key: cord-1036129-izv3cvz5
authors: Chishinga, Nathaniel; Gandhi, Neel R; Onwubiko, Udodirim N; Telford, Carson; Prieto, Juliana; Smith, Sasha; Chamberlain, Allison T; Khan, Shamimul; Williams, Steve; Khan, Fazle; Shah, N Sarita
title: Characteristics and Risk Factors for Mortality by COVID-19 Pandemic Waves in Fulton County, Georgia: A Cohort Study March 2020–February 2021
date: 2022-03-03
journal: Open Forum Infect Dis
DOI: 10.1093/ofid/ofac101
sha: a6ccf970bcaed45ab46b573288c874a24ea2ea44
doc_id: 1036129
cord_uid: izv3cvz5

BACKGROUND: We examined differences in mortality among COVID-19 cases in the first, second and third waves of the COVID-19 pandemic. METHODS: A retrospective cohort study of COVID-19 cases in Fulton County, Georgia, USA, reported to a public health surveillance from March 2020 through February 2021. We estimated case fatality rates (CFR) by wave and used Cox proportional hazards random effects models in each wave, with random effects at individual and long-term-care-facility level, to determine risk factors associated with rates of mortality. RESULTS: Of 75,289 confirmed cases, 4,490 (6%) were diagnosed in wave one (CFR 31 deaths/100,000 person days [pd]), 24,293 (32%) in wave two (CFR 7 deaths/100,000 pd), and 46,506 (62%) in wave three (CFR 9 deaths/100,000 pd). Compared to females, males were more likely to die in each wave: Wave one (adjusted hazard ratio [aHR] 1.5, 95% confidence interval [CI] 1.2–1.8), wave two (aHR 1.5, 95% CI 1.2–1.8), and wave three (aHR 1.7, 95% CI 1.5–2.0). Compared to non-Hispanic Whites, non-Hispanic Blacks were more likely to die in each wave: Wave one (aHR 1.4, 95% CI 1.1–1.8), wave two (aHR 1.5, 95% CI 1.2–1.9), and wave three (aHR 1.7, 95% CI 1.4–2.0). Cases with any disability, chronic renal disease, and cardiovascular disease were more likely to die in each wave compared to those without these comorbidities. CONCLUSIONS: Our study found gender and racial/ethnic disparities in COVID-19 mortality, and certain comorbidities associated with COVID-19 mortality. These factors have persisted throughout the COVID-19 pandemic waves, despite improvements in diagnosis and treatment.

Since the first case of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2), the novel coronavirus that causes coronavirus disease 2019 (COVID- 19) , was detected in the United States (U.S.) in January 2020 [1, 2] , there have been serial waves of the epidemic, with cases rising and falling during the summer and winter holiday weeks.

Simultaneously, there have been substantial improvements in the diagnosis, treatment, and prevention of COVID-19, with several interventions demonstrating reduced morbidity and mortality [3] . However, key risk groups have remained disproportionately affected by COVID- 19 , despite these advances.

Early data from the U.S. demonstrated more severe disease occurring among older persons and those with comorbidities [4] , including hypertension, diabetes, obesity, renal disease, lung disease, and immunosuppression [5] [6] [7] . In addition, persons who were non-Hispanic Black represented 23% of COVID-19 related deaths [8] , despite comprising 13.4% of the U.S population [9] . Although hospitalization rates have been reported as higher among non-Hispanic Black patients, results were mixed when evaluating race and ethnicity as a risk factor for death after adjusting for covariates [10, 11] . Adults in the southern US have higher rates of several medical comorbidities (e.g., diabetes, obesity) than those in other parts of the US, with even higher rates among racial and ethnic minorities [12] , raising concerns for even poorer outcomes from COVID-19 in these populations. for priority populations toward the end of the third wave, it remained critically important to monitor trends in severe disease among the most vulnerable groups. To explore changes in demographics of COVID-19 cases over time, including differential impact on severe disease and mortality by epidemic waves, we evaluated individuals with SARS-CoV-2 in Fulton County, Georgia, USA.

