Linkage of the US National Health Interview Survey to air monitoring data: An evaluation of different strategies

Authors

Jennifer D. Parker, National Center for Health Statistics, CDC, 3311 Toledo Road, Room 6107, Hyattsville, MD 20782, USA
Tracey Woodruff, US Environmental Protection Agency, University of California, San Francisco, USA
Lara J. Akinbami, National Center for Health Statistics, CDC, 3311 Toledo Road, Room 6107, Hyattsville, MD 20782, USA
Natalyia Kravets, NOVA Research Company, Northrop Grumman CITS II Contract, USA

Document Type

Article

Date of this Version

2008

Comments

Published in Environmental Research 106 (2008) 384–392.

Abstract

The goal of this study is to describe linkages between the National Health Interview Survey (NHIS) and Environmental Protection Agency (EPA) air monitoring data, specifically how the linkage method affects characteristics and exposure estimates of study samples and estimated associations between exposure and health. In the USA, nationally representative health data are collected in the NHIS and annual air quality data are collected by the EPA. The linkage of these data for research is not straightforward and the choices made may introduce bias into results. The 2000–2003 NHIS and air quality data for six air pollutants were linked by residential block group and monitor location, which differ by pollutants. For each pollutant, three annual exposure variables were assigned to respondents: (1) average of all monitors in the county, (2) of monitors within a 5-mile radius of the distance between block group and monitor, and (3) within a 20-mile radius. Exposure estimates, study sample characteristics, and association between fine particle exposure and respondent-reported health status were compared for different geographic linkage methods. The results showed that study sample characteristics varied by geographic linkage method and pollutant linked. Generally, the fewer the NHIS respondents linked, the higher is the pollution exposure and lower is the percentage of non-Hispanic whites. After adjustment for sociodemographic and geographic factors, associations between fine particles and health status were generally comparable across study samples. Because exposure information is not available for all potential participants in an epidemiological study, selection effects should be considered when drawing inferences about air quality–health associations. With the current monitoring data system, the study sample is substantially reduced when linkage to multiple pollutants is performed.