Surviving COVID-19 according to race: evidence by a Brazilian retrospective cohort

The COVID-19 pandemic did not impact and still now does not impact people homogeneously. In Brazil, race shows itself as an important difference in health events, including COVID-19 outcomes. We observed, during the pandemic, a higher lethality pattern among black and brown populations. Considering the most important factor for the disease severity in Brazil are, in order of relevance, age, socioeconomic factors and, only then, comorbidities, and that black and brown Brazilians have much poorer socioeconomic conditions compared to white people, we can understand these populations as part of a group under greater risk of aggravation by COVID-19. Besides, it is known that black and brown people face more di�culties to access healthcare services, a way that sometimes they are not even aware of the comorbidities possessed, which can potentially aggravate COVID-19, or present fewer possibilities to control these diseases. Nonetheless, intrinsic, institutional, and structural racism, a health social determinant shown by many indicators such as mortality rates of black and brown populations, presents itself in all healthcare levels in Brazil. Thus, this study aims to analyze the racial differential for COVID-19 survival amongst hospitalized patients in Rio de Janeiro during the COVID-19 pandemic. We performed a survival analysis from selected noti�cations of COVID-19-induced induced Severe Acute Respiratory Syndrome in Rio de Janeiro from the date of the �rst death registered in Brazil to the end of the Public Healthcare Emergency of National Interest, in order to evaluate the times between the �rst symptoms and hospitalization; hospitalization and outcome (death); and �rst symptoms and outcomes (death), relating those to the variable of interest race and the covariables age; sex; presence or absence of major signs/symptoms; presence or absence of multimorbidities; resident of Rio de Janeiro or not; resident of urban/rural areas; ICU hospitalization or not. With that, we aim to characterize hospitalized COVID-19 cases in Rio de Janeiro regarding sociodemographic and clinical variables, describe the course between initial symptomatology and outcomes of the in-patients who utterly passed, analyze the survival probability of in-patients according to their race/skin color; relate social health determinants aspects to the survival rates of the hospitalized patients.


Introduction
COVID-19 has brought incomparable pressure over national health systems worldwide.The pandemics, while deepening socioeconomic inequalities, have been aggravated by them.This feedback generated signi cant consequences for the countries, including di culties in educational and health access, worsening living conditions, loss of employability and income [1].Frequently, this mechanism is understood as a syndemic, when observed under a perspective that extrapolates the clinical practice isolatedly [2].
Brazil has shown itself as a case to be carefully evaluated, since it was presented as the pandemics' epicenter for ve months, from March to June 2021.due to the variant P.1 development on its territory [3], while suffering from a federal government that denied the illness' severity, discouraged social isolation, held forth medications with no scienti c e cacy evidence, and minimized vaccination importance.[4; 5; 6].One of the consequences of this behavior was the requirement of decentralized acting by states and municipalities, under no single command [7], which utterly culminated in peculiar cases whose speci c analysis is plausible.
The city of Rio de Janeiro, for example, has been the Brazilian pandemics' epicenter for the last trimester of 2020 [8] and the whole second semester of 2021 [9].Furthermore, the behavior of lethality rates in the town has been erratic.Its peaks, during the two most signi cant phases of the pandemic, have been, respectively, 16.79% (Epidemiological Week EW 19/2020) and 10.05% (EW 13/2021).During the same epidemiological weeks, world rates were 7.19% and 1.31%.In addition, it's important to mention that, during the pandemics' rst phase, the city performed closer to high-income countries and during the second phase, more like low-income countries did [10].Moreover, we highlight the signi cant in-town social inequity, shown by some indicators.The average income of the 10% richest people is 25.8% higher than that of the poorest 40% [11].Besides, the city measures its social development based on an index that has great variability, the social development index (IDS).Approximately 73.55% of the town's districts and the city itself are less developed than the best raking favela (Favela da Regata), according to the IDS.Regarding the human development index (HDI), the best-ranking district in the city (Gávea) ranks comparatively to Switzerland (0.970 vs. 0.942), while the worse (Complexo do Alemão) ranks similarly to Mongolia (0.700 vs. 0.701) [12; 10].Finally, it's important to emphasize that the regional organization of the Brazilian public health system (SUS) assumes Rio de Janeiro to be responsible to absorb health demands from other 21 towns located in the metropolitan area, which overburdens its local health system [13].
A syndemic approach to COVID-19 links its explanatory models to social determinants of health.That said, we consider race/skin color as one of its most important explanatory components since in many circumstances race presents itself as a proxy of socioeconomic status [14].
In Brazil, the straight relation between social hierarchy, poverty, and race/skin color is clear [15; 16; 17].It is expected, then, that the poorest population, which comprises black/brown people, has been disproportionately affected by the virus and its consequences.Rio de Janeiro's demographic and socioeconomic distribution is similar to the rest of Brazil, regarding race and income [16], though the dynamics of the pandemic has differed from the rest of the country.That said, our objective was to analyze the race/skin color differential impact on the survival of hospitalized COVID-19 patients in Rio de Janeiro during the pandemic.

