Determinants of COVID-19 Incidence and Mortality in the US: Spatial Analysis

SARS-CoV-2, Coronavirus disease 2019, Public policy, Social Determinants of Health, 21 Health status disparities, Spatial analysis 22 Word count of Manuscript: 2,994 23 Word count of Abstract: 249 24 Tables: 2 25 Figures: 3 26 Supplementary Material: 1 27 Number of references: 27 28 29 . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 24, 2021. ; https://doi.org/10.1101/2020.12.02.20242685doi: medRxiv preprint NOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice. 2 Determinants of COVID-19 Incidence and Mortality in the US: Spatial Analysis 30

31

The US continues to account for the highest proportion of the global Coronavirus 32 Disease-2019 (COVID-19) cases and deaths. Currently, it is important to contextualize the spread and 33 success of mitigation efforts. The objective of this study was to assess the ecological determinants 34 (policy, health behaviors, socio-economic, physical environment, and clinical care) of COVID-19 35 incidence and mortality in the US. 36

Data from the New York Times’ COVID-19 repository (01/21/2020-10/27/2020), 2020 37 County Health Rankings, 2016 County Presidential Election Returns, and 2018-2019 Area Health 38 Resource File were used. County-level logged incidence and mortality rate/million were modeled using 39 the Spatial Autoregressive Combined model and spatial lag model. 40

Counties with higher proportions of racial minorities (African American β= 0.007, Native 41 Americans β= 0.008, Hispanics β= 0.015), non-English speakers (β= 0.010), population density 42 ([logged] β= 0.028), and air pollution (β= 0.062) were significantly associated with high COVID-19 43 incidence rates. Subsequently, counties with higher Republican voters (β= 0.017), excessive drinkers 44 (β= 0.107), children in single-parent households (β= 0.018), uninsured adults (β= 0.038), racial 45 minorities (African American β= 0.032, Native Americans β= 0.034, Hispanics β= 0.037), females (β= 46 0.101), and population density ([logged] β= 0.270), air pollution (β= 0.130), and non-Whites/Whites’ 47 residential segregation (β= 0.014) were significantly associated with high COVID-19 mortality rates. 48 Additionally, longer state-level restrictions were associated with lower COVID-19 incidence and 49 mortality rates. 50

The spatial models identified longer state-level restrictions, population density, air 51 pollution, uninsured rate, and race/ethnicity as important determinants of the geographic disparities in 52 COVID-19 incidence and mortality. 53 54

SARS-CoV-2, Coronavirus disease 2019, Public policy, Social Determinants of Health, 55 Health status disparities, Spatial analysis 56 57 . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 24, 2021. ; https://doi.org/10.1101/2020.12.02.20242685doi: medRxiv preprint 3

58 The Coronavirus Disease 2019 (COVID-19) pandemic continues to spread in the US and around 59 the world. The US, as of January 11, 2021 recorded 22.6million COVID-19 cases and 375,141deaths. 1 60 The US cases and deaths continue to account for th e largest share of global cases (25%) and global 61 deaths(20%).1 The COVID-19 pandemic containment in the US was challenging actors due to virus 62 contagion characteristics, its pathophysiology, and socio-political factors. In response to the pandemic, 63 by the first week of April, all but two states and local authorities imposed social distancing measures and 64 restrictions. As of the writing of this study, many states who had eased phase 1 restrictions went on to 65 implement phase 2 restriction policies to curb the surge in COVID-19 infections during the holiday 66 season. 2–5 67 Restriction measures such as the closure of business establishments, stay-at-home, and social 68 distancing mandates severely impacted the economy. In response to the pandemic, the lawmakers passed 69 the Coronavirus Aid, Relief, and Economic Security Act (CARES) Act, 6 with additional relief expected 70 shortly. As of January 11, 2021, Moderna and Pfizer vaccines have received emergency use 71 authorization by the FDA and are rolled out in phases. Other biopharmaceutical companies are 72 conducting clinical trials for fifty-four COVID-19 vaccine candidates.7–9 The current vaccine rollout will 73 continue to gain momentum while other COVID-19 vaccines may receive market approval in the near 74 future. Despite this progress, experts recommend that the public follow COVID-19 safety measures due 75 to the slow initial rollout and uncertainties surrounding COVID-19 vaccines (virus mutations, duration 76 of immunity, real world effectiveness, vaccine uptake, etc.). 