BACKGROUND: In the second week of January, 2014, a polar vortex caused once-in-a-generation cold temperatures across much of the United States. Given the exceptional nature of the weather conditions, it was expected that there would be impacts on public health and the healthcare system. Emergency department visits for cold-induced conditions such as hypothermia or frostbite are among the most likely outcomes, but there are broader questions to consider as well, such as whether the extreme cold would cause an overall increase in emergency department utilization or keep people away. Reviewing time series charts for each county would be prohibitive and not capture any spatial trends, and a shaded map cannot convey temporal patterns. Here we present a visualization for exploring spatio-temporal trends over a large geographic area.
METHODS: The EpiCenter syndromic surveillance system is connected to hospitals in the states of Ohio, Pennsylvania, and New Jersey. Emergency department registrations from these hospitals were collected before, during, and after the polar vortex event. A visualization was developed to display spatially-located time series for this entire region, with additional shading to provide information on temporal trends and temperature measurements. Additional charts stratified by age and gender were also created using the same method.
RESULTS: Exploration of the data revealed regional differences in the effect on emergency department volume. Most of New Jersey saw increased utilization, while most of Pennsylvania saw decreased utilization; these were both confirmed as statistically significant trends via linear regression analysis. Ohio showed a more mixed response, but with some regional patterns that suggested a spatial structure. Stratified analysis indicated difference in emergency department utilization by age; this was also confirmed by linear regression.
CONCLUSIONS: Visualizing large amounts of data in a single display did reveal spatial and temporal trends that would have been difficult to identify in an exploratory fashion via 176 county-level time series charts, or any single summary statistic. The underlying causes of these trends are still being investigated; for example, given the different utilization patterns among different age groups, it is possible that demographics differences in different regions explain some of the spatial patterns in emergency department volume. Understanding the factors that drive healthcare utilization in different circumstances, using tools like these, can improve planning and preparedness for future extreme weather events, and potentially also provide insight into broader usage patterns and how they will be affected by policy changes.