Study design and availability of data and materials
This study has been developed in the city of Rio de Janeiro, the capital of the homonymous state, located in Southeast Brazil.It has approximately 6.775.561inhabitants [18], 47.96% of which declare as brown or black according to the census 2010 [19].The Primary Health Care (PHC) covered, in March 2020, 49.89% of the population, reaching 74.2% in May 2022.[20].
We organized a retrospective cohort in which we evaluated all the noti ed cases of SARS induced by COVID-19 in Rio de Janeiro, from March 12, 2020, date of the rst death registered in Brazil due to the disease [21], to May 23, 2022, when the federal government declared the end of the public health emergency of national interest (ESPIN).All the noti cation data have been obtained on the In uenza Epidemiological Vigilance Information System (SIVEP-Gripe), which also holds COVID-19's data, an open source and public domain database.The dataset supporting the conclusions of this article are available in the SIVEP-Gripe repository, [https://opendatasus.saude.gov.br/dataset/srag-2020] and [https://opendatasus.saude.gov.br/dataset/srag-2021-a-2023].
We described the eligible population selection proceedings in Fig. 1.The cohort comprised 92,058 noti cations regarding hospitalizations in the city of Rio de Janeiro.We excluded patients who developed SARS for any reason other than COVID-19 and those whose noti cation declared asiatic/indigenous race/skin color.

Data analysis
The analyzed independent covariables were sex; age; rural or urban living; Rio de Janeiro or other municipality living; multimorbidities presence or absence; major signs/symptoms on noti cation time presence or absence; intensive-care unit (ICU) hospitalization or not.Age groups were classi ed as 0 to 19 years old; 20 to 39 years old; 40 to 59 years old; 60 to 79 years old; 80 years old and above.As major signs/symptoms, we considered cough, dyspnea, respiratory discomfort, and/or peripheral oxygen saturation lower than 95% on room air.As for multimorbidities, we comprehend in-patients knowingly bestowing at least two previously diagnosed chronic diseases.
The rst stage consisted of an exploratory analysis of absolute and relative frequencies of the categorical variables included in the research.Then, we applied the Kaplan-Meier method to evaluate which independent variables would demonstrate statistical signi cance.For that we performed a log-rank test, comparing the survival curves between groups.Finally, starting from the results found, we proceeded with a time-dependent Cox regression from the forward strategy for those statistically signi cant variables.
The dependent variables were the time between the onset of rst symptoms and hospitalization; the time between hospitalization and outcome; the time between the onset of rst symptoms and outcome.In order to formulate the rst, we considered as the date of entry on the study the "date of rst SARS symptoms" declared by the patient and, as the outcome, the "hospitalization date".To compose the second dependent variable, we used as the entry date the "date of hospitalization" and as the outcome the "date of closure", which indicates the outcome of death or release.As for the last dependent variable, we considered as the entry date on the study the "date of rst SARS symptoms" and, as the outcome, the "date of closure".All mentioned dates were lled in the SARS noti cation sheet for in-patients and there were no cases of non-completion on the analyzed database.We considered as our event of interest the death caused by COVID-19 and censored those whose causes were any other.
To maintain the participants, we obtained variability measures for the times described and excluded all those who were above the upper limit of the distribution presented by boxplots.With this, we discard possible lling errors and database outliers.
The initial diagnosis obtained by Schoenfeld's residue evaluation was that of risk proportionality during the follow-up period.Therefore, we decided on the conventional Cox regression model for our analysis.We developed, then, three models in which time was measured in days.We adjusted the models of Cox's proportional risks in order to evaluate if race was associated with in-hospital survival.The event considered as failure was death by COVID-19, From the proportional risks we estimated the hazard ratio (HR) with con dence intervals of 95% (CI 95%).
All the analyses were executed with IBM SPSS Statistics software, version 26.