10 In summary, fiscal, legislative, and 77 scientific efforts are all underway to address the needs of the population in the current phase of the 78 COVID-19 pandemic. However, the assessment of the impact of ecological contextual factors such as 79 health behaviors, clinical care, socio-economic, and physical environment-related characteristics on the 80 course of COVID-19 pandemic is also necessary. The contextual understanding from such a study is 81 required to gauge which strategies work, to what extent, and for which groups. Having a thorough 82 knowledge of these ecological contextual factors is critical to address the public health and economic 83 challenges and prioritize resources. 84 Studies so far have generated predictive models for growth in COVID-19 incidence and 85 mortality and estimated the impacts of some community-level factors on COVID-19 incidence and 86 mortality. Millett et al. and Khanijahania et al. focused on assessing the ecological determinants of 87 susceptibility to COVID-19 outcomes among predominantly African American counties, while Fielding-88 Miller et al. and Peters et al. performed a similar assessment along the rural-urban continuum. 11–14 Few 89 . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 24, 2021. ; https://doi.org/10.1101/2020.12.02.20242685doi: medRxiv preprint 4 studies had assessed the spatial determinants of COVID-19 transmission. 15–17 Recently, Sun et al. 90 assessed the relationship between various county-level determinants and COVID-19 incidence while 91 Andersen et al. also identified high prevalence clusters. 15,16 These studies had evaluated the incidence 92 and prevalence during the initial phase of the pandemic and limited the analysis to a few county-level 93 factors. The goal of the current study was to assess the impact of county-level ecological factors, using 94 spatial econometric analysis, on the cumulative COVID-19 incidence and mortality rate over an 95 extended phase of the pandemic. 96 97

98 Data source and study design 99 The current study used county-level COVID-19 confirmed cases and deaths data from the New 100 York Times repository extracted as of October 27, 2020, and included data up until that date. 18 The 101 study used county-level characteristics from 2020 County Health Rankings data, 2018-2019 Area Health 102 Resource File data, 2016 County Presidential Election Returns data, and state-level lockdown duration 103 data was generated based on the lockdown start and end date obtained from Ballotpedia as of October 104 27, 2020 19–21. Counties from Alaska and Hawaii and counties with less than 25 cumulative cases as of 105 October 27, 2020, were excluded from the analysis. The US counties ESRI Shapefile were obtained 106 from the US Census Bureau. 22 The study employed a cross-sectional ecological study design to assess 107 the association between county-level characteristics on the cumulative COVID-19 incidence cases and 108 deaths. 109 Outcomes 110 The county-level cumulative COVID-19 cases and COVID-19 deaths as of October 27, 2020 111 (per million population) were the operational definition for COVID-19 incidence and mortality rate, 112 respectively. The resulting incidence and mortality rates were log-transformed due to the skewed nature 113 of the county-level cumulative incidence and mortality rates. 114 Covariates 115 The covariates selected to predict the county-level incidence and mortality were based on the 116 County Health Ranking framework. The framework categorizes health factors into four sub-categories, 117 namely health behaviors, clinical care, socioeconomic factors, physical environment (Figure 1). Each of 118 the sub-categories is further divided into individual components. Although the County Health Ranking 119 model assigns weights to each of the components, they were not utilized in the current analysis as no 120 composite rank score was calculated. The individual components were used as covariates in the 121 . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 24, 2021. ; https://doi.org/10.1101/2020.12.02.20242685doi: medRxiv preprint 5 regression. The final list of county-level covariates included in the model is described in 122 Supplementary Table 1 . In addition to the above county-level factors, length of state-level lockdown 123 duration and month (January/February/March) of the first reported case at the state-level were included 124 as covariates. Finally, the county-level composition of the Republican voters was also added to 125 determine the extent to which the county’s community health was associated with voting preference for 126 the 2016 presidential elections. 