EXPOSURE MISCLASSIFICATION BIAS IN ENVIRONMENTAL EPIDEMIOLOGY: WITHIN-HOUSEHOLD RESPONSE CONCORDANCE ON HOUSEHOLD CHARACTERISTICS IN THE 2008-2010 SURVEY OF THE HEALTH OF WISCONSIN (SHOW)

Le résumé fourni par la source

Background and Aims: Environmental exposure assessment often relies on survey data which can lend itself to misclassification and recall bias that can lead to null associations and unreliable exposure estimates. Large-scale studies such as NHANES randomly choose one participant from each household to obtain information about characteristics of that home. It is assumed that the answers given to basic household characteristic affecting health outcomes would be the same for all residents from that household. The goal of this study was to test the assumption to estimate the extent this leads to discrepant exposure and risk estimates.. Methods: In the SHOW study, the same household characteristic questionnaire was administered to multiple respondents within each household, allowing for analysis of variations in response patterns. The SHOW study consists of a series of annual surveys gathering health-related data on a representative sample of Wisconsin residents age 21-74. A total of 1,043 households have been enrolled in SHOW; of these, 494 households had two or more respondents. Results: For this analysis, we focused on the following household characteristics (percent discrepant between respondents within household): report of peeling paint outside the house (24.6%), recent indoor pesticide use (21.6%), past radon testing (21.4%), black grout or caulk (19.4%), use of water filtration system (19.2%), mildew or mold odors (19.1%), outdoor weed killer/insecticide use (18.0%), water stains on walls or ceilings (17.4%), household smoking policy (16.1%), peeling paint in the home (13.4%), and source of water supply (4.0%). Conclusions: Preliminary analysis indicates that agreement within households may be associated with concordance of individual education (p-value=0.03) and income levels (p-value=0.04) among respondents. For studies designed to look at household characteristics, these results indicate the need to rely on more than one respondent or the potential benefit of external verification of responses to improve exposure estimates.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.