Noti cation sheet completion analysis
Initially, we evaluated the noti cations regarding their data completeness (Table 1).We observed that, among mandatory and non-mandatory variables, there is a signi cant completeness percentage gap.The variables town of residence; sex; date of birth; race/skin color (since July 09th, 2020); ICU hospitalization are mandatory [22].Regarding the completeness of our variable of interest, there was a signi cant proportion of incompleteness on race/skin color.However, it is worth mentioning that its completion has become mandatory on the SARS sheet on July 09th, 2020 [44], in order to follow an Ordinance from the Health Ministry n. 344, from February 1st, 2017, which disposes of race/skin color on health information forms [42].Until that date, there were 20.165 SARS by COVID-19 noti cations in Rio de Janeiro, 8.021 of which had not the mentioned eld completed, which represents 8.71% of all the noti cations that this study comprehends.The other 26.19%, or 24.127 noti cations, were lled with the "ignored" option.
Fields comprehending declared symptoms in admission and patient's comorbidities have never been mandatory and, therefore, we understand this as the reason for such an expressive volume of noti cations in which those pieces of information are absent.However, it is unlikely that oligo or asymptomatic patients required hospitalization.Considering also that 57.31% of the in-patients were 60 years old or above, and that multimorbidities' prevalence in this age group varies from 30.7 to 57% [43], the total of 10.48% noti cations with reported multimorbidities is probably underestimated.
Finally, regarding the necessary information to model time-to-the-interest-event, the elds "hospitalization date" and "SARS' rst symptoms' date" are mandatory according to the noti cation sheet instructions.Also, the eld "closure date", needed to model time-to-outcome is mandatory according to the same document [22].

Exploratory analysis
We performed an exploratory analysis of the valid noti cations by categorically distributed variables (Table 2).Deaths by causes other than COVID-19 were censored.During the period of 802 analyzed days, most of the hospitalization noti cations belonged to male patients (53.9%); 60 years old and above (57.3%);caucasians (51.3%); inhabitants of Rio de Janeiro (85.5%); urban living (90.3%); patients who did not require ICU hospitalization (45.9%); patients with presence of major signs/symptoms at admission (48,4%); and presence of multimorbidities (10.5%).There was a predominance of incidence of hospitalization and death in male patients.However, the mortality rate of female COVID-19 in-patients was higher: 467.9/1.000.versus 462.5/1.000 for males.Age distribution was normal, with a mean age of 61.35 years old and a standard deviation of 18.84.There was a gradient of death incidence the older the patient.An amount of 57.3% of the noti cations referred to patients 60 years old and above.There was a strong predominance of SARS by COVID-19 in the age group from 60 to 79 years old.For males, the age mean was 59.78 years old with a standard deviation of 18.2 and, for females, 63.17 years old, with a 19.39 standard deviation.The incidence difference of SARS by COVID-19 between these groups was statistically signi cant (p < 0.001).(Supplementary material #1).
Regarding the analyzed hospitalizations, 14.9% refer to patients living in other towns than Rio de Janeiro, and only 0.11% in rural areas.Furthermore, 54.5% of in-patients required ICU beds.The mortality rate for this population was 527.7/1.000, while for the ward in-patients was 335.3/1.000.Of the valid noti cations, 52.1% belonged to caucasian patients, 53.63% of which were discharged.In the brown population, 62.22% died, and 61.83% of black people also died.Though there is a discrepancy, then, with caucasians most frequently hospitalized due to SARS by COVID-19 in Rio de Janeiro, we note that both absolute and proportional deaths in this population were less frequent than among black and brown people combined.

