Evaluating Vaccine Safety Data Using Markov Chains

Loading Map....

Date/Time
Date(s) - Sep 13, 2010
2:00 PM - 3:00 PM

Location
UCLA Center for Community Health

Category(ies)


Presented by: 

Martin L. Lee, PhD, Cstat, Csci

UCLA Department of Biostatistics
Abstract: 
The analysis of vaccine safety data from population cohorts poses a number of issues. These include the rarity of adverse outcomes, non-experimental data, and a complex correlation structure. Different approaches have been suggested including case-control and case-cohort designs. Others have used the “risk-interval” method, whereby vaccinees act as their own control and intervals before and/or after vaccination are compared with an at-risk period immediately post-vaccination. We present an alternative approach which models the probability of events within these periods using a first-order Markov chain. This model is simple and provides chi-square tests of the relationship between vaccination and events of interest. It allows for determination of whether the Markov process is stationary and first order. We have applied these methods to a population-based dataset evaluating the relationship of gender to the incidence of adverse events following vaccination. We have demonstrated that our data fit this probability model and also determined associations that have been shown with more complex models using similar data. This model can be used for the evaluation of vaccine safety.
Biography: 
Dr. Lee is the Senior Biostatistician for the Sepulveda Center of Excellence, and Adjunct Professor of Biostatistics at the UCLA School of Public Health as well as Adjunct Professor of Internal Medicine at Charles R. Drew University of Medicine and Science. He has more than 15 years experience in designing studies and analyzing data from VA and non-VA variations studies and multi-site organizational interventions, with special expertise in performance measures relevant to quality improvement activities, in addition to 30 years experience in clinical trial design, conduct, and analysis for the pharmaceutical and biotechnology industries. He has authored or coauthored more than 200 scientific papers and is the co-author of two textbooks on medicine and statistics.