A Simulation to Evaluate the Impact of Design on Model-Based Methods for National Health and Nutrition Examination Survey (NHANES) Data Linked with Environmental Exposures

• Activity Number: 191
• Type: Contributed
• Date/Time: Monday, August 2, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Jennifer Parker*+ and Lester Randolph Curtin and Van Parsons
• Companies: National Center for Health Statistics and CDC and National Center for Health Statistics
• Keywords: NHANES ; sample survey ; air pollution ; survey weights

Abstract

To simulate the impact of the NHANES sample design, we create a finite population defined by combining 9 years of the National Health Interview Survey linked to (correlated, often missing) air quality data. Based on the NHANES sample design, we will draw samples of 5,000 Households from this finite population of approximately 360,000 households and evaluate different analytic methods for linked data files, with attention to the use of the standard survey weights and multi-level models. With the simulated samples we will apply design-based and model-based methods to estimate the relationship between health and air pollution, with and without control for standard and, when appropriate, scaled survey weights. We will assess the impact of correlated and missing exposure data, and of variation in survey design (number, size of PSUs) on inferences to provide guidance to users of linked files.