Survival analysis
The global model evaluated the complete scenario, from the onset of the rst symptoms to the death/discharge of patients hospitalized with SARS by COVID-19 (Table 3).Kaplan-Meier curves disclosed the association between mortality with every analyzed variable, except living zone and major signs/symptoms at admission.We highlight that for race/skin color and age group there was a gradient of time reduction through the analysis categories.Adjusted analysis revealed a higher, non-signi cative, probability of death of brown people (HR = 1.03,CI 95% 0.98-1.08)and signi cant for black patients (HR=.,17,IC 95% 1.09-1.26)when compared to caucasians.We emphasize that, though the brown category has not shown statistical signi cance, the gradient between categories has had (p < 0.001).Also, in this adjusted model we observed that the lowest probability of survival occurred among the elderly, who had a 217% higher probability of death among the 60-79 years old age group (HR = 3.17, CI 95%, p < 0.001) and 324% higher for those of 80 years old and above (HR = 4,24, IC 95%, p < 0.001) when compared to people of 0-19 years old.For those in-patients who required ICU beds, the probability of death was 17% higher (HR = 1.17,IC 95%, p < 0.001) than for those who were hospitalized in ward beds.
Finally, it is worth mentioning that the variable sex disclosed a protecting effect in the raw analysis (HR = 0.96, IC 95%, p = 0.94-0.98)for males and changed its association effect in adjusted analysis (HR = 1.06,IC 95%, p = 1.01-1.10).On the other hand, variable multimorbidity lost its statistical signi cance at the model's adjustment.Based on these results, we verify the interaction between these variables.In order to con rm this hypothesis, we crossed the data between the aforementioned variables and noticed that the prevalence of multimorbidities in females was 50.5%, while in males it was 46.9%.That said, the multivariate model controlled the interaction effect between sex and multimorbidity, displaying the protection factor of being female, in this case.Because of that, we considered this interaction effect on the association estimates of the adjusted model.
The analysis of the course of the disease among in-patients highlighted some variables associated with the time until the outcome of interest (death).However, we are aware of two well-marked phases during this period: the time between the onset of rst symptoms and hospitalization, and the time between hospitalization and outcome.We believe they represent two different approaches, which relate to access to hospitalization (and, eventually, diagnosis) and the disease's prognosis.For that reason, we performed the survival analysis for both of these times (Tables 4 and 5, respectively).Regarding the cohort follow-up comprehending time between rst symptoms and hospitalization, there was a positive association between these and the variables sex; age, being shorter the time the older the patient; UCI hospitalization; race/skin color, being shorter the time the darker the patient's phenotype; presence/absence of multimorbidities (Table 4).In this follow-up, there was a higher probability of hospitalization requirement for brown (HR = 1.23,CI 95% 1.17-1.28)and black patients (HR = 1.27,CI 95% 1.18-1.37),if compared to caucasians.Both categories showed a statistically signi cant gradient between them (p < 0.001).Other variables maintained similar associations to those veri ed on total follow-up time.
Variable sex demonstrated, once more, a protecting effect in the raw analysis (HR = 0.93, CI 95% 0.90-0.95) and changes its effect in the adjusted analysis (HR = 1.04,CI 95% 0.99-1.08).Also, the variable multimorbidity lost its statistical signi cance when adjusting this model.We emphasize that there is an association between sex and multimorbidity, which is the reason for this effect modi cation.
Regarding the follow-up in which we analyzed the time between hospitalization and outcome (death) (Table 5), the variables "town of residence" and "absence or presence of multimorbidities" lost statistical signi cance when adjusted.Furthermore, we consider it important to note that the variable "race/skin color" lost its association gradient.Category "brown" lost statistical signi cance.The "black" category, though, sustained risk association, with a death probability of 10.1% (CI 95%) higher than among white in-patients.Besides, we realized there is a time gradient between hospitalization and death through the age groups, with shorter time the older the patient.It was 375% (CI 95%) more probable for in-patients of age 80 or above to die if compared to those between age 0 and 19 years old.Also, it was 15% (CI 95%) more probable that patients who required ICU beds to die, in comparison to those hospitalized in wards.All the models proved "age" to be the variable with the most important association with outcomes, being more negative the older the patients, since the onset of the rst symptoms.Moreover, we emphasize that in all three models, there was an effect modi cation for the variable "sex" among raw and adjusted analyses.As previously mentioned, given the association strength in which model, we believe there has been an interaction with the variable "multimorbidity", since selfreferred illnesses are knowingly better reported among females.
Beyond that, it's important to highlight that the variables did not maintain equal association strength among the models."Race/skin color", "ICU hospitalization" and "town of living" demonstrated higher association with the model which analyzed the time between the onset of rst symptoms and hospitalization, which reveals access to healthcare prior to the infection by the new coronavirus (SARS-CoV-2) may have been determinant to each outcome.
Once hospitalized, there was no strong association with "race/skin color", which demonstrates that this variable was not a determinant of in-hospital care.
Finally, the variables "presence of major signs/symptoms" and "multimorbidities" were more associated with the model which evaluated the time between hospitalization and outcome.

