key: cord-354009-1ek4s8oe
authors: Wang, Yun; Liu, Ying; Struthers, James; Lian, Min
title: Spatiotemporal Characteristics of COVID-19 Epidemic in the United States
date: 2020-07-08
journal: Clin Infect Dis
DOI: 10.1093/cid/ciaa934
sha: 
doc_id: 354009
cord_uid: 1ek4s8oe

BACKGROUND: A range of near-real-time online/mobile mapping dashboards and applications have been used to track the COVID-19 pandemic worldwide. It remains unknown about small area-based spatiotemporal patterns of COVID-19 in the United States. METHODS: We obtained county-based counts of COVID-19 cases confirmed in the United States from January 22 to May 13, 2020 (N=1,386,050). We characterized the dynamics of COVID-19 epidemic through detecting weekly hotspots of newly confirmed cases using Spatial and Space-Time Scan Statistics and quantifying the trends of incidence of COVID-19 by county characteristics using the Joinpoint analysis. RESULTS: Along with the national plateau reached in early April, COVID-19 incidence significantly decreased in the Northeast (estimated weekly percentage changes [EWPC]: -16.6%), but remained increasing in the Midwest, South and West Regions (EWPCs: 13.2%, 5.6%, and 5.7%, respectively). Higher risks of clustering and incidence of COVID-19 were consistently observed in metropolitan vs rural counties, counties closest to core airports, most populous counties, and counties with highest proportion of racial/ethnic minority counties. However, geographic differences in the incidence have shrunk since early April, driven by a significant decrease in the incidence in these counties (EWPC range: -2.0% – -4.2%) and a consistent increase in other areas (EWPC range: 1.5% – 20.3%). CONCLUSIONS: To substantially decrease the nationwide incidence of COVID-19, strict social distancing measures should be continuously implemented, especially in geographic areas with increasing risks, including rural areas. Spatiotemporal characteristics and trends of COVID-19 should be considered in decision-making on the timeline of re-opening for states and localities.

Since the first cluster of the coronavirus disease 2019 (COVID-19) was reported, 1,2 the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has triggered massive outbreaks and then evolved to a worldwide pandemic of COVID-19. As of May 13, 2020, 4,347,018 confirmed cases and 297,197 COVID-19 related deaths have been reported worldwide. 3 In the United States, the first COVID-19 case was reported on January 21, 2020, 4 and national outbreak of COVID-19 beginning in early March of 2020 had caused 1,386,050 confirmed cases and 83,167 deaths from COVID-19 as of May 13. 5 It is urgent to "flatten the epidemic curve" for COVID-19 in the United States.

Remarkable efforts have been made to map the coronavirus spread using near-real-time interactive online/mobile GIS dashboards, websites, and applications in and out the United States. 3, [5] [6] [7] These maps provide timely information on descriptive statistics of the outbreak situation. However, no studies have comprehensively assessed small area-based characteristics of COVID-19 spreading in the United States. Using government record-based surveillance data, we examined the spatiotemporal variations in COVID-19 as well as its associated geographic characteristics across the country. The results would enhance our understanding of small area-based spatiotemporal dynamics of COVID-19 outbreak, thus help inform multilevel strategies to control the spread of coronavirus and appropriate allocations of public health and healthcare resources in the United States.

A c c e p t e d M a n u s c r i p t 5

We obtained the counts of COVID-19 cases diagnosed from January 22 to May 13, 2020 in the United States from the USAFacts, a not-for-profit initiative standardizing and providing the government record-based data publicly available. 5 The daily-updated numbers were cumulated to form a time-series database of confirmed COVID-19 cases across all the US counties. The study is exempted from the ethics review due to the use of publicly accessible data source.

To identify the characteristics of counties with a high burden of COVID-19, we examined county-level geographic and sociodemographic factors, including rural-urban context, distance to the nearest core airport, population density, percentage of non-white minority population, percentage of population 65 years or older, and percentage of population below the federal poverty line. Using the Rural-Urban Continuum Codes of U.S. Department of Agriculture, 8 rural-urban context was defined as metropolitan (codes 1-3), urban (codes 4-7), and rural areas (codes [8] [9] . There are 30 core airports with the highest volume of traffic across the country. 9 The Euclidean distance from the populationweighted centroid of a given county to its nearest core airport was calculated to measure spatial relationship of that county with core airports. Population density was computed as the number of population per square miles of lands. County-level information on land areas, population sizes, and other three socioeconomic variables were retrieved from combined 2014-2018 American Community Surveys to reduce potential marginal error of survey.

