Assessing Different Measurements of Heat Exposure and Scale: The Spatiotemporal Dynamics of Heat-Related Hospitalizations in Michigan

BACKGROUND: With global warming, heat morbidity and mortality is a growing health concern.  The number of “extreme heat” events or average summer temperature may be statistically correlated to the number of heat-related hospitalizations. By exploring these associations, predictive models may be improved, and decisions about when, where, and how heat warnings and response plans are implemented may be refined.  

METHODS: The National Oceanic and Atmospheric Administration (NOAA) uses a Heat Index (HI) derived from air temperature and humidity to classify “extreme heat” events for individual Michigan counties. The Great Lakes Integrative Sciences and Assessment Center (GLISA) compiles measurements from over 200 climate stations to interpolate average seasonal temperature over time for Michigan’s 10 Climate Divisions. The Michigan Inpatient Database (MIDB) contains discharge records from Michigan’s acute care hospitals.  Michigan resident cases of heat-related hospitalizations between 1999 and 2013, defined by diagnostic (ICD-9 CM) codes 992.0-992.9 and cause of injury code E900, were correlated with number of extreme heat events by county and the average summer temperature per Division to explore the impact of different heat exposure measures and geographic scales.  Cases and heat exposures were mapped using ArcGIS.

RESULTS: During January 1, 1999 to December 31, 2012 there were 2,002 heat-related hospitalizations in Michigan with a linear regression trendline showing a gradual increase. The number of total “extreme heat” events ranged from 0 to 14 events within individual counties from 1999 to 2012. Most cases and events were located in urban, southeastern Michigan, although high case rates were observed in some northern rural areas. Statistically significant correlations were found for associations between county case numbers and number of “extreme heat” events (r² = 0.587, p-value <0.0001), and Climate Division case numbers and average summer temperatures (r² = 0.465, p-value <0.0001). Maps of heat exposures and heat-related illness outcomes visualize the correlations.

CONCLUSIONS: Both NOAA “extreme heat” events and GLISA average summer temperature exposures were useful in establishing relationships and thresholds for heat-related hospitalizations in Michigan, and both could be employed to predict future heat-related disease burden. Thresholds may need to be modified to address local socioeconomic and physical vulnerabilities in order to be effective in preventing heat-related illness. The spatial scale at which “heat” is considered an exposure can also dictate when and where warnings are implemented and should be tailored differently for county and Climate Divisions.