Design of an Observational Study to Evaluate the Role of HIV in Early-Onset Disease in Newborns

• Activity Number: 140
• Type: Contributed
• Date/Time: Monday, August 1, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Elizabeth Zell*+
• Companies: Centers for Disease Control and Prevention
• Keywords: Propensity Score Methods ; Observational Data ; Conditional Associations

Abstract

Frequently in epidemiology, it is important to identify risk factors associated with a specific disease, often from observational studies. Propensity score methods are commonly used in the design of observational studies for causal inference. However, there is also a role for these methods to assess more interesting conditional associations. For example, the question was asked, "Is HIV exposure a risk factor for developing early-onset disease?" We applied propensity score matching to determine if infants born to HIV positive mothers are at increased risk of developing early-onset disease than infants born to HIV negative mothers. For each infant born to an HIV positive mother, we identified an infant born to an HIV negative mother by matching on known risk factors for developing early-onset disease. Propensity score matching allowed us to select an appropriate referent group similar on known or suspected risk factors while blinded to the outcome. The strength of our study was our ability to utilize propensity score methods to better evaluate associations between our comparison groups. Here we illustrate why the matched subset gives the more realistic comparisons between the groups.