What is a COVID-19 hot spot?
Typically, a hot spot is an area that has a higher than average number or intensity of events. In our project, we define a COVID-19 hot spot as a county that exhibits the concentration of high case/death counts (or rates). Similarly, a COVID-19 cold spot indicates a spatial cluster of low case/death counts (or rates). A county is classified as a hot spot when:
- It has a high value and its neighboring counties also have similarly high values
- This clustering pattern is unlikely to be the result of random chance
We perform hot spot analysis to identify COVID-19 clusters (hot and cold spots) at a county level based on:
- The total number of confirmed cases
- Total number of deaths
- Case rates (cases per 100,000 people)
- Death rates (deaths per 100,000 people)
How do we interpret the hot spot maps?
In the maps, hot spots are represented by the red colors while cold spots are marked in blue. The darker the red or blue colors are, the more intense the clusters of high or low values are. In other words, the darker the colors appear, the more confident we can be that such clustering is not the result of a random process. For example, a county colored in dark red suggests that there is only a 1% chance that it might not be a hot spot. A county in medium red indicates that there is a 5% chance that it might not be a hot spot. Similarly, a county in light orange suggests that there is a 10% chance that it might not be a hot spot. Counties with no pattern detected are shaded in gray.
How is a COVID-19 hot spot determined?
We calculate the Getis-Ord Gi* statistic for each county using a geographic information system (GIS) to identify the locations of COVID-19 hot spots in Eastern North Carolina. The Getis-Ord Gi* statistic investigates where high/low values cluster spatially and whether the observed clusters are statistically significant. To be classified as a significant hot spot, a county with a high value should be surrounded by neighbors with high values. More specifically, a county is classified as a hot spot if the local sum for the county is significantly higher than the expected local sum and that difference is too large to be the result of a random process. The local sum for the county is determined by the values of its neighboring counties. We define neighbors of a county as those that share an edge with the county.
What have we found?
- The locations that are most affected by the COVID-19 pandemic have changed over time.
- Although there are monthly variations, the majority of hot spots for the case and death counts were found in the ENC 12-county sub-region (ENC 12; see map below).
- However, when adjusted for population (cases/deaths per 100,000 people), hot spot patterns became different. Most hot spots for case/death rates were identified in the ENC 29-county sub-region (ENC 29; see map below). This indicates that the pandemic is more seriously affecting the population in ENC 29 than those in ENC 12. This pattern has become more prominent over time, particularly since July.
- The burden of the pandemic is greater in ENC 29, which is a region known to exhibit higher mortality rates from diseases, including diabetes, asthma, or heart disease, than the rest of the state in general. It suggests that people in ENC 29 have a higher than average risk associated with COVID-19, given that underlying health conditions are likely to put people at a greater risk for severe illness from the virus.
- The locations of clusters for high case counts/rates did not always correspond to those for high death counts/rates.
- There are a few cold spots in northern areas near the coast and they remain as cold spots or non-significant hot/cold spots even after adjusting for population.
Why is the identification of hot spots important for public health practice?
Identifying the locations of COVID-19 clustering is essential to successful public health responses to the pandemic. Hot spot maps can be used by public health officials, health care providers, or health researchers to better understand the spatio-temporal distribution of the pandemic, target the implementation of effective public health action and control practices to prevent the acceleration of the disease clustering, and explore possible causes of high numbers or rates in hot spots. Because the method we use returns “statistically significant” hot and cold spots, the analysis results can provide public health decision-makers with reliable and actionable information.
Case and death counts: New York Times COVID-19 data (U.S. county-level data) (https://github.com/nytimes/covid-19-data)
Population estimates: US Census Bureau American Community Survey 2014-2018 (5-year) data (https://www.census.gov/newsroom/press-kits/2019/acs-5-year.html)