Using Logistic Regression of Sparse Data to Predict Influenza

• Activity Number: 83
• Type: Contributed
• Date/Time: Sunday, August 1, 2010 : 4:00 PM to 5:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Hector Lemus*+
• Companies: San Diego State University
• Keywords: conditional logistic ; influenza ; sparse data ; Bayesian estimate

Abstract

The influenza virus has long been known to cause a lower respiratory disease that is clinically indistinguishable from many other respiratory infections. Very few factors have been found to be strongly associated with influenza infection. Antivirals can reduce morbidity and mortality if they are administered early in the disease. Given the mediocre performance of most rapid point-of-care influenza diagnostics, clinicians have few tools with which to make diagnostic and therapeutic decisions. If certain symptoms can be modeled to strongly predict the presence or absence of flu, clinicians' management of febrile respiratory illness would be enhanced. This paper develops a model to correct the bias of sparse data in logistic regression and predict the presence of influenza.