key: cord-0708702-ctgiw0qm authors: Rodriguez-Diaz, Carlos E.; Guilamo-Ramos, Vincent; Mena, Leandro; Hall, Eric; Honermann, Brian; Crowley, Jeffrey S.; Baral, Stefan; Prado, Guillermo J.; Marzan-Rodriguez, Melissa; Beyrer, Chris; Sullivan, Patrick S.; Millett, Gregorio A. title: Risk for COVID-19 infection and death among Latinos in the United States: Examining heterogeneity in transmission dynamics date: 2020-07-23 journal: Ann Epidemiol DOI: 10.1016/j.annepidem.2020.07.007 sha: 9707ade30d8409bc6acd79e93f4ac0eebb13a748 doc_id: 708702 cord_uid: ctgiw0qm OBJECTIVES: Ascertain COVID-19 transmission dynamics among Latino communities nationally. METHODS: We compared predictors of COVID-19 cases and deaths between disproportionally Latino counties (>17.8% Latino population) and all other counties through May 11, 2020. Adjusted Rate Ratios were estimated using COVID-19 cases and deaths via zero-inflated binomial regression models. RESULTS: COVID-19 diagnoses rates were greater in Latino counties nationally (90.9 vs. 82.0 per 100,000). In multivariable analysis, COVID-19 cases were greater in Northeastern and Midwestern Latino counties (aRR 1.42, 95% CI 1.11–1.84 and aRR 1.70, 95% CI 1.57–1.85, respectively). COVID-19 deaths were greater in Midwestern Latino counties (aRR, 1.17, 95% CI 1.04-1.34). COVID-19 diagnoses were associated with counties with greater monolingual Spanish speakers, employment rates, heart disease deaths, less social distancing, and days since the first reported case. COVID-19 deaths were associated with household occupancy density, air pollution, employment, days since the first reported case, and age (fewer <35yo). CONCLUSIONS: COVID-19 risks and deaths among Latino populations differ by region. Structural factors place Latino populations and particularly monolingual Spanish speakers at elevated risk for COVID-19 acquisition. The novel coronavirus disease 2019 (COVID-19) pandemic has dramatically impacted the health and well-being of millions of individuals residing in the United States (U.S.) and around the world. As of June 14, 2020, over two million people in the U. S. have been diagnosed with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection and over 114,000 individuals have died from COVID-19-related complications. 1 Much of the discourse regarding disease burden has acknowledged the disparate COVID-19 diagnoses and deaths among differing populations, including ethnic and racial minorities. 2, 3 This is particularly evident in the case of Latinos who comprise approximately 18% of the U.S. population in the U.S. and represent the nation's largest ethnic and racial minority group. 4 Latinos have been disproportionately impacted by COVID-19; 5 as of June 8, 2020, they account for one in three (33%) of all confirmed COVID-19 cases in the U.S. with known ethnicity/race data, while half (52%) of the COVID-19 cases assessed still lacked race or ethnicity data. 1 Furthermore, Latinos have among the highest age-adjusted rates of COVID-19 associated hospitalizations at 117 per 100,000 and account for approximately one in five of all confirmed COVID-19-related deaths in the U.S. with known ethnicity/race data. 6, 7 Research designed to identify the underlying drivers of SARS-CoV-2 transmission and death among the country's vast and diverse Latino populations is urgently needed. A recent epidemiological study provided the first national assessment of the disparity in COVID-19 diagnoses and deaths among Black Americans. 8 To overcome incomplete reporting of COVID-19 diagnoses and deaths by race/ethnicity, the study examined counties with a share of Black Americans above the U.S. average and compared these counties to all other U.S. counties. This was shown to be a useful proxy for examining case report data and offered real-time guidance for enhancing the public health response to COVID-19 among Black Americans. In the present study, we applied the same county-level approach to better understand the patterns in COVID-19 diagnoses and deaths among Latino populations nationally, including Puerto Rico. We considered the heterogeneity of the US Latino population and explored whether differences in a set of core demographic variables attenuate drivers of COVID-19 diagnoses and deaths. We assessed the differences in county-level characteristics of counties with a greater share of Latino residents than the U.S. average (>17.8% Latino population; hereafter, disproportionately Latino counties) compared to all other counties (<17.8% Latino population). Additionally, we examined the association between the proportion of Latino residents and COVID-19 cases and deaths. All data used in these analyses were from publicly available datasets. Puerto Rico was included as a single additional "county" in our analysis because Puerto Rico does not have counties as administrative units. Additionally, we assessed intra-regional