Estimation of Odds Ratio with Rare Event Case: A Simulation Study Based on Hospitalized Flu Vaccine Study

• Activity Number: 592
• Type: Contributed
• Date/Time: Wednesday, August 4, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Qingxia Chen+ and Yuwei Zhu* and Helen K.B. Talbot and Marie R. Griffin and Kathryn M. Edwards
• Companies: Vanderbilt University and Vanderbilt University and Vanderbilt University and Vanderbilt University and Vanderbilt University
• Keywords: penalized regression ; LASSO ; propensity score ; vaccine effectiveness

Abstract

Influenza vaccine effectiveness can be estimated by (1-odds ratio)*100% using patients hospitalized with confirmed influenza as cases and similar hospitalized patients who are influenza negative as controls. Challenges exist when estimating VE for individual years due to the fact that there may not be sufficient sample size (influenza events or casese) to control for the long list of potential confounders. We conducted a series of simulations based on the hospitalized flu vaccine study to compare the performance of different approaches, including conditional logistic regression with propensity score matching, logistical regression with propensity score adjustment, logistic regression based on penalized maximum likelihood, lasso logistic regression, and logistic regression without penalization, in the estimation of vaccine effectiveness under different complex and realistic settings.