127 [INSERT FIGURE 1] 128 Statistical Analyses 129 Descriptive univariate statistics of the weighted county-level characteristics were generated. The 130 presence of spatial correlation was confirmed by performing Moran’s I test for correlation of ordinary 131 least square regression residuals. Due to the spatial correlation in the data, the current study employed a 132 spatial regression analysis approach. Two island counties were excluded from the spatial regression 133 analysis. Prior research on spatial analysis of COVID-19 used the spatial autoregressive combined 134 model (SAC) model to evaluate the association between period prevalence of COVID-19 and county-135 level characteristics 15. The SAC model was also adopted for modeling cumulative incidence and deaths 136 if both the spatial lag parameter (rho) and spatial error parameter (lambda) were statistically significant. 137 However, model simplification was attempted when one of the parameters was not significant. A first-138 order queen spatial weight matrix was employed for all spatial models. The queen matrix defines 139 neighbor relationships if the counties either share a border or a vertex. All analysis was performed in 140 SAS Studio University Edition (Cary, NC), QGIS v 3.16.0 (Berne, Switzerland), and RStudio (R) v 141 4.0.3 (Boston, Massachusetts). 142 143

144 The final analysis included data from 3,101 counties from the mainland US. Between January 20 145 to October 27, 2020, the population-weighted cumulative incidence and mortality rates for the mainland 146 US were 26,576 and 632 per million, respectively. Table 1 demonstrates descriptive statistics of the 147 incidence and mortality rates and county-level determinants, namely, health behaviors, clinical care, 148 socio-economic, and physical environment factors. Fifteen states had a length of the lockdown of 59 149 days or more, and 39 states reported their first case in the month of March. Overall, 2,499 counties 150 leaned Republican (50% of the votes in the county were for the Republican presidential candidate in 151 2016). The weighted proportion of adult smokers, adults with obesity, adults with physical inactivity, 152 Medicare enrollees that were administered influenza vaccines and adults with some college education 153 . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 24, 2021. ; https://doi.org/10.1101/2020.12.02.20242685doi: medRxiv preprint 6 were 15%, 29%, 23%, 19%, and 65%, respectively. Premature (<75 years) age-adjusted mortality was 154 344 per 100,000. Among socio-economic factors, unemployment was 4%, 12% of adults were 155 uninsured, the mean income inequality ratio was 5. About 12% were African Americans, 18% 156 Hispanics, 1% Native Americans, and 4% population not proficient in English. The mean percentage of 157 the population older than 65 years and less than 18 years were 16% and 23%, respectively. On average, 158 19% of counties were rural, the homeownership rate was 65%, 33% of the children lived in single-parent 159 households, and 18% of households had severe housing problems. At the county-level, the rate of 160 primary care physicians (logged) and preventable hospitalization (logged) was 4 and 8, respectively. 161 Importantly, no significant multicollinearity was observed in the analysis, as the VIF for the selected 162 variables was less than 7. 163 [INSERT TABLE 1] 164 165 Figure 2 presents the spatial distribution( septiles) of logged COVID-19 incidence rates. In the 166 West, Oregon High desert and Twin Falls areas, New Mexico’s Navajo and Apache areas, and 167 Washington’s Spokane areas had clusters of high incidence rates. In the Midwestern region, Tribal lands 168 and the Great Plains regions, including North Dakota, South Dakota, Nebraska and Iowa (congruence of 169 these borders near Sioux Falls area), Minnesota, Wisconsin (specifically Superior Upland area), and 170 Kansas regions had clusters with high incidence rate clusters. In the South, the Delta region along the 171 Mississippi river including states of Mississippi, Louisiana, Tennessee, central Alabama, and Arkansas 172 had clusters of high incidence rates. The Corpus Christie area and area along the US-Mexico border had 173 clusters of high incidence rates clusters. Additional high incidence rate clusters were found in the 174 Tallahassee, Jacksonville, Savannah area, and Miami-Dade County area. Conversely, low incidence 175 rates were in the Pacific-Northwest areas. The spatial distribution of the logged COVID-19 mortality 176 rates (by septiles) presented in Figure 3 demonstrated similar patterns. However, in the Northeastern 177 region, high mortality rate clusters were found between the greater Philadelphia area and Boston 178 Massachusetts. Contrarily, the mortality rate clusters in Wisconsin’s Superior Upland area were smaller 179 compared to the incidence rate clusters. 180 181 [INSERT FIGURE 2] 182 [INSERT FIGURE 3] 183 184 . