2015