key: cord-0705030-6ycu46r5
authors: Sung, Baksun
title: A spatial analysis of the effect of neighborhood contexts on cumulative number of confirmed cases of COVID-19 in U.S. Counties through October 20 2020
date: 2021-02-17
journal: Prev Med
DOI: 10.1016/j.ypmed.2021.106457
sha: 3c2f0397cd459898f3eb1127b6b42b3131bf039d
doc_id: 705030
cord_uid: 6ycu46r5

COVID-19 has become a nationwide public health crisis in the United States and the number of COVID-19 cases is different by U.S. counties. Also, previous studies have reported that neighborhood contexts have an influence on health outcomes. Therefore, the objective of this study was to examine the association between neighborhood contexts and cumulative number of confirmed COVID-19 cases (per 100,000) in U.S. counties. Cumulative number of COVID-19 cases gained from USA FACTS and variables related to neighborhood contexts gained from the 2018 5-Year American Community Survey at the county level. Data were analyzed using spatial autoregressive models. According to the present results, firstly, larger population, high poverty rate, higher % of bachelor's degree, higher % of no health insurance, higher employment rate, higher % of manufacturing jobs, higher % of primary industry jobs, higher % of commute by drove alone, higher % of foreign born, higher % of Hispanic, and higher % of Black are positively associated with higher cumulative number of COVID-19 cases. Secondly, higher income, higher % of cash assistance recipient, higher % of SNAP recipient, higher unemployment rate, higher % of commute by walked, higher % of Asian, and higher % of senior citizen are negatively associated with higher cumulative number of COVID-19 cases. In conclusion, there exist geographical differences in cumulative number of COVID-19 cases in U.S. counties, which is influenced by various neighborhood contexts. Hence, these findings emphasize the need to take various neighborhood contexts into account when planning COVID-19 prevention.

Coronavirus disease 2019 (COVID-19) is a highly contagious respiratory disease, which is caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) (CDC, 2020a). The COVID-19 virus is more infectious than other corona viruses such as SARS and MERS even though it is less severe (CDC, 2020a) . The COVID-19 virus is a person to person transmitted disease that spreads mainly through respiratory droplet and saliva of infected people (CDC, 2020b) . The United States have one of the most cumulative confirmed cases of COVID-19 in the world. Cumulative confirmed COVID-19 case totals for the United States was 8,202,847 through October 20, 2020 (USA FACTS, 2020 . Although COVID-19 virus has spread across the United States, the number of COVID-19 cases is different by counties (USA FACTS, 2020).

Neighborhood contexts and regional characteristics might lead to geographical difference in the number of COVID-19 cases.

Previous studies have reported that poor neighborhood contexts particularly those characterized by disadvantaged socioeconomic status are associated with increased body mass index and mortality (Andersen et al., 2018; Frank et al., 2006; Wong et al., 2018) . Also, residential segregation has a negative effect on health status among ethnic minorities (Greer et al.,

Moreover, previous research studies have reported that poor neighborhood contexts particularly those characterized by social vulnerability, minorities, and low socioeconomic status are associated with increased risk of COVID-19 (Khazanchi et al., 2020; van Holm et al., 2020; Yancy et al., 2020) . However, these studies are based on the initial outbreak of COVID-19 that may underestimate cases of COVID-19 in some counties. In addition, these studies did not deal with occupational variables. Neighborhood contexts are classified by occupational composition (Clarke et al., 2013 ) that a certain type of occupational variable can lead to higher number of COVID-19 cases. Also, other related studies have focused mainly on the concepts and techniques of GIS (Mollalo et al., 2020; Sun et al., 2020) . More research is needed, therefore, to reexamine the association between various neighborhood contexts including occupational variables and the number of COVID-19 cases from public health and social perspectives. Thus, the objective of this study was to examine the association between neighborhood contexts and cumulative number of confirmed COVID-19 cases (per 100,000) in 3,136 U.S. counties through October 20, 2020.

This study is designed to analyze cumulative number of confirmed COVID-19 cases as on time period with pre-existing cross-sectional neighborhood contexts in the United States. Ordinary least squares (OLS) regression model is vulnerable to bias caused by potential spatial autocorrelation because it presupposes that all variables are independent which ignores potential spatial dependencies (Anselin, 1988; Conway et al., 2010) . According to the primary law of geography, even though every spatial unit is linked to everything else, spatial units near to each other are more strongly linked (Anselin, 1988) . To control potential bias related to spatial autocorrelation, therefore, spatial autoregressive models were used to analyze the association between neighborhood contexts and cumulative number of COVID-19 cases.

Firstly, the spatial lag model demonstrates that result in one spatial unit is linked to result in another spatial unit (Liu, 2020). The main function of this model is to remedy for spatial dependence by adopting a term for the impact of the spatially lagged Y on Y (Anselin et al., 1996; Conway et al., 2010) . It means that the result variable in location a is affected by neighboring location b (Anselin et al., 1996; Conway et al., 2010) . It can be summarized as follows: Y = ρWY + Xβ + ε, Y: dependent variable ρ: lag coefficient W: spatial weight matrix β: coefficient for a vector of neighborhood contexts ε: error term Secondly, the spatial error model demonstrates that unobserved factors in one spatial unit are linked to unobserved factor in another spatial unit (Liu, 2020). It means that the error term in location a is affected by neighboring location b (Anselin et al., 1996; Conway et al., 2010) . It can be summarized as follows: Table 1 shows the descriptive statistics. First, the average poverty rate in 3,136 U.S. counties is 11.24%. Second, the average percentage of bachelor's degree in 3,136 U.S. counties is 14.01%. Third, the average percentage of no health insurance in 3,136 U.S. counties is 6.27%. Table 2 Results from spatial autoregressive models 

American Community Survey (ACS). 2020. 2014-2018 ACS 5-Year Data Profile

Associations between neighborhood environment, health behaviors, and mortality

Lagrange multiplier test diagnostics for spatial dependence and spatial heterogeneity

Simple diagnostic tests for spatial dependence

Social context explains race disparities in obesity among women

COVID-19) 2020 Interim Case Definition

Centers for Disease Control and Prevention (CDC). 2020b. Coronavirus Disease 2019 (COVID-National Center for Education Statistics (NCES). 2020. FAST FACTS. Employment rates of college graduates

75th W.E. UPJOHN INSTITUTE FOR EMPLOYMENT RESEARCH. Food Stamps and Unemployment Compensation in the COVID-19 Crisis

A spatial analysis of the COVID-19 period prevalence in U.S. counties through

Health behaviors and allcause mortality in African American men

2020. BLS REPORTS. A profile of the working poor

US Coronavirus Cases and Deaths. Track COVID-19 data daily by state and county

Forestry, Fishing, Hunting, and Mining Occupations