Modeling age-adjusted mortality rate as a function of socioeconomic and environmental predictors. A map of mortality rate in U.S. counties is shown at center and seven classes of predictor variables are shown as concentric rings, with the strongest predictors closest to the center.

By Felisha Walls and Daniel McGarvey.

Read the full article here.

Human health research often focuses on local socioeconomic conditions, and rightly so. Financial stability, access to healthcare, food security – these factors have immediate and profound consequences for our daily lives. Less is known, however, about connections between human health and the environment. Epidemiology and toxicology research have confirmed the harmful effects of specific disruptions and pollutants, but researchers are only now developing a broad understanding of direct and indirect connections between human health and the environment. This work must proceed in earnest because the environment is changing at an unprecedented rate.

            Our study uses a novel framework to identify potential links between human health and a variety of socioeconomic and environmental variables. Specifically, we combine national-scale data on human mortality rate with a set of 76 socioeconomic and environmental variables, then use a machine learning algorithm called “random forest” to predict county-level mortality rate and identify the strongest predictor variables. All the data we used in this study are publicly available from standardized, vetted sources such as the United States Centers for Disease Control and Prevention, and the U.S. Environmental Protection Agency.

            Relative to other published models, our model does an excellent job of predicting human mortality rate. It accounts for more than 75% of the observed variation in mortality rate, with consistently small prediction errors (deviations between predicted and observed mortality rates for individual counties). Socioeconomic variables, including smoking, food insecurity, and level of physical activity, are identified as the strongest individual predictors of mortality rate. But environmental variables are also shown to have important effects. Strong influences of increasing air temperature and precipitation are particularly noteworthy, as they reinforce the growing concern that climate change will threaten human well-being.

            We hope this study will inspire others to pursue integrative research on the diverse effects that socioeconomic and environmental influences have on human health. By combining large, mixed datasets with modern machine learning tools, we can begin to anticipate human responses to changing conditions and, hopefully, take proactive steps to mitigate the most damaging outcomes.