key: cord-0992051-spdpv9lf
authors: Riley, A. R.; Chen, Y.-H.; Matthay, E. C.; Glymour, M. M.; Torres, J. M.; Fernandez, A.; Bibbins-Domingo, K.
title: Excess deaths among Latino people in California during the COVID-19 pandemic
date: 2020-12-19
journal: medRxiv : the preprint server for health sciences
DOI: 10.1101/2020.12.18.20248434
sha: 56b98b2e975321c1730f215ca2c6a261021ab47c
doc_id: 992051
cord_uid: spdpv9lf

Background: Latinos in the US are experiencing higher excess deaths during the COVID-19 pandemic than any other racial/ethnic group, but it is unclear which subgroups within this diverse population are most affected. Such information is necessary to target policies and programs that prevent further excess mortality and reduce inequities. Methods: Using death certificate data for January 1, 2016 through February 29, 2020 and time-series models, we estimated the expected weekly deaths among Latinos in California from March 1 through October 3, 2020. We quantified excess mortality as observed minus expected deaths and risk ratios (RR) as the ratio of observed to expected deaths. We considered subgroups defined by age, sex, place of birth, education, occupation, and combinations of these factors. Results: During the first seven months of the pandemic, Latino deaths in California exceeded expected deaths by 10,304, a 31% increase. Excess death rates were greatest for individuals born in Mexico (RR 1.44; 95% PI, 1.41, 1.48) or Central America (RR 1.49; 95% PI, 1.37, 1.63), with less than a high school degree (RR 1.41; 95% PI, 1.35, 1.46), or in manufacturing (RR 1.62; 95% PI, 1.52, 1.72) or food-and-agriculture occupations (RR 1.50; 95% PI, 1.40, 1.61). Disparities in excess death by place of birth and education were magnified among Latinos in essential occupations. Conclusions: Foreign-birth, low formal education, and work in an essential sector put some Latinos in California at higher risk of death during the COVID-19 pandemic. Interventions should reduce the disproportionate impact of the pandemic on Latino immigrants and Latinos in unsafe working conditions; these may include early vaccination, workplace safety enforcement, and expanded access to medical care.

Throughout the coronavirus disease 2019 (COVID- 19) pandemic, Latinos in the US have been dying at disproportionately high rates. National all-cause mortality increased 53.6% for Latinos from the start of the pandemic through October -more than any other group and over four times more than the percent increase for non-Latino Whites 1 . However, aggregate national and state-level statistics obscure the vast diversity among Latinos in the US, making it difficult to intervene effectively to prevent further pandemic mortality 2, 3 . Identifying subgroups at greatest risk for excess death could help guide targeted interventions including workplace protections, testing, and vaccination.

California is a critical setting to evaluate the drivers of Latino mortality during the COVID-19 pandemic. One quarter of U.S. Latinos live in California. Latinos are 39% of California's population of 39.5 million residents, but they account for 48% of COVID-19-confirmed deaths through December 2020 4 . Even this remarkable death toll likely understates the impact of the pandemic on Latinos because COVID-19-confirmed deaths are understood to be a subset of the total deaths due to COVID-19 5 .

The current study estimates the impact of the COVID-19 pandemic on deaths among Latino individuals in California by estimating excess deaths that occurred between March 1, 2020 and October 3, 2020 compared to the four years prior to the COVID-19 pandemic. By examining death record data, this study overcomes the limitations of official COVID-19 death counts 6 . In addition, with the overarching goal of assisting public health efforts, we evaluate heterogeneous risk within the Latino population by examining excess death during the pandemic by age, sex, place of birth, education, and occupation.

In order to forecast expected mortality trends during our time-period of interest, we drew on death certificate data from the California Department of Public Health for the period from January 3, 2016 to February 29, 2020 (pre-pandemic period) and from March 1, 2020 to October 3, 2020 (pandemic period). Our analyses included all deaths that occurred within the state of California for which Hispanic/Latino ethnicity was designated. (Only 0.29% of deaths during the pandemic period and 0.25% of deaths during the pre-pandemic period lacked information on ethnicity). Our analytic sample All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted December 19, 2020. ;  https://doi.org/10.1101/2020. 12.18.20248434 doi: medRxiv preprint consisted of 43,564 total deaths among Latinos in California during the pandemic period and 220,984 during the pre-pandemic period. There was little missingness for the variables of interest and exclusions due to missingness were done on an available-case basis. Due to delays in complete cause of death data, we did not distinguish specific causes of death for this analysis.

We defined Latinos as any individual identified with Latino/a or Hispanic ethnicity on the death certificate, irrespective of race or country of birth. We do not use the more inclusive, gender-neutral terms Latinx or Latine in this paper because they were not used in death records.

