Using Artificial Intelligence to Identify Sources and Pathways of Lead Exposure in Children

Context: Sources and pathways of lead exposure in young children have not been analyzed using new artificial intelligence methods. Objective: To collect environmental, behavioral, and other data on sources and pathways in 17 rural homes to predict at-risk households and to compare urban and rural indicators of exposure. Design: Cross-sectional pilot study. Setting: Knox County, Illinois, which has a high rate of childhood lead poisoning. Participants: Rural families. Methods: Neural network and K-means statistical analysis. Main Outcome Measure: Children's blood lead level. Results: Lead paint on doors, lead dust, residential property assessed tax, and median interior paint lead level were the most important predictors of children's blood lead level. Conclusions: K-means analysis confirmed that settled house dust lead loadings, age of housing, concentration of lead in door paint, and geometric mean of interior lead paint samples were the most important predictors of lead in children's blood. However, assessed property tax also emerged as a new predictor. A sampling strategy that examines these variables can provide lead poisoning prevention professionals with an efficient and cost-effective means of identifying priority homes for lead remediation. The ability to preemptively target remediation efforts can help health, housing, and other agencies to remove lead hazards before children develop irreversible health effects and incur costs associated with lead in their blood.