differences based on U.S. Census statistical regions: Midwest, Northeast, South, and West. 9 Puerto Rico was excluded from the regional assessments. Demographic data: County-level data from the Census Bureau American Community Survey 5-Year 10 were collected for select demographics (county population, percent of Latinos, percent of the population over the age of 65, percent of the population under the age of 35, percent of the under-65 population without health insurance, occupants per room, and language spoken at home and ability to speak English). For language spoken at home, the percentages were calculated based on monolingual English, monolingual Spanish (speak Spanish at home and ability to speak English "not well" or "not at all"), and dual-lingual (speak Spanish at home and ability to speak English "well" or "very well"). The annual average county unemployment rates were obtained from the Bureau of Labor Statistics. 11 COVID-19 data and comorbidities: COVID-19 cases and deaths at the county-level were downloaded from USAFacts through May 11. While more recent data were available, we chose May 11, 2020 as a cut-off date as it precedes many states' and localities' re-opening of their economies, which could introduce additional bias such as increased exposure due to occupational risk or daily life activities. The rates of diagnosed diabetes among adults aged 20+ were downloaded from the Centers for Disease Control and Prevention (CDC) Diabetes Atlas (2016). 12 The heart disease deaths per 100,000 were accessed from CDC's Interactive Atlas of Heart Disease and Stroke for those 35 years of age and older . 13 The combined rates of cerebrovascular and hypertension deaths per 100,000 were sourced from the CDC WONDER (2018). 14 The estimates of people living with diagnosed HIV per 100,000 among adults and adolescents 13 and older were derived from the CDC ATLAS (2018). 15 For counties with missing HIV data in Kentucky and Alaska, archived HIV data from AIDSVu.org 16 from 2015 were used, given that a state data-sharing agreement between these two states restricts the CDC from releasing comparable data. Social/environmental data: Following Wu, 17 we used county-level estimates of fine particulate matter (PM 2.5 ) to assess air quality. Social distancing grades were drawn from Unacast's county measures on May 22 and were coded as A+/-= 1, B+/-= 2, C+/-=3, D+/-=4, and F+/-=5. 18 Higher scores were associated with poorer social distancing. The U.S. Department of Agriculture Urban-Rural Classification Scheme was used to assess urbanicity (index from 1-6, with 1 being the most urban). 19 Puerto Rico was assigned an urbanicity score based on the weighted average of all municipalities. First, we categorized counties into two groups: disproportionately Latino counties (>17.8% of county population) and other counties (<17.8% of county population). We summarized the characteristics of counties in each group using medians and interquartile ranges. Additionally, we calculated and mapped county-level COVID-19 diagnosis and death rates to assess geographic patterns for each group. Next, we analyzed the proportion of Latino residents in a county as a continuous variable and plotted the proportion of Latino residents by the rate of COVID-19 diagnoses, adjusted for days since the first infection was reported for each U.S. Census region. To assess whether the observed associations between the proportion of Latino residents and COVID-19 cases and deaths were confounded by other factors, we conducted a multivariate analysis using Bayesian hierarchical models. We fit zeroinflated negative binomial models with logarithmic link functions separately to COVID-19 cases and deaths using integrated nested Laplace approximations. 20 All models included the proportion of the population that is Latino as a continuous variable, the county population as an offset and included a spatially structured state-level random effect. 21 We first estimated an unadjusted rate ratio for the proportion of the population that is Latino before fitting a model that only controlled for the proportion of population that is younger than 35. Finally, we present a fully-adjusted model that includes all other county-level characteristics (see Table 1 for the characteristics) and an interaction term for the region by proportion of Latino residents to estimate the adjusted rate ratios for each region. To control for potential temporal confounding, each model also included a variable representing the number of days since the first case of COVID-19 was reported in each county. We presented the adjusted rate ratios of COVID-19 cases and deaths, comparing the observed third quartile to the observed first quartile of each variable. Rate ratios greater than one indicate that the higher levels of a given characteristic were associated with higher rates of COVID-19 cases or deaths. All analyses were