Real-Time Automated Hazard Detection Framework for Health Information Technology Systems.

An increase in the reliability of Health Information Technology (HIT) will facilitate institutional trust and credibility of the systems. In this paper, we present an end-to-end framework for improving the reliability and performance of HIT systems. Specifically, we describe the system model, present some of the methods that drive the model, and discuss an initial implementation of two of the proposed methods using data from the Veterans Affairs HIT and Corporate Data Warehouse systems. The contributions of this paper, thus, include (1) the design of a system model for monitoring and detecting hazards in HIT systems, (2) a data-driven approach for analysing the health care data warehouse, (3) analytical methods for characterising and analysing failures in HIT systems, and (4) a tool architecture for generating and reporting hazards in HIT systems. Our goal is to work towards an automated system that will help identify opportunities for improvements in HIT systems.

Augmenting Epidemiological Models with Point-Of-Care Diagnostics Data.

Although adoption of newer Point-of-Care (POC) diagnostics is increasing, there is a significant challenge using POC diagnostics data to improve epidemiological models. In this work, we propose a method to process zip-code level POC datasets and apply these processed data to calibrate an epidemiological model. We specifically develop a calibration algorithm using simulated annealing and calibrate a parsimonious equation-based model of modified Susceptible-Infected-Recovered (SIR) dynamics. The results show that parsimonious models are remarkably effective in predicting the dynamics observed in the number of infected patients and our calibration algorithm is sufficiently capable of predicting peak loads observed in POC diagnostics data while staying within reasonable and empirical parameter ranges reported in the literature. Additionally, we explore the future use of the calibrated values by testing the correlation between peak load and population density from Census data. Our results show that linearity assumptions for the relationships among various factors can be misleading, therefore further data sources and analysis are needed to identify relationships between additional parameters and existing calibrated ones. Calibration approaches such as ours can determine the values of newly added parameters along with existing ones and enable policy-makers to make better multi-scale decisions.

Analyses of school commuting data for exposure modeling purposes.

Human exposure models often make the simplifying assumption that school children attend school in the same census tract where they live. This paper analyzes that assumption and provides information on the temporal and spatial distributions associated with school commuting. The data were obtained using Oak Ridge National Laboratory's LandScan USA population distribution model applied to Philadelphia, PA. It is a high-resolution model used to allocate individual school-aged children to both a home and school location, and to devise a minimum-time home-to-school commuting path (called a trace) between the two locations. LandScan relies heavily on Geographic Information System (GIS) data. With respect to school children attending school in their home census tract, the vast majority does not in Philadelphia. Our analyses found that: (1) about 32% of the students walk across two or more census tracts going to school and 40% of them walk across four or more census blocks; and (2) 60% drive across four or more census tracts going to school and 50% drive across 10 or more census blocks. We also find that: (3) using a 5-min commuting time interval - as opposed to the modeled "trace" - results in misclassifying the "actual" path taken in 90% of the census blocks, 70% of the block groups, and 50% of the tracts; (4) a 1-min time interval is needed to reasonably resolve time spent in the various census unit designations; and (5) approximately 50% of both the homes and schools of Philadelphia school children are located within 160 m of highly traveled roads, and 64% of the schools are located within 200 m. These findings are very important when modeling school children's exposures, especially, when ascertaining the impacts of near-roadway concentrations on their total daily body burden. As many school children also travel along these streets and roadways to get to school, a majority of children in Philadelphia are in mobile source-dominated locations most of the day. We hypothesize that exposures of school children in Philadelphia to benzene and particulate matter will be much higher than if home and school locations and commuting paths at a 1-min time resolution are not explicitly modeled in an exposure assessment. Undertaking such an assessment will be the topic of a future paper.