Discussion
The SARS-CoV-2 pandemic is an emergent public health issue.However, its behavior reproduces the same pattern observed in other transmissible and nontransmissible diseases.Socioeconomic conditions, access to healthcare and structural, institutional, and implicit racism do in uence the health-illness binomial and, certainly, are determinants for negative outcomes in most vulnerable populations [23].
The variable "race/skin color" was the mandatory completion SARS sheet eld which showed the worst completeness percentage.The inadequate completion of this data demonstrates the lack of importance given by health professionals from Rio de Janeiro regarding the racial theme.It is important to acknowledge that some authors [24; 25; 26] consider this lack of lling of the variable "race/skin color" as a central element to explain the occurrence of the disease.We consider it structural because the lack of knowledge by health professionals regarding the purpose of completing sociodemographic data on noti cation sheets typi es non-intentional error, but is implicitly constructed by considering the health situation detached from the racial context.
Given the three developed models and comparing the variables' performance among each other, we realized that "race/skin color", "ICU hospitalization" and "town of living" had a stronger association with the model time between the onset of the rst symptoms and hospitalization.That said, we can assume these variables are mostly associated with factors related to access to healthcare rather than the post-hospitalization clinical prognosis.
Among patients who were hospitalized in Rio de Janeiro and died due to COVID-19, we observed a decreasing gradient between the time of onset of rst symptoms and death for caucasian, brown, and black patients, respectively.Due to the discrepancy between brown and black populations, we justify the need to categorize these separately.Therefore, even considering that other variables, such as "age" and "ICU hospitalization" have had an important association with obtained results, we realize that "race/skin color" has been determinant for negative outcomes.Our ndings demonstrate that, probably, both black and brown populations were admitted to healthcare services presenting more advanced disease.One of the hypotheses for this fact is less schooling, greater exposure to the virus due to the impossibility to keep social distancing given the need for in-person working, higher rates of public transport usage, in-home agglomerations, and population density of the neighborhoods they live in [27].
There was also a gradient regarding "age", which has been proven to be more probable for both hospitalization and death the older the patient during the whole period of analysis, despite the "rejuvenation" of the hospitalizations by COVID-19 observed during the pandemics, due to the staggered vaccination approach by age in Rio de Janeiro. [28].This predominance of death probability among elders 60 years old and above in Rio de Janeiro, corresponds to the results found in the rest of the country [29]."Age" was the variable that showed a stronger association with this negative outcome.
Variables "presence of major signs/symptoms" and "multimorbidities" were more signi cant to the model time between hospitalization and death.Major signs/symptoms have acted as protecting factors in the model time between the onset of rst symptoms and hospitalization, and as a risk factor in the model time between hospitalization and death, as expected.These ndings mean that the early-admitted patients showed a smaller probability of death, while those who were admitted in a critical condition had the worst prognosis.
Still, there was not a signi cant positive association between the presence of previous comorbidities and a higher probability of death in the adjusted analysis.This result contradicts former research which relates multimorbidity isolatedly to the aggravation by COVID-19, both in Brazil [30], and in the world [31].This demonstrated that it is inadequate to trace a direct causal relationship between these factors.However, our ndings reinforce those of Pedro Baqui et col (2021), in which it was observed that the factors mostly involved with the probability of aggravation of COVID-19 are, respectively, age, socioeconomic conditions, and, nally, presence of comorbidities [32].Our results also con rm those about black, browns, and elders being the most affected by the disease [33].
It should be noted that it was not possible to categorize the noti cations regarding the nature of public or private health institutions in which patients were hospitalized, since the database for the "hospitalization unit" variable is nominal and composed of 250 different institutions.Among these, there are federal, municipal, and state public administration units; public and mixed economy companies; business entities; non-pro table units; self-administered and extraterritorial institutions, which makes the manual classi cation of all of them impracticable.Even so, the racial characterization of the users of healthcare services is an important aspect to be discussed.