performed using R 3.5.0. and Bayesian hierarchical models were fit using the inla package (www.r-inla.org). (Table 1) . Across regions, disproportionately Latino counties comprised people younger than those in other counties, were more likely to lack health insurance, had a greater number of persons per room in a household, had fewer monolingual English-speaking Latino residents, and had a greater proportion of monolingual Spanish-speaking or bilingual (Table 1 ; disaggregation by region in Table 1 can be found in Supplemental Table 1 ). There were no major differences in the underlying health conditions or social distancing scores and varying associations by region for other variables. In Table 2 To our knowledge, this is the first nationwide analysis of Latino-specific COVID-19 diagnoses and deaths in the U.S., including Puerto Rico. We found that the COVID- outbreaks. 25 Various reports have found that a large number of Latino migrants are hired to work in the meatpacking industry and that COVID-19 outbreaks and deaths have been rising in meatpacking plants. 26 Exercising his authority under the Defense Production Act, in April 2020, President Trump issued an executive order that designated meatpacking workers as "essential workers", thereby forcing many plants to re-open despite the ongoing transmission of COVID-19 in the plants and surrounding communities. 27 Since that time, there have been documented increases in COVID-19 deaths in meatpacking plants; those deaths may be reflected in our data. Our findings suggest that occupational exposure plays an important role in Latino COVID-19 cases and deaths. Up to 59% of the Latino community in the U.S. is under the age of 35 compared to 46% of the total population. 28 Relative to other racial and ethnic groups, Latinos are disproportionately young and overrepresented in a segment of the workforce that experiences ongoing occupational exposure. In addition to meatpacking plants, majority of the Latino community is employed in factories or in other service-based industries that have been deemed "essential" in recent months. 25, 26 Furthermore, counties with a high proportion of monolingual Spanish speakers were more likely to have higher rates of COVID-19 cases (compared to counties with fewer monolingual Spanish speakers), but not more likely to have increased COVID-19 death rates. Again, monolingual Spanish speakers are more likely to be occupationally exposed through involvement in factory or "service industry" jobs, but they are also more likely to be healthier, younger, and undocumented, thereby providing insight into the observed rates of greater Latino COVID-19 cases and no deaths in this analysis. 24, 26, 29 COVID-19 testing access for Latinos warrants mention. In general, COVID-19 testing has prioritized older populations with clinical indications for testing or those with underlying health conditions. 30, 31 Migrants, who are likely to be monolingual, may be less likely to get tested. Those who are tested are likely to be sick and, thus, have clinical indications for testing. Historically, Latinos have experienced patterns of decreased access to health care due to lack of insurance, inadequate insurance coverage, and fear of deportation. 32 , 33 Furthermore, recent policies that have deterred immigrants from receiving federal safety net assistance may influence decisions to seek testing or care even when needed. 34 Given this, it is likely that immigrants would have more undiagnosed cases than other groups and would be under-reported in official case counts. Additional barriers to testing among Latinos may include provider bias and less access to culturally appropriate healthcare services. 35, 36 As with the previous analysis of COVID-19 among Black Americans, the social determinants of health such as housing, employment and working conditions were better predictors of Latino COVID-19 cases and deaths than underlying health conditions (except heart disease). Disproportionately Latino counties had a greater proportion of individuals residing in densely populated households, which was associated with COVID-19 cases. Employment was slightly lower in disproportionally Latino counties when compared to other counties, which was associated with lower COVID-19 cases and death rates after controlling for other variables. These findings might be ascribed to the fact that most Latinos continued working during this period, were not working from home, and many jobs held by Latino populations place them at greater risk exposure to COVID-19. 