In this project, we will examine the influence of long-term, seasonal, and daily weather and climate changes on hospital and emergency department admissions in Virginia health care facilities.
- Weather, climate, and human health
- Human bioclimatology
RDRobert E. DavisCollege of Arts and Sciences
HKHyojung KangSchool of Engineering and Applied Science
KEKyle B. EnfieldSchool of Medicine
Time series of daily hospital and emergency department (ED) admissions exhibit components of long-term, intra-annual, and daily variability, and many of these fluctuations are associated with weather and climate factors. For example, seasonal variations in air quality and pollen impact respiratory distress; heat and cold waves exert physiological strain on the cardiovascular and respiratory systems; and influenza transmission has been linked to periods of cold and dry air. The daily admissions changes from environmental factors are typically large enough to impact staffing and resource delivery, including beds. Since many of the weather and climate components are predictable, we can incorporate seasonal modeling and short-term weather forecasts to provide daily predictions of ED and hospital admissions. We propose to develop and test machine learning models by combining patient electronic medical records and daily hospital utilization data from facilities in Virginia with daily weather data from National Weather Service observation stations throughout the Commonwealth. Each ED or hospital admission would be georeferenced to the nearest weather station. Findings of this study would be useful to optimize hospital utilization over the course of a year and should result in more optimal hospital and ED resource allocation while simultaneously improving the patient experience.
Improved understanding of the influence of weather and climate fluctuations on human health.
Site-specific forecasts of medical facility utilization rates by coupling historic weather and admissions patterns with short-term forecasts.
Development of machine-learning tools to optimize model development and forecasting.
Most of the funding will be used for student stipends.
Both graduate and undergraduate students will participate in all aspects of the project, including the development of hypotheses and dataset design to modeling and model testing.
Student participants will be coauthors on all papers and presentations that arise from this research.