A c c e p t e d M a n u s c r i p t 6 Statistical Analysis:

We first created an epidemic curve to visualize the progression of newly confirmed COVID-19 cases by four US government-defined Regions (Northeast, Midwest, South, and West) over 11 distinct time periods from January 22 through May 13, 2020, including the first six epi-weeks in combination (January 22 nd -March 4 th ) and individual epi-weeks from March 5 th to May 13 th .

Using a Spatial and Space-Time Scan Statistics (SaTScan), 10, 11 we examined spatiotemporal clustering of confirmed COVID-19 cases through detecting the higher-than-expected geographic hotspots across the country. The SaTScan applies a pre-defined circular window with varied sizes and time periods to scan the study area and identify the most likely clusters of the event of interest using a space-time permutation statistical model, and uses a Monte Carlo simulation approach to generate 999 random datasets in computing the statistic for the statistical inference of a cluster. In this study, we defined the parameters of the scanning window as 150 miles of maximum geographic radius and the day as the minimum temporal scanning unit. Geographic clustering was detected in each of 11 time periods to characterize the dynamics of geographic clustering of newly confirmed COVID-19 cases.

The most likely high-risk clusters/hotspots were captured based on the Monte Carlo Rank with P<0.05. We further examined the associations of county characteristics with COVID-19 clustering using logistic regressions. The outcome was whether a given county was identified as part of a hotspot or not. The analysis was performed separately for each of the 7 th -16 th epi-weeks. Considering the colinearity between county characteristics, county-level variables were not mutually adjusted for.

Statistical significance was tested as two-sided with P<0.05. 

As of May 13, 2020, a total of 1,386,050 COVID-19 cases were confirmed in the United States over 16 epi-weeks. Figure 1A shows the overall temporal trend of weekly counts of newly confirmed COVID-19 cases by four US Regions. COVID-19 had occurred sporadically until early March (first six epi-weeks); 103 confirmed cases were reported mainly in the West Region. The number of weekly confirmed COVID-19 cases subexponentially increased across the country from the 7 th to 11 th epi-week, followed by a slowly decrease over recent five weeks. During the entire observation period, the largest proportion of cases was from the Northeast (48.6%), followed by the 

In the first six epi-weeks, COVID-19 cases were reported in 27 counties from the west coast and Northeast states with the highest county-level incidence of 3.4/100,000 persons. Starting in the 7 th through the 16 th epi-weeks, SARS-CoV-2 spread to broad geographic areas. As of May 13, 92.0% of US counties had the confirmed COVID-19 cases, and the median county-level cumulative incidence rate was 88.0/100,000 persons (interquartile range: 36.1-219.3/100,000 persons) with the highest reaching 14,426/100,000 persons (Supplemental Figure 1) . The incidence of COVID-19 reached the peak in the Northeast in the 12 th epi-week (214.2/100,000 persons), followed by a significant reduction of 16.6% weekly until the 16 th epi-week. However, the incidence consistently increased in the Midwest, South and West regions from the 10 th to 16 th epi-weeks with a significant EWPC of 13.2%, 5.6% and 5.7%, respectively. Overall, COVID-19 incidence reached the national plateau in the epi-week 11 (66.6/10,000 persons), followed by a slight and insignificant decrease in recent five weeks ( Figure 1B and Table 2 ). Figure 3 illustrates the trends in the incidence of COVID-19 by county characteristics. Over 16 epi-weeks, the incidence was significantly higher in metropolitan vs urban/rural areas, areas closest vs farthest to core airports, most vs least populous areas, and areas with highest vs lowest percentage of minorities, and lowest vs highest percentage of population aged 65 years and older ( Figure 3A-E) .