We evaluated subgroups defined by: age at death in years (0-24, 25-54, 55-64, 65-74, 75-84, or 85+); sex (male or female); place of birth (4-categories: U.S., Mexico, Central America, or other; 2-categories: U.S., or foreign-born); occupational sector (see below); educational attainment (4-categories: no high school degree and no GED, high school degree or GED, some college or Associate's degree, Bachelor's degree or beyond; 2-categories: no high school degree and no GED, or high school degree or more).

We defined excess deaths as observed minus expected deaths due to all causes, with expected deaths based on historical trends over the previous four years using time series methods described below.

For analyses stratified by educational attainment, following standard practice, we restricted our analysis to decedents at least 25 years of age so most people would have completed schooling. For the analyses stratified by occupational sector, we grouped individuals by the primary occupation over their life, a field included in the death certificate, and restricted analyses to decedents 18 to 65 years of age. We drew on the California Department of Public Health designations for essential work 7 to sort the occupation into one of seven sectors of high-risk or essential work, a category for non-essential and non-retail work, or a category for unemployed or missing data on occupation. Using this designation, we also compared individuals in essential occupational sectors to individuals in non-essential occupational sectors.

For each subgroup of interest, we repeated the following procedure using the forecast package in R 8 .

First, we aggregated the data to weekly death counts. Next, to estimate expected deaths, we fit dynamic All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted December 19, 2020. ; https://doi.org/10.1101/2020.12.18.20248434 doi: medRxiv preprint harmonic regression models with autoregressive integrated moving average (ARIMA) 9 errors for the number of weekly deaths, using deaths occurring pre-pandemic. We iterated through models with up to 25 Fourier terms and selected the model that minimized the corrected Akaike information criterion, following an approach developed by Hyndman and Khandakar 8 . We inspected the residuals from the preferred model using plots and the Ljung-Box test to confirm that minimal autocorrelation remained.

Then, using the preferred model, we forecast the number of weekly deaths and the corresponding 95% prediction intervals (PI) for the pandemic period: March 1, 2020 through October 3, 2020. These forecasts provided the expected weekly deaths during the pandemic period based on historical trends.

To estimate the weekly excess deaths, we subtracted the number of expected deaths from the observed deaths for each week. We constructed a 95% PI for excess deaths by simulating the expected deaths model 10,000 times (varying based on uncertainty in the estimated coefficients for the time-series prediction model), selecting the 2.5 and 97.5 percentiles, and subtracting the total number of observed deaths. Risk ratios were estimated as observed deaths divided by expected deaths, with the pandemic as the exposure, and are also known as observed-to-expected mortality ratios. Risk ratios present the same information as percent excess deaths, as reported in previous studies 1, 10, 9 . We also calculated per capita excess mortality by dividing excess deaths by the corresponding population size, using estimates from the 2019 American Community Survey. We conducted all analyses in R version 3.6.3.

Between March 1 and October 3, 2020, Latino residents in California experienced a 31% increase in mortality compared with historical trends, with an estimated 10,304 (95% PI, 9,558, 11,018) excess deaths (Table 1) . COVID-19 was listed as a cause of death on 7672 death certificates (74% of the estimated excess deaths). The number of excess deaths varied by week such that excess mortality was lowest during California's shelter-in-place period (March 19, 2020-May 7, 2020), then increased to a peak at the end of July with over 600 excess deaths per week, and began to decline by mid-August ( Figure 2 ), but the time trends varied by subpopulation (Figures 3 and 4) . The observed excess death rate for Latinos aged 25 or older implies 1106 additional deaths per million individuals. Excess deaths in Latinos were observed in all regions in California ( Figure S1 ).

As shown in Table 1 , relative increases in mortality ranged from 26-42% for all adults 25 and older and was of similar magnitude across adult age groups. The magnitude of excess mortality varied by country of birth and education. Foreign-born Latinos had higher relative excess mortality than U.S.-All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. Figure 3 ).

Disadvantages in excess mortality associated with foreign-birth and low education were more pronounced among Latinos in essential occupations and in the 55 to 74 age group. Among Latinos who worked in essential sectors, the percent excess in mortality for immigrants (RR 1.57; 95% PI, 1.47, 1.69) was more than double the percent excess for U.S.-born Latinos (RR 1.23; 95% PI, 1.16, 1.31) ( Figure 1 ). This 34-percentage-point immigrant disadvantage was reduced to 18-percentagepoints among Latinos in non-essential occupations. Among Latinos in essential occupational sectors who lacked a high school credential, mortality was 51% (95% PI, 43%, 59%) higher than expected, whereas for Latinos in the same occupational sectors with a high school credential or more, mortality was 32% (95% PI, 23%, 42%) higher than expected. The elevated risk of excess death associated with low education nearly disappeared among Latinos in non-essential occupations. Finally, pandemic mortality was 46 percent (RR 1.46; 95% PI, 1.40, 1.52) above expected in the subgroup of Latinos who were both foreign-born and without a high school degree, and only slightly lower in the subgroup of Latinos who were foreign-born with a high school degree or more (RR 1.38; 95% PI, 1.29, 1.48) (Table S1 ).