We conducted a retrospective cohort study analysis of surveillance data of individuals diagnosed with laboratory-confirmed SARS-CoV-2 infection from March 2, 2020 when the first case of COVID-19 was notified in Georgia [13] to February 28, 2021. A laboratory confirmation of SARS-CoV-2 infection was defined as having a positive result on a real-time reverse transcriptase-polymerase chain reaction (RT-PCR). We included residents of Fulton County, Georgia, which includes 90% of the city of Atlanta. The population in Fulton County is 1.06 million people and represents 10% of the state of Georgia's population [14] . During this study period, we identified three waves of the COVID-19 pandemic that were used in our analyses: The first wave was 90 days from March 2, 2020, to May 30, 2020; the second wave was 119 days from May 31, 2020, to September 26, 2020; and the third wave was 155 days from September 27, 2020, to February 28, 2021 ( Figure 1 ). These time periods correspond with the overall peaks and troughs for the U.S., as reported by the Centers for Disease Control and Prevention. [15] 

Data were extracted from the State Electronic Notifiable Disease Surveillance System (SENDSS), an electronic database used by the Georgia Department of Public Health to track patients with notifiable diseases, including COVID-19 cases. The extracted data for each case included in this study were: date of first SARS-CoV-2 positive specimen collection, age, gender, race and ethnicity, medical comorbidities, residence in a long-term care facility A c c e p t e d M a n u s c r i p t 5 (LTFC), hospitalization, intensive care unit (ICU) admission, and death. Where applicable, dates related to hospitalization, discharge, death, and the length of hospital stay were also extracted and used to determine hospitalization status for records with missing data. To have complete case investigations and reporting to the surveillance system, we included cases that tested positive for COVID-19 up to and including February 28, 2021. Given known delays in disease progression and case investigation for individuals diagnosed in the latter portion of the study period, we allowed for a four-week lag to March 31, 2021, for extraction of data from SENDSS.

The primary outcome measure was mortality, measured as the case fatality rate (CFR), rate ratio, and adjusted hazard ratio (aHR). The CFR was defined as the number of deaths per follow-up time in days (person-days) among cases in each wave [16] . The rate ratio was defined as the ratio of the CFRs between risk groups among confirmed cases in each wave. The aHR was defined as the ratio of the hazard rates of mortality between risk groups among confirmed cases in each wave while accounting for covariates. The secondary outcomes, measured among hospitalized cases only, were the proportion admitted to the intensive care unit (ICU), hospital length of stay, and hospital discharge disposition in each wave.

We described demographic characteristics of all cases as medians and interquartile ranges (IQR) for continuous variables, or frequencies and proportions (%) for categorical variables. Differences in the distributions of the baseline characteristics across the three waves were assessed using Kruskal-Wallis test for continuous variables and Pearson Chisquared test for categorical variables. Missing data on covariates were shown in the descriptive table and excluded in the analyses. In each wave, there were < 1% of cases with missing age or gender, and < 11% of cases with missing race and ethnicity combined. Cases with missing data are included in the descriptive analysis (Table 1 ) but excluded from models.

In the primary outcome analyses, we used Kaplan Meier curves to compare mortality by the COVID-19 pandemic waves, and to compare mortality by gender and race/ethnicity groups in each wave. To determine factors associated with mortality, we obtained crude (unadjusted) rate ratios by gender groups, race/ethnicity groups, and by medical comorbidity (any disability, immunocompromised, chronic renal disease, cardiovascular disease, diabetes mellitus, chronic lung disease, and chronic liver disease). In the adjusted analyses, we fit multivariable Cox proportional hazards regression random effects models by wave, with individuals and long-term care facilities (LTCFs) as random effects (shared frailty models) that contained age, gender, race and ethnicity (combined), and all the recorded medical comorbidities. We included random effects for clustering at individual and LTCF level in the adjusted models. Cases from LTCFs have been shown to contribute high proportions of COVID-19-related hospitalizations and deaths [17] . We anticipated a nonlinear association between age and mortality over time and found the association between age and mortality to be squared and cubic across the three waves for all models (P>0.05 for all models). We therefore added these fractional polynomials for age to the adjusted models [18] . In each of the adjusted Cox models, we used the likelihood ratio test to test the proportional hazards assumption for potential interaction between each variable and time.

