|
Abstract:
|
Analyses of spatio-temporal exposures with health outcomes have historically been limited to aggregated outcome data with crude adjustment for person-level information (e.g., stratification). The availability of longitudinal patient electronic health records (EHR) data makes detailed person-level data readily accessible and linkable to spatio-temporal exposures. Mixed spatio-temporal data contains data that vary only temporally (e.g., temperature), data that varies by observation (e.g., sex or race), and data that varies both temporally and spatially (e.g., tree cover). While EHR and environmental exposure data can be efficiently collected, both data sources have a number of analytic concerns that must be addressed (e.g., neither data source is primarily collected or organized for research purposes). Furthermore, combining these data in inferential models presents additional challenges. We will detail strategies for addressing data concerns as well as different statistical approaches to combine mixed spatio-temporal data in inferential models. We will provide an example of a study linking EHR data on pediatric asthma exacerbation to ambient air quality exposures in Durham, NC.
|