Using data on all deaths among Latinos in California from January 1, 2016 through October 3, 2020, All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted December 19, 2020. ; https://doi.org/10.1101/2020.12.18.20248434 doi: medRxiv preprint we found 10,304 additional deaths occurred among California Latinos in the first 7 months of the pandemic compared with trends in the four years prior. In relative terms, this was an excess of 31 percent. Our findings reveal disparities in excess mortality by Latino subgroups. Immigrant Latinos, those with less than a high school degree, and those working in essential occupations, in particular manufacturing and food/agriculture, were at markedly higher risk of death, especially in the presence of more than one of these risk factors. Participation in an essential occupational sector appeared to magnify existing social vulnerabilities.

We have previously shown that over the first 5 months of the pandemic, Latinos in California experienced higher excess mortality than other racial/ethnic groups, with a per capita excess nearly double that of non-Latino White people in California 11 . Our current results show that the pandemic's impacts on mortality have been far worse for Central American-born and Mexican-born Latinos than U.S.-born Latinos; a disparity apparent within every age stratum, but worse among 55 to 74 year-olds.

Latino immigrants comprise 37% of California's Latino population, but an estimated 72 percent of Latino excess deaths during the pandemic period occurred among immigrants.

The differential impact on foreign-born Latinos is notable given previously documented immigrant health advantages 12 . The literature suggests recent immigrant Latinos are less likely than their U.S.born counterparts to have an underlying medical condition that would increase COVID-19 severity 13 .

However, foreign-born Latinos may be more vulnerable to COVID-19 because of their concentration in high-exposure occupations with limited workplace protections 14 , barriers to health care and other social protections due to legal status and public charge inadmissibility 15 , and chilling effects on healthcare-seeking behavior due to anti-immigrant rhetoric and policy 16 .

The elevated mortality and stark disparities observed in essential occupational sectors underscores the potential role of workplace settings and worker marginalization as contributors to excess deaths among Latino individuals 17, 18 . The risk ratios for excess death among those in health and emergency jobs or government and community jobs -where workplace safety measures are more consistently adopted 19 --were substantially lower than the risk ratios among those in sectors characterized by lowwage jobs such as manufacturing and agriculture. Notably, the foreign-born disadvantage and the loweducation disadvantage in pandemic mortality were both magnified among Latinos in essential occupational sectors (Figure 1 ). For example, although pandemic mortality was 46 percent higher than All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted December 19, 2020. ; https://doi.org/10.1101/2020.12.18.20248434 doi: medRxiv preprint usual in the foreign-born and low education subgroup, it was 51 percent higher than usual in those with both low education and essential occupation, and 57 percent higher than usual in those who were both foreign-born and in the essential occupation subgroup. Within these intersecting factors are many opportunities for intervention.

The two main strategies for pandemic control that have been endorsed thus far are individual behavior change (e.g., masking, distancing) and economic shutdowns to slow transmission. Our results suggest that neither pandemic control strategy has sufficiently protected Latinos from excess mortality. In fact, as our previous study demonstrated and our current study reinforces, California began reopening in May when excess deaths among Latinos had yet to decline and, as we show here, when excess death were rising for vulnerable subgroups. Distinct policies and investment are needed to protect immigrant Latinos and reduce disparities in COVID-19-related death 26-28 .

Our findings draw attention to four domains of promising policy response: workplace conditions, financial supports, healthcare coverage, and vaccine distribution. First, expanded action is needed, particularly in the manufacturing and food & agriculture sectors, to regulate and enforce workplace modifications to protect employees from SARS-CoV-2 exposure, provide workers with accurate All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted December 19, 2020. This study is limited by the accuracy and comprehensiveness of the data on death certificates. The interpretation of excess deaths that did not list COVID-19 as a contributing cause is unclear; these deaths may be undiagnosed COVID-19 cases or they may reflect increased mortality due to other causes exacerbated by the pandemic. The proportion of COVID-confirmed deaths compared with overall excess mortality that we observed (approximately 74%) is similar or slightly higher than in national studies. Our analysis is also limited by the occupation field on death certificates which indicates primary occupation for most of the decedent's lifetime rather than most recent occupation.

