Air pollution mixtures and stillbirth: k-means cluster analysis

• Four air pollution mixtures were identified using k-means cluster analysis. • Patients associated with the mixture highest in PM 2.5 and SO 2 had an increased odds in stillbirth. • Individual dose-response relationships to air pollutants differed from mixtures. Preventing stillbirths starts with identifying modifiable risk factors, such as air pollution (AP). Most research has focused on individual components of AP (e.g. PM 2.5 ) to assess the linear relationship between air pollutant concentration and stillbirth, with mixed results. However, the air we breathe is a mixture of gases and particles. The purpose of this study was to assess how mixtures of criteria air pollutants are associated with stillbirth in the United States. Air pollution data from the US Environmental Protection Agency and individual US birth and fetal death data from the CDC National Vital Statistics System were used. Air pollutant exposures were clustered using k-means clustering. A categorical variable linking each patient to one of these clusters was used in a regression analysis. Four AP clusters were identified through k-means. Exposure to air pollution mixtures included in Clusters 1, 2, and 4 during pregnancy was associated with a 40%, 13%, and 43% increase in the odds of stillbirth, compared to Cluster 3 (reference), (OR = 1.40, 95% CI: 1.36-1.44, p<0.001; OR = 1.13, 95% CI: 1.09-1.16, p<0.001; OR = 1.43, 95% CI: 1.38-1.49, p<0.001), respectively. The air pollution cluster associated with the greatest odds of stillbirth was high in PM 2.5 and SO 2 , which may indicate industrial sources. To our knowledge, this is the first study to assess the impact of air pollution mixtures on stillbirth using k-means clustering. This method can be used to identify non-linear relationships between mixtures of air pollution and a health outcome.