Bayesian hierarchical modeling of cardiac response to particulate matter exposure

Studies have linked increased levels of particulate air pollution to decreased autonomic control, as measured by heart rate variability (HRV), particularly in susceptible populations such as the elderly. In this study, we use data obtained from the 1998 USEPA epidemiology-exposure longitudinal panel study of elderly adults in a Baltimore retirement home to examine the relationship between HRV and PM₂.₅ personal exposure. We consider PM₂.₅ personal exposure in the aggregate and personal exposure to the components of PM₂.₅ as estimated in two ways using receptor models. We develop a Bayesian hierarchical model for HRV as a function of personal exposure to PM₂.₅, which integrates HRV measurements and data obtained from personal, indoor and outdoor PM₂.₅ monitoring and meteorological data. We found a strong relationship between decreased HRV (HF, LF, r-MSSD and SDNN) and total personal exposure to PM₂.₅ at a lag of 1 day. Using personal exposure monitoring (PEM) apportionment results, we examined the relative importance of ambient and non-ambient personal PM₂.₅ exposure to HRV and found the effect of internal non-ambient sources of PM₂.₅ on HRV to be minimal. Using the PEM apportionment data, a consistent effect of soil at short time scales (lag 0) was found across all five HRV measures, and an effect of sulfate on HRV was seen for HF and r-MSSD at the moving average of lags 0 and 1 days. Modeling of ambient site apportionment data indicated effects of nitrate on HRV at lags of 1 day, and moving averages of days 0 and 1 and days 0-2 for all but the ratio LF/HF. Sulfate had an effect on HRV at a lag of 1 day for four HRV measures (HF, LF, r-MSSD, SDNN) and for LF/HF at a moving average of days 0-2.