We did not have data to directly examine a number of plausible contributors to excess mortality among Latinos, for example household size or legal status. Nor did we have access to data on indigenous ethnicity, another potentially important subgroup with high vulnerability 32 . Finally, our results rest on the time-series model premise, that we can predict mortality rates in 2020 based on trends observed in the prior four years, accounting for factors such as seasonality and autocorrelation.

In summary, our study used data from individual death records to reveal that immigrant Latinos, Latinos with low education, and Latinos who participate in essential work have experienced particularly high excess mortality during the COVID-19 pandemic. Policies enforcing occupational safety, especially for immigrant and low-educated workers, and policies extending high-quality healthcare to all people, including immigrants, may have the highest benefit for reducing COVID-19 mortality.

All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted December 19, 2020. Risk ratio (observed/expected) for death during the pandemic * ± All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted December 19, 2020. ; https://doi.org/10.1101/2020.12.18.20248434 doi: medRxiv preprint

Note: Risk ratio is observed/expected for all-cause mortality, comparing the pandemic period (March 1, 2020 to October 3, 2020) to previous years (January 3, 2016 to February 29, 2020); Source: California Department of Public Health Death Records. Risk ratio All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted December 19, 2020. ; https://doi.org/10.1101/2020.12.18.20248434 doi: medRxiv preprint 

All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted December 19, 2020. ; https://doi.org/10.1101/2020.12.18.20248434 doi: medRxiv preprint Excess deaths per week All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted December 19, 2020. ; https://doi.org/10.1101/2020.12.18.20248434 doi: medRxiv preprint All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted December 19, 2020. ; https://doi.org/10.1101/2020.12.18.20248434 doi: medRxiv preprint

Excess Deaths Associated with COVID-19, by Age and Race and Ethnicity -United States

Structural racism is shaping the Latino COVID-19 experience in San Diego County

Lessons We've Learned -Covid-19 and the Undocumented Latinx Community

All Cases and Deaths Associated with COVID-19 by Race and Ethnicity

COVID-19: a need for real-time monitoring of weekly excess deaths. The Lancet

Understanding excess mortality: What is the fairest way to compare COVID-19 deaths internationally? The Health Foundation

Essential Critical Infrastructure Workers list

Automatic time series forecasting: The forecast Package for R

Time Series Analysis: Forecasting and Control

Excess Deaths From COVID-19 and Other Causes

Explaining the Immigrant Health Advantage: Self-selection and Protection in Health-Related Factors Among Five Major National-Origin Immigrant Groups in the United States

Length of Residence in the United States is Associated With a Higher Prevalence of Cardiometabolic Risk Factors in Immigrants: A Contemporary Analysis of the National Health Interview Survey

Bureau of Labor Statistics. FOREIGN-BORN WORKERS: LABOR FORCE CHARACTERISTICS-2019. Accessed

Undocumented U.S. Immigrants and Covid-19

Policies of Exclusion: Implications for the Health of Immigrants and Their Children

The Color of COVID-19: Structural Racism and the Disproportionate Impact of the Pandemic on Older Black and Latinx Adults

Racial Capitalism: A Fundamental Cause of Novel Coronavirus (COVID-19) Pandemic Inequities in the United States. Health Educ Behav

COVID-19: US federal accountability for entry, spread, and inequities-lessons for the future

Contesting Inequality: The Impact of Immigrant Legal Status and Education on Legal Knowledge and Claims-Making in Low-Wage Labor Markets

Mexicans decline to less than half the U.S. unauthorized immigrant population for the first time

Facts on Latinos of Guatemalan origin in the U.S. Pew Research Center's Hispanic Trends Project. Accessed

Serious Challenges And Potential Solutions For Immigrant Health During COVID-19

Access-To-Care Differences Between Mexican-Heritage And Other Latinos In California After The Affordable Care Act

Risk for COVID-19 infection and death among Latinos in the United States: examining heterogeneity in transmission dynamics

Nativity and Country of Origin Variations in Life Expectancy With Functional Limitations Among Older Hispanics in the United States

Latinos as the "Living Dead": Raciality, expendability, and border militarization

Occupational Safety and Health Administration (OSHA) and Worker Safety During the COVID-19 Pandemic

State of California Department of Industrial Relations. Cal/OSHA Emergency Regulations to Protect Workers from COVID-19 in Effect

Coronavirus Disease 2019 and the Case to Cover Undocumented Immigrants in California

One Oakland neighborhood has seen a surge in positivity rates among Latino and Indigenous populations. Why? San Francisco Chronicle

High school or more 1696 (1269, 2107) high school degree/GED 2839 (2361, 3296) 1.38 (1.30, 1.47)

Occupational analyses limited to decedents 25 to 65 years of age at death, otherwise among decedents 25 years of age and older; Risk ratio is observed/expected for all-cause mortality, comparing the pandemic period

Risk ratio*