The incidence dramatically increased from the 7 th epi-week and reached the peak in the 11 th epi-week in metropolitan areas (75.7/100,000 persons), counties closest to core airports (91.7/100,000 persons), most populous counties (79.6/100,000 persons), and counties with highest percentage of minorities (100.9/100,000 persons), followed by a significant decrease thereafter (EWPC= 2.0%, 2.8%, 2.6%, A c c e p t e d M a n u s c r i p t 10 and 4.2%, respectively) ( Table 2) . Notably, the incidence consistently increased from epi-week 7-16 in rural (0.04 to 35.0/100,000 persons) and urban areas (0.1 to 37.5/100,000 persons) ( Figure 3A ). Unlike metropolitan areas, the incidence remained increasing in rural and urban areas after the 11 th epi-week with a significant EWPC of 17.8% and 18.1%, respectively (Table 2) . Similarly, a consistent increase in the incidence of COVID-19 from epi-week 7-16 was also observed in counties farther from core airports and in less populous and lower minority counties ( Figure 3B -D and Table 2 ). The incidence in counties with lowest or highest percentage of elderly people increased from epi-week 7-11 and remained steady thereafter. Overall, geographic disparities in the incidence of COVID-19 by county characteristics had decreased since the 11 th epi-week. There was no significant difference in the incidence of COVID-19 for highest vs lowest percentage of population below the federal poverty line ( Figure 3F ).

Using a national time-series database of confirmed COVID-19 cases, we examined the spatiotemporal patterns of COVID-19 in the United States during the starting 16 epi-weeks. COVID-19 cases sporadically occurred in the west coast and Northeast states in the first six epi-weeks and increased rapidly across the country thereafter until the 11 th epi-week, and then slightly decreased since the 12 th epi-week. Despite a remarkable reduction in newly confirmed cases from the Northeast in recent four weeks, the risk of COVID-19 infection remained consistently increasing in the Midwest, South and West Regions. Geographic clustering of COVID-19 was first identified in southern and northern California, and then rapidly expanded nationwide. Higher risks of COVID-19 clustering and incidence were observed in metropolitan vs rural counties, counties closest to core airports, most populous counties, and counties with highest proportion of racial/ethnic minority. However, the differences have shrunk since the 11 th epi-week, which was driven by a significant decrease in the incidence in these counties and a consistent increase in other areas in recent five weeks. It might be a result of social distancing measures well implemented recently in high-risk areas in early stage of the A c c e p t e d M a n u s c r i p t 11 outbreak, and also suggests that recent region-to-region spreading and community transmission occurred in other areas. Further studies are needed to assess the effectiveness of public health and behavioral interventions on COVID-19 infection and implemental barriers, which is essential for promoting the strict adherence to social distancing guidelines and enhancing personal protections (including appropriately wearing face masks as needed and timely washing hands) to prevent the SARS-CoV-2 spreading and thus substantially decrease the incidence of COVID-19 locally and nationwide.

A significant association between short distance to core airports and COVID-19 clustering suggests a critical role of air transportation in SARS-CoV-2 spreading across the country. Air transportation was believed to accelerate and amplify the spread of influenza, SARS-CoV or MERS-CoV. 13 A recent study showed that the rail transport is related to the transmission and regional outbreak of COVID-19 in China. 14 In the United States, the airports may contribute substantially to the early travel-related region-to-region transmission. From March 18 to April 22, at least 42 states, 3 counties, 10 cities, the District of Columbia and Puerto Rico joined Illinois, New York, and California in the lockdown orders. 15 However, airlines, as one of essential transportation services, are generally exempted from the orders and still operating. We flagged the importance of airports in spreading COVID-19 even after the lockdown of most regions in mid-March. Airlines have put in place stringent safeguards for those still flying, including supercharged cleaning, reduced in-flight services, and the spacing out of passengers on flights. It is crucial to maintain strict management and monitoring of major airports to maximize the reduction of region-to-region transmission.

While COVID-19 incidence in metropolitan areas has decreased since the 11 th epi-week, we identified a consistent increase in the incidence of COVID-19 in rural areas over the 16 epi-weeks.

This was probably a sign of that the local spread of COVID-19 extended beyond metro/urban enclaves and the secondary community transmissions took place around geographic hotspots and spread to rural areas. Rural areas with a lower population density are not safe for this pandemic because rural residents tend to be older and have limited access to health care. 16, 17 Therefore, the A c c e p t e d M a n u s c r i p t 12 restrictive social-distancing measures are necessary in rural areas, and adherence to social distancing should be enhanced for rural residents.

