The Effect of Missing Industry and Occupation Codes on the Assessment of Health Outcomes in the 2016 Behavioral Risk Factor Surveillance System (BRFSS) Survey

• Activity Number: 358 - Contributed Poster Presentations: Section on Statistics in Epidemiology
• Type: Contributed
• Date/Time: Tuesday, July 30, 2019 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Jia Li* and Matthew Groenewold and Sara E. Luckhaupt and Marie H. Sweeney and James M. Boiano
• Companies: NIOSH and NIOSH and NIOSH and NIOSH and NIOSH
• Keywords: missing data; bias; sample weight adjustment

Abstract

Increasingly, states have used information on industry and occupation (I/O) of employment in the annual Behavioral Risk Factor Surveillance System (BRFSS) telephone survey to evaluate worker health. However, a relatively high proportion (13%) of I/O codes are missing in the BRFSS data, and not completely at random. In this study, we examined the potential bias caused by nonresponse and coverage errors due to missing I/O in the 2016 BRFSS data. We used non-response adjustment and raking methods to adjust the BRFSS sample weight for employed respondents so that the marginal distribution of a set of auxiliary variables matched those for this subpopulation according to the American Community Survey. We estimated the prevalence of six selected health behaviors and health outcomes by occupation with both the original and adjusted weights. No statistically significant difference was observed between the estimates. The results suggest that the proportion of missing I/O in the 2016 BRFSS data does not introduce significant bias. Nevertheless, future efforts should be made to increase response rates in I/O in the BRFSS data.