Improved Decision Making for Water Lead Testing in U.S. Child Care Facilities Using Machine-Learned Bayesian Networks

Area of Science:

• Environmental Health
• Public Health
• Data Science

Background:

• Existing tap water lead testing programs in U.S. child care facilities require enhanced methods for identifying high-risk locations.
• Limited resources necessitate optimized strategies for prioritizing facilities for lead testing.

Purpose of the Study:

• To develop and evaluate machine-learned Bayesian network (BN) models for predicting building-wide water lead risk in North Carolina child care facilities.
• To compare the performance of BN models against common heuristics for identifying high-risk facilities for lead testing.

Main Methods:

• Bayesian network (BN) models were developed to predict water lead risk in over 4,000 child care facilities using data from 22,943 taps.
• Model performance was assessed by comparing BN predictions to established risk factors like building age, water source, and Head Start status.
• The study analyzed variables associated with building-wide water lead, including facility demographics, water source, and number of taps.

Main Results:

• BN models identified key risk factors: facilities serving low-income families, using groundwater, and having more taps showed higher lead risk.
• Models predicting single-tap lead exceedance performed better than those predicting clustered high-risk taps.
• BN models demonstrated superior performance (118-213% higher Fβ-scores) compared to alternative heuristics, potentially increasing high-risk facility identification by 60% and reducing sampling needs by 49%.

Conclusions:

• Machine-learning approaches, specifically BN models, offer a valuable tool for accurately identifying high water lead risk in child care facilities.
• Implementing BN model-informed sampling can significantly improve the efficiency and effectiveness of national lead testing programs.
• This study highlights the potential of advanced data analysis to enhance public health protection strategies for lead exposure.