Using Machine Learning to Differentiate Second- and Thirdhand Smoke Exposure

April 5, 2023
By Avery Crosley

This study used machine learning techniques to classify children into three different groups of reported tobacco exposure: no tobacco smoke exposure, thirdhand smoke exposure, and second- and thirdhand smoke exposure. Of the 4,485 nonsmoking 3–17 year-olds in the study, the machine learning model reported that 76% of children were classified as no tobacco smoke exposure, 16% were classified as thirdhand smoke exposed, and 8% were classified as mixed second- and thirdhand smoke exposed.

Compared to the true results from biomarkers taken from the children, the prediction model reported accuracies of 100% for no tobacco smoke exposure, 88% for thirdhand smoke exposed, and 71% for mixed second-and thirdhand smoke exposed. This new technique may help us better understand complicated exposure, such as when children are exposed to both second- and thirdhand smoke exposure.

Image: Merianos et. al. (2023). Distinguishing Exposure to Second- and Thirdhand Smoke among U.S. Children Using Machine Learning: NHANES 2013−2016. Environ. Sci. Technol., 57, 5, 2042–2053

Click here to read the research study.

More Must Read Stories

Recent Articles

pipe tobacco

Does smoking a pipe create thirdhand smoke?

Many people think that smoking a pipe is safer than smoking regular cigarettes, but pipe tobacco is also harmful to our health. Pipe smokers have an increased risk of cancers of the head, neck, liver, and lung.

collaborative consortium on thirdhand smoke research projects announcement

This series features the Consortium’s newly funded projects, which engage in groundbreaking research about the nature and health consequences of thirdhand smoke. Thirdhand smoke is the chemical residue that is left behind on clothes, skin, furniture, walls, and other surfaces after someone smokes.

Share This
Tweet This
Email This

Stay Informed

Get the latest thirdhand smoke news and research delivered straight to your inbox, or follow us on social media: