A new study has used an artificial intelligence method to help predict factors that influence the lead level in the blood of children living in Broken Hill.
The Macquarie University study analysed its previous research, which looked at lead levels in blood collected from 1991 to 2015, and found that the year of testing had the greatest influence on levels.
It also found that children had higher levels if they lived within a kilometre of the city's central mine area or 1.37km of the railroad, and that levels increased faster in Aboriginal children than in non-Aboriginal children at nine-to-10 months of age and 12-to-18 months of age.
Professor Mark Taylor said researchers used a "machine-learning model," which was part of the artificial intelligence branch, for the study.
"It has no prior assumptions. You basically throw all the variables into the model and it spits out the strongest associations between blood lead and all the different variables that you assessed," Professor Taylor said.
Study 'first of its kind'
Lead researcher Xiaochi Liu said he believed the study was the first of its kind.
He said the kind of interpretable machine learning techniques the team used in the study had only been around for several years.
Mr Liu said the study's methods could also be applied elsewhere.
"The key thing here is the techniques we used in this study … can achieve both prediction accuracy and model interpretability," he said.
"It can identify previously unconnected variables associated with the childhood blood lead levels.
Levels have 'fallen over time'
Professor Taylor said the year of testing had been a particularly strong statistic to emerge from the research.
"That tells us the interventions over time and the increased knowledge of the risks over time have worked," he said.
Professor Taylor said the study could also assist with the work the Broken Hill Environmental Lead Program did.
"This provides extra evidence, or ammunition as it were, to the Broken Hill Environmental Lead Program where it might allow them to think a bit more about who do we target [and] what do we target in terms of minimising exposure," he said.