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 24, 2021. ; https://doi.org/10.1101/2020.12.02.20242685doi: medRxiv preprint 7 Table 2 presents the results from spatial regression models that assessed the impact of ecological 185 determinants on COVID-19 incidence and mortality rates (logged). For the incidence rate SAC 186 regression, the rho and lambda model parameters both were significant. However, only the rho 187 parameter was significant for the mortality rate regression model. Therefore, the SAC model was used 188 for incidence rate analysis, while the spatial lag model was used for mortality rate analysis. The analysis 189 further tested the spatial correlation for the residuals of both models using Moran’s I statistic. The 190 Moran’s I for the model residuals was not significant for the incidence rate (SAC model) and mortality 191 rate (spatial lag model) regression, which indicated no significant spatial autocorrelation of residuals. 192 The length of lockdown order, a policy-related factor, was significantly associated with the incidence 193 rate. Compared to the counties that had no orders implemented, counties with an order length of 1-27 194 days, 36-58 days, and 59 or more days had a 0.290, 0.228, and 0.335 unit decrease in logged COVID-19 195 incidence rates, respectively. Among clinical care-related factors, a unit increase in logged primary care 196 physicians’ rate, logged preventable hospitalization rates and logged hospital beds were all associated 197 with 0.011, 0.045, and 0.004 unit increase in logged COVID-19 incidence rates, respectively. A 198 percentage point increase in influenza vaccinations and Republican political leaning was associated with 199 a 0.003 and 0.002-unit increase in logged COVID-19 incidence rates, respectively. Among other socio-200 economic factors, a percentage point increase in adults with some college education, children in single-201 parent households, and social association rate and logged median income resulted in 0.004, 0.003, 0.003, 202 and 0.158 unit decreases in logged incidence rates, respectively. Specifically, among the racial/ethnic 203 factors, a percentage point increase in African Americans, Native Americans, Hispanics, and those not 204 proficient in English was associated with 0.007, 0.008, 0.015, and 0.010-unit increase in logged 205 incidence rates, respectively. Conversely, a percentage point increase in 65 years or older and those less 206 than 18 years of age was associated with a 0.020 and 0.011 unit decrease in logged incidence rates. A 207 unit increase in African American and white residential segregation and a percentage point increase in 208 rurality were associated with a 0.002 and 0.001 unit decrease in logged incidence rates. On the other 209 hand, a unit increase in white and non-white residential segregation and particulate matter air pollution 210 were associated with an 0.003 and 0.062-unit increase in incidence rates, respectively. 211 [INSERT TABLE 2] 212 Subsequently, when compared to the counties that had no orders implemented, counties with 213 order length of 28-35 days, 36-58 days, and 59 or more days had a 0.824, 0.866, and 1.006 unit decrease 214 in logged COVID-19-related mortality rates, respectively. Compared to counties from states that first 215 reported COVID-19 cases in January, those with the first reported case in February were associated with 216 . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 24, 2021. ; https://doi.org/10.1101/2020.12.02.20242685doi: medRxiv preprint 8 a 0.628-unit decrease in logged mortality rates. Among health behaviors, a 1% increase in excessive 217 drinking, influenza vaccinations resulted in a 0.107 and 0.039 increase in logged mortality rates. Logged 218 primary care physicians’ rate and hospital beds resulted in 0.234 and 0.033 units increase in logged 219 mortality rates. Among socio-economic factors, a 1% increase in Republicans, females, children in 220 single-parent households, and uninsured adults resulted in a 0.017, 0.101, 0.018, and 0.038 units 221 increase in logged mortality rates, respectively. However, a 1% increase in adults with some college 222 education was negatively associated i.e. 0.041 decreases in logged mortality rates. Among racial/ethnic 223 composition, a 1% increase in African Americans, Native Americans, and Hispanics were associated 224 with 0.032, 0.034, and 0.037 unit increase in logged mortality rates. The logged population density was 225 also positively associated with incidence and mortality rates by 0.028 and 0.270-units, respectively. 226 Additionally, a 1% increase in air pollution, and residential segregation- Non-White/Whites resulted in a 227 0.130 and 0.014 increase in logged mortality rates. Whereas African Americans/White residential 228 segregation, and rurality were negatively associated with mortality rates by 0.010 and 0.014 units. 229