Brazil is a deeply marked country concerning social and economical inequalities.According to World Inequality Database, in 2021 it ranked #17 out of 169 countries analyzed [34].Considering racial imbalances, the census 2010 reported Brazil had 50.7% of its population self-declared as people of color, 41.3% of which as brown and 7.6% as black [19].Even so, of the 78% of people who depend on the Brazilian public health system (SUS), 65% are black or brown [35], which demonstrates racial disproportionality.Furthermore, for each person of color with private health insurance in Brazil, there are two caucasians.These facts clearly show that access to healthcare, despite being constitutionally granted universally in the country [36], is utterly unequal according to each Brazilian's skin color.Hospitalization in public or private institutions demonstrates important repercussions over mortality, which is higher in SUS services [32].
People of color in Brazil have a per capita income 50% lower than caucasians and live in the most densely inhabited areas of the country [17].In Rio de Janeiro, particularly, the life expectancy of a person of color is up to 10 years shorter than caucasians [37].Circa 18.96% of their residences have no sewer facilities whatsoever [38].Since individuals who cannot maintain social isolation are most subject to contamination by the SARS-CoV-2 and, knowing that black and brown people are the ones mostly informally occupied and relying on public transportation [17], we can deduce that they are the most exposed to the virus.That said, knowing that the pandemics affected black and brown people unequally [39], we aimed to analyze if the disease's aggrievement and its negative effects have also been imposed disproportionately over this population in Rio de Janeiro.
According to the model time between the onset of rst symptoms and hospitalization, we can assume there were no di culties for black and brown people to access secondary and tertiary health care services.The creation of eld hospitals and the opening of municipal beds during the pandemic probably minimized this access restriction [40].However, we can question the role played by the primary health care (PHC) in the resolution of health demands of the most vulnerable, since black and brown people, the main users of SUS and PHC [41], have been aggravated by COVID-19 faster than caucasians.These individuals, who historically have had more di culties accessing healthcare services [24], are the same who showed a greater probability of hospitalization and death by COVID-19.We also question if the PHC politics of the city of Rio de Janeiro is indeed pointed towards the reduction of inequalities, which is a premise of any public politics, and if there is a racial orientation in their actions.
This study has limitations.Firstly, we acknowledge the loss of information regarding multimorbidity's occurrence, once its information is provided with no clinical dispute whatsoever.In the absence of diagnostic opportunities for pre-existing conditions, its completeness is compromised.However, we assume it's a non-selective loss, since we approached hospitalization data and, once admitted, patients with clearly manifested illnesses would have these investigated and managed, despite the main cause of hospitalization.Also, we have not retrieved information regarding the nancing nature of the institutions of hospitalization (public or private).However, the results supported the hypothesis that "race/skin color" was a more important variable in order to explain access to healthcare issues, and not the clinical intervention itself.Besides, the high-complexity public healthcare system is substantially larger than the private one.Given that, we believe this absence of information does not jeopardize our results.Finally, we excluded the Asiatic and indigenous populations from the study.Though it means an analysis loss for these sub-populations, this methodological option allowed us to evaluate the association gradient for black and brown patients, enlightening the colorism issue as a social determinant.It is worth reminding that Asiatic and indigenous populations accounted for only 0.6% of the hospitalizations by COVID-19 in Rio de Janeiro.That said, it's a non-important loss for the resulting associations.New studies which consider these points may impact the results.However, we consider the relevance of this research concerning the racial issue, given the signi cance of our ndings.