We examined if having additional comorbidities to an already existing comorbidity further increased the risk of mortality. We therefore fit Kaplan-Meier curves to determine survival functions by no comorbidity, one comorbidity, two comorbidities, and three or more comorbidities at COVID-19 diagnosis. We used the log-rank test to examine statistical difference among these groups.

In the secondary outcome analyses, we examined the proportions and median times among hospitalized cases and compared them across waves. A two-sided P value of <0.05 was considered statistically significant. Statistical analyses were performed in Stata software A c c e p t e d M a n u s c r i p t 7

As a public health surveillance activity in response to the COVID-19 emergency, this activity was determined to be exempt by Georgia Department of Public Health Institutional Review Board (IRB). The Emory University IRB approved this activity with a waiver of informed consent. The overall CFR per 100,000 person-days decreased from 31 in the first wave to 7 in the second wave but increased to 19 in wave three (Table 1) . Compared to wave one, there was a statistically significant decline in CFR by 80% in wave two (case fatality rate ratio, 0.2; 95%CI 0.2-0.3) and by 40% in wave three (case fatality rate ratio, 0.6; 95%CI 0.5-0.7).

In the unadjusted analyses, the CFR per 100,000 person-days for non-Hispanic Black persons was higher as compared to non-Hispanic White persons across all waves: 43.7 non-Hispanic Black vs 27.2 non-Hispanic White in wave one; 9.7 vs 7.6 in wave two; and 28.3 vs 22.8 in wave three, respectively (Table 1) . Also, the CFR ratio (rate ratio) for non-Hispanic Black compared non-Hispanic White persons was higher in each wave: 1.6 (95%CI 1.3-2.1) in wave one, 1.3 (95%CI 1.0-1.6) in wave two, and 1.2 (95%CI 1.1-1.5) in wave three (Table 1 ).

In the adjusted analyses, compared to females, males were more likely to die in wave one (aHR 1.5, 95% CI 1.2-1.8), wave two (aHR 1.5, 95% CI 1.2-1.8), and wave three (aHR 1.7, 95% CI 1.5-2.0). Compared to non-Hispanic White persons, non-Hispanic Black persons were more likely to die in wave one (aHR 1.4, 95% CI 1.1-1.8), wave two (aHR 1.5, 95% CI 1.2-1.9), and wave three (aHR 1.7, 95% CI 1.4-2.0). Cases with any disability, chronic renal disease, and cardiovascular disease, were more likely to die across all waves compared to those without these comorbidities. Furthermore, cases with immunocompromised status in wave one, and those with chronic lung diseases in wave three were more likely to die than cases without these comorbidities in the respective waves ( Figure 2 ). Kaplan-Meier analysis showed that mortality was strongly associated with having more than one comorbidity at COVID-19 diagnosis in wave one (P <0.0001), wave two (P <0.0001) and wave three (P <0.0001) (Figure 3 ).

Of 4,582 cases that were hospitalized during the study period, 990 (22%) were in Compared to wave one, the proportion of hospitalized cases that died in the subsequent waves were significantly lower (25% in wave one vs. 12% in wave two vs. 16% in wave three, P <0.001). Compared to wave one, the proportion of cases admitted to ICU decreased significantly in the subsequent waves (26% vs. 18% vs. 16% respectively, P <0.001). There were no statistically significant differences in length of hospital stay among the waves (Table 2) .