230 To the best of our knowledge, this is the first spatial analysis study that captured and assessed the 231 cumulative incidence and deaths during the majority of the year 2020 (January 21 to October 28, 2020) 232 of the COVID-19 pandemic in the US. The study found that counties with higher proportions of 233 Republican voters, social association rates, racial minorities (African Americans, Native Americans, and 234 Hispanics), those not proficient in English and counties with higher residential segregation between non-235 Whites & Whites, population density, pollution-particulate matter were associated with higher incidence 236 rates. However, counties with longer length of stay-at-home orders, a higher proportion of adults with 237 some college education, high-income, elderly, children, rurality, and higher segregation between African 238 Americans & Whites had lower incidence rates. Correspondingly, counties with higher Republican 239 voters, excessive drinkers, children in single-parent households, uninsured adults, racial minorities 240 (African Americans, Native Americans, and Hispanics), females, and higher population density, 241 pollution-particulate matter, and residential segregation between non-Whites & Whites had higher 242 COVID-19 mortality rate. 243 Few studies have assessed the impact of different ecological factors on COVID-19 incidence and 244 mortality rates in the US and worldwide. Similar to the current study, Peters et al. reported higher age, 245 population density, uninsured rate to be associated with increased susceptibility to COVID-19 outcomes 246 13. Khazanchi et al. and Nayak et al. also reported a similar association between higher county-level 247 susceptibility score and higher incidence of COVID-19 incidence and deaths 12,23. Similar to our study, 248 . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 24, 2021. ; https://doi.org/10.1101/2020.12.02.20242685doi: medRxiv preprint 9 Liang et al. also assessed the impact of air pollution level and found increased air pollution to be 249 associated with an increased mortality rate 24. While Fielding-Miller et al. reported higher COVID-19 250 mortality for counties with higher non-English speaking populations, our study showed higher incidence 251 rates among this group 14. Also, in agreement with the current study, other studies have reported high 252 incidence and mortality rates in counties with greater proportions of racial minorities (Native 253 Americans, Hispanics, and African Americans) 12,15,25. Moreover, with an increase in population density, 254 incidence and mortality rates increased, which has been reported by Sun et al. Unlike this study, Mollalo 255 et al. spatial analysis found that increase in household income was associated with an increased 256 incidence rate 26. However, that same study reported that an increase in providers (for example nurse 257 practitioners) increased incidence rates, which was similar to our findings. Allcott et al. and Goolsbee et 258 al. also have pointed out ‘partisanship’ as a risk factor for non-adherence to preventive guidelines and 259 mask use. Our study also observed that higher Republican-leaning counties were associated with higher 260 incidence and mortality 27,28. Interestingly, similar to Sun et al., the current study also reported a lower 261 incidence among counties with higher proportions of people over 65 years of age 15. 262 The study has important limitations. First, log transforming the outcomes i.e. cumulative 263 incidence and mortality rates as a linear dependent variable may mask the variations across counties. To 264 our understanding, there is no software package currently available that runs a spatial lag model with a 265 dependent variable with Poisson or Binomial distribution and thus, this study transformed the outcomes 266 in their logarithmic form. Given the cross-sectional nature of the study, no causal inferences can be 267 made. There are considerable differences in the testing rates across regions and counties and can 268 influence the observed incidence rate. The list of variables is by no means comprehensive and does not 269 include several other factors such as mobility, local restriction policies (county or city-level), 270 compliance with local and federal prevention guidelines. The current analysis is ecological in nature and 271 no direct inferences can be drawn at the individual level. 272