37 Taken together, these findings suggest that a sole focus on individual-level factors such as risk behaviors is inadequate for implementing an effective national public health response to COVID-19 prevention and control within Latino communities. Our analyses have limitations. First, as valuable as our analyses are in highlighting communities that are at a higher risk for COVID-19 acquisition and death, our study is an ecological analysis. Associations at the county level do not necessarily reflect associations at the individual level. The COVID-19 analytical methods we used, however, have been rigorously reviewed and published in another COVID-19 analysis 8 and others have used county-level data to assess COVID-19 impacts in racial minority communities. 38, 39, 40 Another limitation is the COVID-19 data itself. Some have criticized existing COVID-19 surveillance data as inaccurate, but these same publicly available data are also used by the CDC to compile COVID-19 reports and issue public health recommendations. 41 Our analyses likely underestimated COVID-19 cases in disproportionately Latino counties because Latino communities, particularly those in low-income areas, are less likely to have access to COVID-19 testing, 42, 43 and some Latinos fear accessing healthcare due to their immigration status. Moreover, the COVID-19 surveillance systems have inadequately captured race and ethnicity data which is important to describe the heterogeneity in the Latino population. Third, due to data limitations, we included Puerto Rico as one county, which constrained our ability to characterize the nuances in the island's epidemic. Fourth, our results were subject to a specific date and time, and consequently, statistically significant associations may weaken or strengthen as the pandemic grows. Table 1 : Distribution of the demographics, chronic diseases, social/environmental factors, and COVID-19 diagnoses and deaths in disproportionally Latino counties (>17.8% of the total population) and in all other counties (<17.5% of the total population). COVID-19) in the U.S. Centers for Disease Control and Prevention COVID-19 and Racial/Ethnic Disparities Disparities in the Population at Risk of Severe Illness from COVID-19 by Race/Ethnicity and Income American Community Survey 1-Year Estimates Assessing racial and ethnic disparities using a COVID-19 outcomes continuum for New York State COVIDView: A Weekly Surveillance Summary of U.S. COVID-19 Activity Assessing Differential Impacts of COVID-19 on Black Communities Census regions and divisions of the United States American Community Survey 5-Year United States Census Bureau Centers for Disease Control and Prevention. Diabetes Atlas Centers for Disease Control and Prevention. Interactive Atlas of Heart Disease and Stroke Data Visualization Promotes Sound Public Health Practice: The AIDSvu Example Exposure to air pollution and COVID-19 mortality in the United States: A nationwide cross-sectional study. medRxiv Social Distancing Scoreboard--Counties Approximate Bayesian inference for latent Gaussian models by using integrated nested Laplace approximations Bayesian image restoration, with two applications in spatial statistics Portrait of Unauthorized Immigrants in the United States Health care use among undocumented Latino immigrants Use of preventive care services among Latino subgroups COVID-19 Among Workers in Meat and Poultry Processing Facilities -19 States Update: COVID-19 Among Workers in Meat and Poultry Processing Facilities -United States Executive Order on Delegating Authority Under the DPA with Respect to Food Supply Chain Resources During the National Emergency Caused by the Outbreak of COVID-19. The White House American Community Survey 5-Year United States Census Bureau Racial Capitalism within Public Health: How Occupational Settings Drive COVID-19 Disparities Coronavirus Philadelphia: Positive Tests Higher In Poorer Neighborhoods Despite Six Times More Testing In Higher-Income Neighborhoods Spatial Inequities in COVID-19 outcomes in Three US Cities. medRxiv. Published online The Health of Undocumented Latinx Immigrants: What We Know and Future Directions The Impact of Local Immigration Enforcement Policies on the Health of Immigrant Hispanics/Latinos in the United States Policies of Exclusion: Implications for the Health of Immigrants and Their Children Implicit Racial/Ethnic Bias Among Health Care Professionals and Its Influence on Health Care Outcomes: A Systematic Review Hispanic health in the USA: a scoping review of the literature Job flexibilities and work schedules -2017-2018. Data from the American Time Use Survey Coronavirus is Decimating Racially Diverse Communities Large and Small Black Americans face disproportionate share of disruption from coronavirus COVID-19 More Prevalent, Deadlier in U.S. Counties with Higher Black Populations Covid-19: Testing Inequality Large Texas Cities, Access to Coronavirus Testing May Depend On Where You Live. NPR.org. Accessed Each author made substantial contributions to all phases of article development and revision; contributed to study, analysis, and article plans; interpreted data; and approved the final version prior to submission Authors have no conflict of interest to disclose.