Fig. 2.
Overall joint effect of the PM2.5 metal mixture for 0–7 day moving averages with QTc interval length estimated by Bayesian Kernel Machine Regression (BKMR). This figure compares the estimated change in QTc interval length when all predictors are at a certain quantile with the value when all of them are at their 50th percentile. BKMR models were adjusted for PM2.5 mass, daily ozone, age (years), race, maximum years of education, BMI (kg/m2), total cholesterol (mg/dL), mean arterial pressure (mmHg), diabetic status, use of beta blocker medication, alcohol intake (2 or more drinks per day or less than 2 drinks per day as reference), smoking status (current, former or never as reference), census tract percent of population age 25 years or older with less than a high school diploma, air temperature (°C), relative humidity (%) and seasonality (sine and cosine).