Identifying the Best Questions for Rapid Screening of Secondhand Smoke Exposure Among Children

Abstract

Abstract Introduction Many children suffer from secondhand smoke exposure (SHSe), which leads to a variety of negative health consequences. However, there is no consensus on how clinicians can best query parents for possible SHSe among children. We employed a data-driven approach to create an efficient screening tool for clinicians to quickly and correctly identify children at risk for SHSe. Methods Survey data from mothers and biospecimens from children were ascertained from the Neurodevelopment and Improving Children’s Health following Environmental Tobacco Smoke Exposure (NICHES) study. Included were mothers and their children whose saliva were assayed for cotinine (n = 351 pairs, mean child age = 5.6 years). Elastic net regression predicting SHSe, as indicated from cotinine concentration, was conducted on available smoking-related questions and cross-validated with 2015–2016 National Health and Nutrition Examination Survey (NHANES) data to select the most predictive items of SHSe among children (n = 1670, mean child age = 8.4 years). Results Answering positively to at least one of the two final items (“During the past 30 days, did you smoke cigarettes at all?” and “Has anyone, including yourself, smoked tobacco in your home in the past 7 days?”) showed area under the curve = .82, and good specificity (.88) and sensitivity (.74). These results were validated with similar items in the nationally representative NHANES sample, area under the curve = .82, specificity = .78, and sensitivity = .77. Conclusions Our data-driven approach identified and validated two items that may be useful as a screening tool for a speedy and accurate assessment of SHSe among children. Implications The current study used a rigorous data-driven approach to identify questions that could reliably predict SHSe among children. Using saliva cotinine concentration levels as a gold standard for determining SHSe, our analysis employing elastic net regression identified two questions that served as good classifier for distinguishing children who might be at risk for SHSe. The two items that we validated in the current study can be readily used by clinicians, such as pediatricians, as part of screening procedures to quickly identify whether children might be at risk for SHSe.