Final Considerations
The SARS-CoV-2 pandemic has brought, in addition to an alarming death toll, severe damage to the most vulnerable populations in Brazil and the world.The elderly, the poor, and the black and brown Brazilians have found themselves often more exposed to the virus itself and the secondary effects brought by the pandemics in education, physical and mental health, employability, and income.Social determinants exert in uence in the whole health context, from the impossibility to access hygiene and healthcare, preventing social isolation of those non-formally employed and essential workers, and, nally, making those who did get infected by the virus, have a higher probability of severity, due to the intersectionalities that per pass them.In Rio de Janeiro, we noted a larger amount of caucasian patients who required hospitalization, but a mortality rate considerably higher and a lower probability rate to survive among black and brown patients, which con rms these populations as more vulnerable to COVID-19.
An important aspect regarding access to public healthcare is that much of the effort targeting complications derived from the SARS-CoV-2 infections was made by secondary and tertiary healthcare.However, we consider it strategic for PHC to assume its role as gatekeeper of the health system, organizing it.This strategy aims to recover the original health politics intention to stimulate equity, especially towards black and brown populations.PHC must behave as the healthcare coordinator not only, but particularly of these patients, developing primary prevention actions, health education, and managing mild to moderate COVID-19 cases, in order to reduce the probability of hospitalization and death.
We consider it fundamental for health professionals to understand the importance of accounting for the racial issues in all of their services, regardless of the patient's complaint, since, even if a person is not only of color, he/she always is, and many times this is what de nes its survival or death.The individual contribution of health professionals, though, is not enough to minimize the vulnerabilities the black and brown populations are subject to.Every primary care policy must be racially oriented.It is of utmost importance for every public policy to consider and be assertive in race/skin color issues in Brazil.Otherwise, we will never reach the ideal of equity determined by the original laws of SUS, and it is improbable a decrease in social disparities which are already historical in the country.

Declarations
Ethical Approval and Consent to participate: Microdata used in this study are public and unidenti ed.Therefore, this research does not require approval from a research ethics committee (REC).The dismissal has been formally provided by an o cial letter from Fiocruz's REC, though (attachment #1).
Consent for publication: I, Danielle Cristina Lourenço dos Santos Pastura, as the corresponding author on behalf of all authors, hereby give our consent to publish our article titled "Surviving COVID-19 according to race: evidence by a Brazilian retrospective cohort" in International Journal for Equity in Health.We con rm that the article is original, has not been previously published, and is not under consideration for publication elsewhere.
We assure you that all authors have made substantial contributions to the article, have reviewed and approved the nal version, and have agreed to be accountable for all aspects of the work.Furthermore, we con rm that the article does not contain any defamatory or unlawful content, and that all necessary permissions and approvals for conducting the research have been obtained.

Table 2
SARS by COVID−19 noti cations in Rio de Janeiro by the outcome

Table 3
Survival model for time between onset of rst symptoms and death.Riode Janeiro, 2020de Janeiro,  -2022

Table 4
Survival model for time between onset of rst symptoms and hospitalization.Riode Janeiro, 2020de Janeiro,  -2022