Are there joint effects of different air pollutants and meteorological factors on mental disorders? A machine learning approach.

Abstract

Exposure to air pollutants is considered to be associated with mental disorders (MD). Few studies have addressed joint effect of multiple air pollutants and meteorological factors on admissions of MD. We examined the association between multiple air pollutants (PM2.5, PM10, O3, SO2, and NO2), meteorological factors (temperature, precipitation, relative humidity, and sunshine time), and MD risk in Yancheng, China. Associations were estimated by a generalized linear regression model (GLM) adjusting for time trend, day of the week, and patients' average age. Empirical weights of environmental exposures were judged by a weighted quantile sum (WQS) model. A machine learning approach, Bayesian kernel machine regression (BKMR), was used to assess the overall effect of mixed exposures. We calculated excess risk (ER) and 95% confidence interval (CI) for each exposure. According to the effect of temperature on MD, we divided the exposure of all factors into different temperature groups. In the high temperature group, GLM found that for every 10μg/m3 increase in O3, PM2.5 and PM10 exposure, the ERs were 1.926 (95%CI 0.345, 3.531), 1.038 (95%CI 0.024, 2.062), and 0.780 (95% CI 0.052, 1.512) after adjusting for covariates. Temperature, relative humidity, and sunshine time also reported significant results. The WQS identified O3 and temperature (above the threshold) had the highest weights among air pollutants and meteorological factors. BKMR found a significant positive association between mixed exposure and MD risks. In the low temperature group, only O3 and temperature (below the threshold) showed significant results. These findings provide policymakers and practitioners with important scientific evidence for possible interventions. The association between different exposures and MD risk warrants further study.