Modeling Elevated Children's Blood Lead Levels across Five U.S. States: A Statistical Approach to Improve Predictions...

Area of Science:

• Environmental Health
• Biostatistics
• Geospatial Analysis

Background:

• Children's elevated blood lead levels (EBLL) pose significant public health risks.
• Existing predictive models for EBLL lack interpretability and real-world applicability, hindering effective health interventions.
• Accurate prediction of EBLL is crucial for targeted prevention and mitigation strategies.

Purpose of the Study:

• To develop and evaluate interpretable and applicable statistical models for predicting census tracts with elevated blood lead levels in children.
• To identify key demographic and housing factors associated with EBLL across different U.S. states.
• To enhance predictive accuracy by incorporating spatial autocorrelation into modeling approaches.

Main Methods:

• Utilized publicly available population data and EBLL rates from five U.S. states (Michigan, Ohio, California, Minnesota, Wisconsin) at the census tract level.
• Developed and compared spatial and nonspatial random forest and logistic regression models to predict EBLL risk.
• Employed a 10-fold cross-validation approach to evaluate model performance, focusing on spatial logistic regression.

Main Results:

• Spatial logistic regression models demonstrated superior predictive performance, achieving 82% to 90% accuracy.
• The percentage of homes built before 1940 was a consistent and significant predictor (p < 0.05) across models, increasing EBLL odds by 1-7%.
• Accounting for spatial autocorrelation improved prediction by capturing local clustering effects of EBLL.

Conclusions:

• Spatial statistical modeling offers enhanced interpretability and applicability for predicting children's EBLL.
• The developed models can effectively inform public health interventions and guide lead poisoning prevention efforts, especially in areas with sparse EBLL data.
• This geospatial approach provides a valuable tool for identifying at-risk communities and allocating resources for environmental health initiatives.