273 The findings of this study are more insightful than the mere coronavirus count meters and data 274 visualizations that depict the spread of the COVID-19 pandemic. The current spatial models 275 incorporated a comprehensive list of factors to ensure that the results, when parsed, offer a multi-faceted 276 explanatory power. For illustration, these models helped identify factors including policy-related factors 277 (i.e. length of order), health behaviors (example: excessive drinking), clinical care (example: preventable 278 hospitalization rate), socio-economic factors (example: race/ethnicity, median income, uninsurance rate, 279 education), and physical environment (example: population density, air pollution-particulate matter, 280 . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 24, 2021. ; https://doi.org/10.1101/2020.12.02.20242685doi: medRxiv preprint 10 rurality, residential segregation among races, homeownership) as some of the important determinants of 281 the geographic disparities in COVID-19 incidence and mortality. As per prior research, this study 282 reaffirms that policy restrictions have helped to limit the COVID-19 incidence and mortality. This study 283 highlights the plausible effect of one’s residential location, vicinity, and local policymakers; and the 284 connectivity to the neighboring counties on the incidence and mortality of COVID-19. As the country is 285 facing the next wave of pandemic, the study findings have important policy implications and guidance 286 on identifying areas at greater risk of infection and mortality. As the pandemic gains momentum in the 287 rural areas, especially in the Midwest and South, contextualized policies at the local level that align with 288 state and federal policies will be necessary to contain the next wave. 289 290 . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 24, 2021. ; https://doi.org/10.1101/2020.12.02.20242685doi: medRxiv preprint 11 291

292 1. Cases in the U.S. CDC, https://www.cdc.gov/coronavirus/2019-ncov/cases-updates/cases-in-293 us.html (accessed 15 November 2020). 294 2. Map of COVID-19 case trends, restrictions and mobility. USA Today , 27 November 2020, 295 https://www.usatoday.com/storytelling/coronavirus-reopening-america-map/ (27 November 296 2020). 297 3. Yasmeen Wafai. What you can and can’t do under Washington’s newest coronavirus stay-home 298 restrictions. Seattle Times, 16 November 2020, https://www.seattletimes.com/life/what-you-can-299 and-cant-do-following-washingtons-newest-november-coronavirus-lockdown-restrictions/ (16 300 November 2020). 301 4. Matt Olberding. Hospitals to further restrict elective surgeries due to COVID-19. Lincoln Journal 302 Star, 13 November 2020, https://journalstar.com/lifestyles/health-med-fit/health/hospitals-to-303 further-restrict-elective-surgeries-due-to-covid-19/article_3e0b67ee-c06e-5967-aa16-304 25731057af40.html (13 November 2020). 305 5. Travel restrictions issued by states in response to the coronavirus (COVID-19) pandemic, 2020-306 2021 - Ballotpedia, 307 https://ballotpedia.org/Travel_restrictions_issued_by_states_in_response_to_the_coronavirus_(C308 OVID-19)_pandemic,_2020-309 2021#Travel_restrictions.2C_advisories.2C_and_recommendations_by_state (accessed 21 310 January 2021). 311 6. CHAPPELOW J. Coronavirus Aid, Relief, and Economic Security (CARES) Act. Investopedia . 312 7. Corum J, Wee S-L, Zimmer C. Covid-19 Vaccine Tracker: Latest Updates - The New York 313 Times. The New York Times, https://www.nytimes.com/interactive/2020/science/coronavirus-314 vaccine-tracker.html?auth=login-google (2020, accessed 27 November 2020). 315 8. Pfizer’s covid vaccine is more than 90 percent effective in first analysis, company reports - The 316 Washington Post, https://www.washingtonpost.com/health/2020/11/09/pfizer-coronavirus-317 vaccine-effective/ (accessed 27 November 2020). 318 9. Promising Interim Results from Clinical Trial of NIH-Moderna COVID-19 Vaccine | National 319 Institutes of Health (NIH), https://www.nih.gov/news-events/news-releases/promising-interim-320 results-clinical-trial-nih-moderna-covid-19-vaccine (accessed 27 November 2020). 321 10. Scipioni J. Fauci: Why still need masks, social distancing after Covid-19 vaccine. CNBC Make it, 322 . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 24, 2021. ; https://doi.org/10.1101/2020.12.02.20242685doi: medRxiv preprint 12 16 November 2020, https://www.cnbc.com/2020/11/16/fauci-why-still-need-masks-social-323 distancing-after-covid-19-vaccine.html (16 November 2020, accessed 27 November 2020). 324 11. Millett GA, Jones AT, Benkeser D, et al. Assessing differential impacts of COVID-19 on black 325 communities. Ann Epidemiol 2020; 47: 37–44. 326 12. Khanijahani A. County-Level Proportions of Black and Hispanic populations, and Socioeconomic 327 Characteristics in Association with Confirmed COVID-19 Cases and Deaths in the United States. 328 medRxiv 2020; 1–23. 329 13. Peters DJ. Community Susceptibility and Resiliency to COVID-19 Across the Rural-Urban 330 Continuum in the United States. J Rural Heal 2020; 36: 446–456. 