Bayesian Methodology Applied on Blood Lead Data for Children

• Activity Number: 621 - Beyond Linear Regression: Nonlinear Association, Quantile Regression and Generalized Linear Models
• Type: Contributed
• Date/Time: Thursday, August 1, 2019 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Shailendra Banerjee* and Yu Sun
• Companies: Centers for Disease Control and Georgia Department of Public Health
• Keywords: Bayesian Methodology; Blood Lead Level; Credible Set

Abstract

Bayesian methodology is applied on blood lead level (BLL) data for children in the State of Georgia to estimate the number of children with BLLS 5-9 µg/dL. Assuming that the data for BLLs follow a Poisson distribution and the Poisson parameter follows a gamma distribution, it is possible to estimate the number of children with BLLs 5-9 µg/dL in a county of a public health district of Georgia, based on the data in surrounding counties. The outcome investigated in this study is BLL ? 5 µg/dL, which is the CDC blood lead reference level in children. We used Georgia Department of Public Health (DPH) blood lead surveillance data for children aged < 72 months who were tested for lead at least once during 2015. Based on 2015 BLL data for children in 13 of 18 health districts in Georgia DPH, we estimated the number of children with BLLs 5-9 µg/dL and its credible set in a targeted county in each of these health districts. In the Northwest public health district, it was estimated that the number of children < 72 months of age with BLLs 5-9 µg/dL in targeted county A was 15 with 95% credible set (7, 25) as compared to actual value 5. Other models and priors will be further tested.