The goal of this project is to develop a series of models which describe the distribution of childhood blood-lead concentrations as a function of demographic data available through the 1990 U.S. Census. The analysis will be based on blood-lead screening data observed in four states (Rhode Island, New York, Illinois and Wisconsin) from 1994 through 1998. Blood-lead concentrations of children in each census tract will be categorized into one of the following discrete blood-lead groups to form a multinomial distribution:
|
Group |
Blood Lead Levels ( m g/dL) |
|
1 2 3 4 5 6 7 |
0 to 4 5 to 9 10 to 14 15 to 19 20 to 24 25 to 44 44 + |
A polytomous logistic regression model will then be used to predict blood-lead category as a function of Census information. The presentation will show participants how to use this model to (1) develop risk indices (an objective basis for prioritizing resources), (2) project the number of lead poisoned children expected under different screening scenarios, and (3) identify geographic areas with low current screening penetration that likely require additional screening resources.
Learning Objectives: The talk is designed to show lead poisoning prevention programs how to link childhood blood-lead screening data to information available in the 1990 US Census, via geocoding. The participant will then learn how to develop a predictive model, and how to use this model to (1) develop risk indices (an objective basis for prioritizing resources), (2) project the number of lead poisoned children expected under different screening scenarios, and (3) identify geographic areas with low current screening penetration that likely require additional screening resources
Keywords: Statistics, Lead
Presenting author's disclosure statement:
Organization/institution whose products or services will be discussed: None
I do not have any significant financial interest/arrangement or affiliation with any organization/institution whose products or services are being discussed in this session.