331 14. Fielding-Miller RK, Sundaram ME, Brouwer K. Social determinants of COVID-19 mortality at 332 the county level. medRxiv Prepr Serv Heal Sci 2020; 2020.05.03.20089698. 333 15. Sun F, Matthews SA, Yang T-C, et al. A spatial analysis of the COVID-19 period prevalence in 334 U.S. counties through June 28, 2020: where geography matters? Ann Epidemiol. Epub ahead of 335 print 2020. DOI: 10.1016/j.annepidem.2020.07.014. 336 16. Andersen LM, Harden SR, Sugg MM, et al. Analyzing the spatial determinants of local Covid-19 337 transmission in the United States. Sci Total Environ; 754. Epub ahead of print 2021. DOI: 338 10.1016/j.scitotenv.2020.142396. 339 17. Fortaleza CMCB, Guimarães RB, De Almeida GB, et al. Taking the inner route: Spatial and 340 demographic factors affecting vulnerability to COVID-19 among 604 cities from inner São Paulo 341 State, Brazil. Epidemiol Infect. Epub ahead of print 2020. DOI: 10.1017/S095026882000134X. 342 18. The New York Times. Coronavirus (Covid-19) Data in the United States. Github. 343 19. HRSA Bureau of Health Workforce. Area Health Resources File. HRSA Data Warehouse. 344 20. MIT Election Data and Science Lab. County Presidential Election Returns 2000-2016. Harvard 345 Dataverse. 346 21. Ballotpedia. States that did not issue stay-at-home orders in response to the coronavirus (COVID-347 19) pandemic, 2020, https://ballotpedia.org/Status_of_lockdown_and_stay-at-348 home_orders_in_response_to_the_coronavirus_(COVID-19)_pandemic,_2020 (2020, accessed 349 15 November 2020). 350 22. USCB. Cartographic Boundary Files - Shapefile, https://www.census.gov/geographies/mapping-351 files/time-series/geo/carto-boundary-file.html (2019, accessed 27 October 2020). 352 23. Nayak A, Islam S, Mehta A, et al. Impact of Social Vulnerability on COVID-19 Incidence and 353 Outcomes in the United States. medRxiv Prepr Serv Heal Sci. Epub ahead of print 2020. DOI: 354 . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 24, 2021. ; https://doi.org/10.1101/2020.12.02.20242685doi: medRxiv preprint 13 10.1101/2020.04.10.20060962. 355 24. Fu L, Wang B, Yuan T, et al. Clinical characteristics of coronavirus disease 2019 (COVID-19) in 356 China: A systematic review and meta-analysis. J Infect 2020; 80: 656–665. 357 25. Rodriguez-Diaz CE, Guilamo-Ramos V, Mena L, et al. Risk for COVID-19 infection and death 358 among Latinos in the United States: examining heterogeneity in transmission dynamics. Ann 359 Epidemiol. Epub ahead of print 2020. DOI: 10.1016/j.annepidem.2020.07.007. 360 26. Mollalo A, Vahedi B, Rivera KM. GIS-based spatial modeling of COVID-19 incidence rate in the 361 continental United States. Sci Total Environ; 728. Epub ahead of print 2020. DOI: 362 10.1016/j.scitotenv.2020.138884. 363 27. Allcott H, Boxell L, Conway J, et al. Polarization and public health: Partisan Differences in 364 Social Distancing during the Coronavirus Pandemic. w26946, 2020. 365 28. Goolsbee A, Luo NB, Nesbitt R, et al. COVID-19 Lockdown Policies at the State and Local 366 Level. 2020–116, 2020. Epub ahead of print 2020. DOI: 10.2139/ssrn.3682144. 367 368 . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 24, 2021. ; https://doi.org/10.1101/2020.12.02.20242685doi: medRxiv preprint 1 Table 1: Descriptive statistics of factors used in this study, as of October 27, 2020 (n = 3,101) using AHRF, NYT, CPER and CHR 369 datasets. 370 371 Ecological Factors Mean SD Incidence rate in per 1 Million (logged) 10.07 0.53 Incidence rate in per 1 Million 26,575.75 12,216.59 Mortality rate in per 1 Million (logged) 6.18 0.80 Mortality rate in per 1 Million 631.59 477.64 COVID-19 Policy- related factors Length of order (quintles) [n, %]* n % 0 days (No order) 438.00 14.12 1 - 27 days 469.00 15.12 28 - 35 days 717.00 23.12 36 - 58 days 714.00 23.02 59 or more days 736.00 24.60 Calendar month of the first reported case in the state [n, %]* n % January 214.00 6.90 February 497.00 16.03 March 2,390.00 77.07 Health behaviors % Adult Smoking 15.36 3.46 % Physical Inactivity 23.25 5.37 % Excessive Drinking 18.76 2.79 % Influenza Vaccinations^ 46.27 6.91 Clinical Care Primary Care Physicians Rate (log) 4.23 0.51 Preventable Hospitalization Rate (log) 8.38 0.29 Hospital Beds (log) 6.45 3.29 Socio-economic factors % Republicans 46.63 16.77 Counties leaning Republican (>50% votes) [n, %] 2,500.00 80.62 . