STUDYING THE ASSOCIATION OF ASTHMA HOSPITAL VISITS WITH PARTICULATE MATTER EXPOSURES USING SATELLITE-DERIVED NEW DAILY PM2.5 DATA PRODUCTS

PM2.5 presents serious health concerns as these tiny particles can penetrate deeply into the lungs and other body organs, thereby increasing the risk for many health conditions including respiratory illnesses. Spatio-Temporal studies of PM 2.5 concentration over a long period of time can help establish benchmarks for environmental policies. Moreover, studying associations with health outcomes helps develop prophylactic measures for patient care. However, such studies require reliable data for both PM2.5 and health outcome. Various networks of ground-based sensors routinely measure PM2.5 but with sparse spatial coverage. Attempts have been made to overcome this limitation utilizing satellite-derived Aerosol Optical Depth (AOD) to estimate ground-level PM2.5 (GLP). Estimating GLP utilizing atmospheric AOD is difficult due to: 1) vertical variations within the atmospheric column; and 2) the nonlinear relationship between GLP, AOD, humidity, pressure, wind speed and land surface type.

In this project, a new method to estimate GLP based on satellite observations, meteorological analysis and machine learning procedure is applied for the period of 2000 to 2013. For health outcome data, state-wide asthma hospital patient data for Mississippi from 2003 to 2011 are utilized. Initial resolutions of are: 1) GLP on a 0.1°x0.1° grid level, 2) asthma hospital visits geocoded to patients’ addresses, 3) demographic and socioeconomic status (SES) data at the Census Block Group level, and 4) seasonal attributes of temperature and relative humidity from NOAA ground locations. All four datasets are resolved to a common 0.1°x0.1° grid to support spatio-temporal analysis of associations between GLP and asthma hospital visits in Mississippi.

The association between hospital visits and the estimated GLP is modeled using a mixed effect model that accounts for spatial autocorrelation of residual error terms while controlling for age, race, admission type, SES and seasonal impact on hospital admission. This model supports inference concerning the magnitude and significance of this association on observed patterns in health outcomes and also allows prediction of changes in the expected number of hospital visitations associated with increased or decreased levels of PM2.5 exposure