The pandemic of COVID-19 poses different challenges for the US states currently designing its coping strategies. The population density is a key driver for transmission of infectious disease. We observed that predominately minority counties were at higher risk of COVID-19. This was consistent with the reports of African Americans accounting for about 70% of COVID-19 related deaths but just about 30% of the population in Chicago, Milwaukee County and Louisiana. 18 The disproportionate burden of COVID-19 in minority populations may largely result from inequities in adherence to social distancing measures. Our analysis indicates that the incidence of COVID-19 was lower in areas with higher percentage of elderly people. It could partly result from the lower mobility of older vs younger people. However, we found a comparable risk of COVID-19 clustering in counties with highest vs lowest percentage of elderly people in recent weeks. It might be relevant to the outbreaks of COVID-19 in nursing homes in some geographic areas. 19, 20 Therefore, senior-heavy areas should not be ignored in the allocation of prevention efforts on COVID-19 because senior people typically have multiple chronic health conditions and higher risk of developing more serious complications from COVID-19. 21 Lack of a stable pattern in the association between poverty and the risk of COVID-19

indicates that socioeconomic factors might not play a critical role in the risk of coronavirus infection.

The study has some limitations. The confirmed cases of COVID-19 might not reflect the actual number of persons infected with SARS-CoV-2 due to unknown/untested asymptomatic cases. [22] [23] [24] We used the reliable governmental records of lab-confirmed cases of COVID-19 in the first 16 

World Health Organization. Pneumonia of unknown cause -China. Disease outbreak news

A Novel Coronavirus from Patients with Pneumonia in China

An interactive web-based dashboard to track COVID-19 in real time. The Lancet Infectious diseases

First Case of 2019 Novel Coronavirus in the United States. The New England journal of medicine

Centers for Disease Control and Prevention. Cases of Coronavirus Disease (COVID-19) in the U.S. 2020

Latest Map and Case Count

Rural residence and cancer outcomes in the United States: issues and challenges. Cancer epidemiology, biomarkers & prevention : a publication of the American Association for Cancer Research

Air Traffic Control Modernization: Progress and Challenges in Implementing NextGen

Spatial disease clusters: detection and inference

A space-time permutation scan statistic for disease outbreak detection

Permutation tests for joinpoint regression with applications to cancer rates

The roles of transportation and transportation hubs in the propagation of influenza and coronaviruses: a systematic review

The association between domestic train transportation and novel coronavirus (2019-nCoV) outbreak in China from 2019 to 2020: A datadriven correlational report. Travel medicine and infectious disease

The New York Times. See which states and cities have told residents to stay at home

Access and issues of equity in remote/rural areas. The Journal of rural health : official journal of the American Rural Health Association and the National Rural Health Care Association

Rural-urban differences in usual source of care and ambulatory service use: analyses of national data using Urban Influence Codes

COVID-19 and African Americans

Nursing Home Care in Crisis in the Wake of COVID-19

Long-Term Care Policy after Covid-19 -Solving the Nursing Home Crisis. The New England journal of medicine

Clinical course and risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study

Transmission of 2019-nCoV Infection from an Asymptomatic Contact in Germany. The New England journal of medicine

Asymptomatic cases in a family cluster with SARS-CoV-2 infection. The Lancet Infectious diseases

Centers for Disease Control and Prevention. Social distancing, quarantine, and isolation: keep your distance to slow the spread

How will countrybased mitigation measures influence the course of the COVID-19 epidemic?

Hospitalization Rates and Characteristics of Patients Hospitalized with Laboratory-Confirmed Coronavirus Disease 2019 -COVID-NET, 14 States

Can China's COVID-19 strategy work elsewhere?

Governmental Public Health Powers During the COVID-19 Pandemic: Stay-at-home Orders, Business Closures, and Travel Restrictions. Jama. 2020. 30. Pan A

Public Health Interventions for COVID-19: Emerging Evidence and Implications for an Evolving Public Health Crisis

Reopening Society and the Need for Real-Time Assessment of COVID-19 at the Community Level

Population Density (Quartile)

We thank the GIS and spatial statistics supporting from the Health Behavior, and Communication & Outreach Core, which is affiliated with Washington University Institute of Clinical Translational Sciences funded by the National Center for Advancing Translational Sciences, National Institutes of Health (UL1 TR002345) and Washington University Alvin J. Siteman Cancer Center funded by the National Cancer Institute, National Institutes of Health (P30 CA091842).

A c c e p t e d M a n u s c r i p t 15 A c c e p t e d M a n u s c r i p t A c c e p t e d M a n u s c r i p t 20