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 24, 2021. ; https://doi.org/10.1101/2020.12.02.20242685doi: medRxiv preprint 2 % Adult Obesity 29.23 5.45 Premature Age Adjusted Mortality# 343.74 92.96 Income Inequality Ratio 4.67 0.68 % Adults with Some College Education 65.28 9.79 % Unemployment 3.95 1.13 % Children in Single-parent Households 32.99 8.58 Social Association Rate 9.09 3.57 % Uninsured Adults 12.42 6.00 Median Income per $1,000 (log) 4.14 0.26 % 65 years old and over 16.05 4.08 % less than 18 years of age 22.49 2.91 % African American 12.30 12.56 % Native Americans 1.23 3.22 % Hispanic 18.06 17.19 % Female 50.72 1.22 % Not proficient in English 4.19 4.20 Physical environment Air pollution-particulate matter (2.5) 10.09 2.13 % Severe Housing Problem 17.72 5.41 Residential Segregation- African Americans/White 50.67 12.82 Residential Segregation- non-White/White 36.73 11.25 % Rural 19.18 24.18 % Homeownership 64.57 9.54 Population density per 100 sq. miles (log) 6.12 1.55 Abbreviations: AHRF: 2018-2019 Area Health Resource Files; 2020 NYT: New York Times; 2020 CHR: County Health Rankings; and CPER: 2016 County Presidential 372 Election Returns; SD: Standard Deviation. 373 * Unweighted as this represents the number of counties; ^: Among Medicare enrollees; #: Number of deaths among residents under age 75 per 100,000 population (age-374 adjusted). 375 376 . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 24, 2021. ; https://doi.org/10.1101/2020.12.02.20242685doi: medRxiv preprint 3 Table 2: SAC model for the incidence rates (logged) and Spatial Lag Model for mortality rates (logged) as of October 27, 2020 using 377 NYT, AHRF, CPER and CHR datasets. 378 Ecological Factors Incidence Rates Mortality Rates Estimate SE Estimate SE COVID-19 Policy- related factors Length of order (quintles) 0 days (No order) Reference Reference 1 - 27 days -0.290 *** 0.079 -0.483 0.257 28 - 35 days -0.135 0.070 -0.824 *** 0.218 36 - 58 days -0.228 ** 0.071 -0.866 *** 0.206 59 or more days -0.335 *** 0.081 -1.006 *** 0.229 Calendar month of the first reported case in the state January Reference Reference February -0.056 0.092 -0.628 * 0.297 March 0.054 0.083 -0.249 0.274 Health behaviors % Adult Smoking -0.003 0.005 0.036 0.033 % Physical Inactivity 0.002 0.002 0.017 0.014 % Excessive Drinking 0.000 0.005 0.107 *** 0.026 % Influenza Vaccinations^ 0.003 *** 0.001 0.039 *** 0.007 Clinical Care Primary Care Physicians Rate (log) 0.011 ** 0.004 0.234 *** 0.031 Preventable Hospitalization Rate (log) 0.045 * 0.022 0.183 0.166 Hospital Beds (log) 0.004 ** 0.001 0.033 ** 0.011 Socio-economic factors % Republicans 0.002 * 0.001 0.017 * 0.007 % Adult Obesity -0.001 0.002 -0.007 0.013 Premature Age Adjusted Mortality# 0.000 0.000 -0.001 0.001 Income Inequality Ratio 0.018 0.011 0.120 0.094 % Adults with Some College Education -0.004 *** 0.001 -0.041 *** 0.008 % Unemployment -0.014 0.008 -0.019 0.055 % Children in Single-parent Households -0.003 * 0.001 0.018 * 0.009 Social Association Rate 0.003 * 0.001 0.011 0.010 % Uninsured Adults 0.001 0.003 0.038 * 0.019 Median Income per $1,000 (log) -0.158 * 0.078 -0.393 0.509 . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 24, 2021. ; https://doi.org/10.1101/2020.12.02.20242685doi: medRxiv preprint 4 % 65 years old and over -0.020 *** 0.003 -0.043 0.025 % less than 18 years of age -0.011 ** 0.004 -0.025 0.030 % African American 0.007 *** 0.002 0.032 *** 0.008 % Native Americans 0.008 *** 0.002 0.034 ** 0.013 % Hispanic 0.015 *** 0.002 0.037 *** 0.009 % Female -0.007 0.004 0.101 ** 0.034 % Not proficient in English 0.010 * 0.005 -0.019 0.038 Physical environment Air pollution-particulate matter (2.5) 0.062 *** 0.012 0.130 ** 0.046 % Severe Housing Problem 0.004 0.003 0.009 0.021 Residential Segregation- African Americans/White -0.002 ** 0.001 -0.010 * 0.005 Residential Segregation- non-White/White 0.003 *** 0.001 0.014 * 0.006 % Rural -0.001 ** 0.000 -0.014 *** 0.003 % Homeownership 0.002 0.001 0.051 *** 0.011 Population density per 100 sq. miles (log) 0.028 * 0.012 0.270 *** 0.077 /i1 (spatial lag parameter) -0.563 *** 0.179 *** λ (spatial error parameter) 0.936 *** AIC 2993.60 15382.00 Observed Moran’s I for residuals -0.032 -0.006 379 Abbreviations: SAC: Spatial Autoregressive Combined; AHRF: 2018-2019 Area Health Resource Files; 2020 NYT: New York Times; CHR: 2020 County Health 380 Rankings; 2016 CPER: County Presidential Elections Results. 381 ^: Among Medicare enrollees; #: Number of deaths among residents under age 75 per 100,000 population (age-adjusted). 382 Level of significance: *p-value < 0.05, **p-value < 0.01, ***p-value <0.001. 383 384 385 . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 24, 2021. ; https://doi.org/10.1101/2020.12.02.20242685doi: medRxiv preprint . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 24, 2021. ; https://doi.org/10.1101/2020.12.02.20242685doi: medRxiv preprint . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 24, 2021. ; https://doi.org/10.1101/2020.12.02.20242685doi: medRxiv preprint . CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 24, 2021. ; https://doi.org/10.1101/2020.12.02.20242685doi: medRxiv preprint