An Association Study Including Measurement Errors Using Linked NHIS and EPA Modeled Data

• Activity Number: 179
• Type: Contributed
• Date/Time: Monday, August 1, 2016 : 10:30 AM to 12:20 PM
• Sponsor: Survey Research Methods Section
• Author(s): Rong Wei* and Van Parsons and Jennifer Parker and Ambarish Vaidyanathan and Pavlina Rumcheva
• Companies: CDC/NCHS and CDC/NCHS and CDC/NCHS and National Center for Environmental Health and CDC/NCHS
• Keywords: air pollutant ; health status ; downscaler air measures ; linked complex survey data ; measurement smoothed errors

Abstract

The CDC conducts the National Health Interview Survey (NHIS) and maintains a database of modeled air quality (AQ) concentrations, which are generated from a model developed by Environmental Protection Agency and its partners, linkable to NHIS data at the census tract level. A previous study by the authors used this linked dataset to model health status using complex survey design features, but under the assumption that the modeled AQ concentrations were deterministic. However, AQ concentrations were actually modeled using a Bayesian space-time framework, providing a prediction and its standard error, which in part is based on distances of census tract centroids to air monitors. As the geographical locations of households in the NHIS and closest air monitor vary greatly, associations of population health with predicted air qualities might be more accurately modeled if the standard errors of predictions are incorporated into the health analysis model. This study focuses on such methodologies in estimating air quality effects on population health. Results from these methods will be compared and discussed.