We examined characteristics associated with death by COVID-19 pandemic waves in Fulton County, Georgia, a densely populated, diverse urban center that includes most of the city of Atlanta and its suburbs. These data provide valuable insights from one of the largest cohorts of COVID-19 cases in the Southeast U.S, a region that showed rapidly rising numbers of COVID-19 in each pandemic wave. Despite improvements in overall epidemiological and clinical outcomes during this period, we found several groups with persistently greater risks of mortality, namely males, non-Hispanic Black individuals, and persons with medical co-morbidities. Even with the widespread availability of vaccines that occurred after the time of this study, there have been persistent disparities in vaccine uptake (and boosting) that are likely to further exacerbate the clinical outcomes observed in our study. Given these groups comprise large proportions of the US population, it is critical that A c c e p t e d M a n u s c r i p t 10 COVID-19 interventions are designed to specifically address their health needs to turn the tide of this epidemic We found that the CFR was high in the first wave of the pandemic but decreased over subsequent waves, coinciding with greater availability of COVID-19 testing and improvements in COVID-19 prevention and treatment. Despite this overall trend, non-Hispanic Black persons had a disproportionately higher CFR and a persistently high rate ratio across all waves, adding to the findings from studies conducted during the early part of the COVID-19 pandemic in California and Louisiana [11, 19] . In addition, while non-Hispanic Black persons comprise 44% of the population in Fulton County [20], 55% of all confirmed COVID-19 cases and 76% of those that died in the first wave were non-Hispanic Black.

Notably, the increased risk of death in this group was independent of age, gender, and medical comorbidities in the adjusted analyses. These data support known inequities in access to and utilization of healthcare and testing services among non-Hispanic Blacks, in part caused by long-standing medical mistrust and experiences of racism [21] . In addition, non-Hispanic Black persons comprise a higher proportion of frontline and essential workers in the U.S., placing them at greater risk for SARS-CoV-2 exposure and infection [22, 23] .

Our findings raise important concerns that, despite awareness of racial and ethnic disparities in COVID-19 disease burden and outcomes since early in the pandemic, these gaps have persisted throughout subsequent waves. With data on similar disparities in COVID-19 vaccine uptake, our study supports the continued need to intensify diagnosis, treatment, and prevention efforts to close gaps in morbidity and mortality.

Consistent with early trends of the COVID-19 pandemic across the U.S. [24] , we found an increased risk of mortality among COVID-19 cases with comorbid medical conditions across all pandemic waves. Furthermore, there was an increased risk of mortality for each increase in the number of comorbidities an individual had. This suggests that having additional comorbidities further complicates the management of COVID-19, which in turn results in increased mortality. Our findings underscore the need to ensure optimized treatment of comorbid conditions-particularly as health services have been disrupted for A c c e p t e d M a n u s c r i p t 11 nearly two years-outreach for COVID-19 vaccine administration, and ongoing transmission prevention measures among individuals with these risk factors. It is important to note that our study period was during a time when the Alpha and Beta variants predominated and before vaccines were widely available [25] . However, the subsequent Delta variant (July-November 2021, -fourth wave‖) was more transmissible, coupled with the general public's COVID fatigue in rigorously maintaining precautions. This resulted in more hospitalizations, despite vaccine availability, further underscoring the need for targeted outreach to high-risk groups. Indeed, as subsequent variants (e.g., Omicron) have demonstrated immune evasion leading to many vaccinated individuals becoming infected, the findings from our study remain highly relevant for monitoring groups who are likely to bear a disproportionate burden of disease.

Lastly, we found that among hospitalized cases, the proportion that died in the hospital decreased in waves two and three, compared to wave one. Several factors likely contributed to the observed improvements. First, limited access to testing in the early parts of the pandemic likely resulted in delays in diagnosis until individuals developed more severe, persistent COVID-19 symptoms or became extremely ill. As testing for COVID-19 became more widely available and policies expanded to allow testing of all age groups, regardless of symptoms, there was an increase in number of cases being diagnosed earlier in the disease course, including persons with mild or asymptomatic infection [26] . Second, improved understanding of the pathophysiology of the COVID-19 disease, expansion of hospital capacity and inpatient supportive treatment, and use of more effective biomedical treatments have resulted in improvements in disease outcomes despite higher caseload in subsequent waves. These improvements in outcomes among hospitalized cases could be jeopardized with the recent Omicron wave (December 2021 onwards) that has resulted in profound staffing shortages, hospitals operating under crisis conditions, and ongoing supply chain issues despite being a milder variant.

Our study is subject to limitations that are inherent to the use of routinely collected public health surveillance data. This includes gaps in reporting of confirmed cases by A c c e p t e d M a n u s c r i p t 12 providers and testing sites, in addition to reporting lag time of up to several weeks for severe outcomes (hospitalization and deaths). To minimize this, we limited case inclusion to cases reported as of February 28, 2021, to allow for sufficient time for reporting and completing case investigations up to March 31, 2021. Also, we could not utilize hospitalization data fully because, as the COVID-19 pandemic progressed, data on hospitalization became less consistently available in the state electronic disease notification surveillance database. In addition, gaps in implementation of testing may have led to an under ascertainment of the true number of cases particularly in the early outbreak period of the COVID-19 pandemic and in the period after change in testing policy. Specifically, the testing policy in the early phase was restricted to symptomatic persons and specific demographics. After the policy was changed to be more inclusive, there were variations in the implementation of testing of all cases. As with other analyses of COVID-19 disparities [27] , our surveillance data were incomplete for age, gender, race and ethnicity. However, compared to other studies, in each wave we had ≤11% of cases missing race and ethnicity, and ≤1% missing age and gender, strengthening the robustness of our findings. We did not include vaccination status as a covariate in our analysis due to delayed linkage between the disease and vaccination surveillance systems. Nonetheless, widespread availability of COVID-19 vaccine in Fulton County did not occur until March 2020 (the end of our study period); thus, the improving trend we observed was unlikely to be due to the protective effect of vaccination.

In conclusion, as the COVID-19 pandemic progressed in Fulton County, Georgia, there were notable improvements in CFR and rate ratios in subsequent waves. Nonetheless, important gaps persisted among males and non-Hispanic Black persons, despite adjusting for age and comorbid medical conditions. Our study is among the largest to examine trends over time in mortality and confirms the early findings of factors associated with mortality, which include gender and race/ethnic disparities, and the presence of any disability, chronic renal disease, and cardiovascular disease that persisted across the three COVID-19 pandemic waves. As access to COVID-19 vaccines increase across the U.S., similar gaps have been observed, raising concerns for further widening of disparities in morbidity and M a n u s c r i p t 13 mortality for vulnerable groups. As the SARS-Cov-2 continues to evolve, the time is now to redouble efforts by clinicians, public health providers, and policy makers to ensure timely prevention, diagnosis, treatment, increased vaccination, and outreach to turn the tide of this pandemic that has shown to surge in waves even among individuals that have been previously vaccinated. A c c e p t e d M a n u s c r i p t 24 Figure 3 

Centers for Disease Control and Prevention. Coronavirus (COVID-19). Available at

First Case of 2019 Novel Coronavirus in the United States

Inferring the effectiveness of government interventions against COVID-19

Severe outcomes among patients with coronavirus disease 2019 (COVID-19) -United States

Presenting Characteristics, Comorbidities, and Outcomes Among 5700 Patients Hospitalized With COVID-19 in the

COVID-19 in kidney transplant recipients

Characteristics of Hospitalized Adults With COVID-19 in an Integrated Health Care System in California

Characteristics and Clinical Outcomes of Adult Patients Hospitalized with COVID-19 -Georgia

Patients In A Large Health Care System In California

Health and Health Coverage in the South: A Data Update

Officials Confirm Two Cases of COVID-19 in Georgia

Estimating mortality from COVID-19

State Reports of Long-Term Care Facility Cases and Deaths Related to COVID-19 (as of

Building multivariable regression models with continuous covariates in clinical epidemiology--with an emphasis on fractional polynomials

Hospitalization and Mortality among Black Patients and White Patients with Covid-19

American Community Survey Demographic and Housing Estimates

A decade of studying implicit racial/ethnic bias in healthcare providers using the implicit association test

Racial Disparities in COVID-19 Mortality Among Essential Workers in the United States

Essential and undervalued: health disparities of African American women in the COVID-19 era'

Coronavirus Disease 2019 Case Surveillance -United States

SARS-CoV-2 Variant Classifications and Definitions

The Proportion of SARS-CoV-2 Infections That Are Asymptomatic : A Systematic Review

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