Effect of income on the cumulative incidence of COVID-19: an ecological study *

Objective to analyze the relationship between per capita income and the cumulative incidence of COVID-19 in the neighborhoods of the city of Rio de Janeiro, RJ, Brazil. Method an ecological study using neighborhoods as units of analysis. The cumulative incidence rate per 100,000 inhabitants and the median of potential confounding variables (sex, race, and age) were calculated. Multiple analysis included quantile regression, estimating the regression coefficients of the variable income for every five percentiles from the 10th to 90th percentiles to verify the relationship between income and incidence. Results the city’s rate was 36.58 new cases per 100,000 inhabitants. In general, the highest rates were observed in the wealthiest regions. Multiple analysis was consistent with this observation since the per capita income affected all percentiles analyzed, with a median regression coefficient of 0.02 (p-value <0.001; R2 32.93). That is, there is an increase of R$ 0.02 in the neighborhood’s per capita income for every unit of incidence. Conclusion cumulative incident rates of COVID-19 are influenced by one’s neighborhood of residency, suggesting that access to testing is uneven.


Introduction
With the declaration of the new Coronavirus disease  pandemic by the World Health Organization on March 11 th , 2020, the escalation of studies addressing potential strategies to cope with the disease has reached a new level (1)(2) . A COVID-19 vaccine, which is the primary means to decrease the population's susceptibility, and consequently the best way to flatten epidemic curves, is inexistent up to now. Thus, only measures to intervene in the social structure as a way to reduce the speed at which COVID-19 spreads remain (3) .
The need to identify the more vulnerable groups and the disease's behavior has gained space in the editorials of the leading international and Brazilian periodicals. Studies addressing early diagnosis, treatment, and containment mechanisms have guided debates worldwide (4)(5) . Much uncertainty remains though regarding the disease's social dynamics in developing countries and more impoverished regions.
The reason is that accumulated theoretical knowledge regarding infectious diseases indicates that there is a potential relationship between per capita income and the incidence of the disease; its spread dynamics are different in European and North American regions, where a significant portion of studies addressing COVID-19 originated (6) . Studies addressing HIV, tuberculosis, and leprosy have already reported this relationship so that it remains to be known how this variable behaves in the case of the COVID-19 pandemic, a critical gap so far (7)(8)(9) .
Based on the international debate on the topic, this study postulates three other factors related to income, mainly due to their possible role as confounding variables in analytical models. The first postulate found in the Brazilian literature is that per capita income is related to race. Economic and social inequalities produce harmful effects on illness and access to health services (10) . To some extent, this relationship can be reproduced in the  context, so that these effects need to be investigated in analytical models. There is abundant literature reporting that women more frequently access health services, either because of the way they conceive the health-disease construct and understand their selfassessment of health, or because of the -already agreed upon -harmful effects of the traditional view of gender on the male population (11)(12) . In this sense, the distribution of sex in the composition of the population may somehow affect the epidemic's behavior.
Bringing up this debate in which race and socioeconomic class are intimately involved in the production of health and disease, an American newspaper presents a report on the possibility of more significant contagion and lethality among Afrodescendent individuals (13) . If we consider that the relationship of income distribution in Brazil is affected by race, investigations are needed to address these variables in explanatory models.
Finally, similar to other respiratory infections, advanced age (the elderly) has been appointed as the variable leading to the most severe manifestations of the disease in most papers investigating the COVID-19 pandemic (14)(15) . In this sense, it seems essential to control models intended to investigate the relationship between income and COVID-19 for sex, race, and age, to perform a more complex explanatory analysis, an aspect that is lacking in the knowledge acquired so far. Therefore, this study presents an alternative hypothesis, that the cumulative incidence rates are directly linked to the per capita income of neighborhoods, regardless of the remaining predictors (composition according to sex, age, and race).
Therefore, this study's objective is: to analyze the relationship between per capita income and cumulative incidence of the COVID-19 in the neighborhoods in the city of Rio de Janeiro, RJ, Brazil.

Method
This is an ecological study, the units of analysis of which were 159 neighborhoods in the city of Rio de Janeiro, located in the metropolitan region of Rio de Janeiro. The option to investigate at a neighborhood level, especially in the city of Rio de Janeiro, is because of the possibility to understand social nuances and inequalities, which may ultimately behave similarly to other Brazilian regions.  this application were: sex, age, date of report, city of residence, neighborhood, and test result. Two aspects need to be clarified: 1) the unit of observation used by the state government system is the individual; 2) the report of the number of cases, which is updated daily, is only available after laboratory confirmation of positive testing -information regarding suspected cases discarded upon laboratory exam is not disclosed.
Data were collected from April 7 th to 13 th 2020.
Considering the speed at which data are produced due to the progression phase of the COVID-19 epidemic in Brazil, it is important to note that data concerning the disease occurred on April 13 th , totaling 2,323 observations. Of these, 11 were excluded: nine because they referred to another city, and two were imported cases (allochthonous). This database contains community transmissions from April 5 th to 13 th , 2020. After data collection, variables were  and 828, respectively. In both cases, the p-value was below 0.001. In addition to this test, the authors also performed visual tests of distribution, confirming that the distribution of data was not normal. Therefore, the authors needed to propose models of analysis that considered the characteristics of this distribution.
Hence, due to the non-normal distribution, the univariate analyses consisted of calculating the median values of the percentages of women, of those 60 years or older, and afro-descendent individuals, and per capita income in Reais (R$) aggregated in administrative regions, according to criteria established by the city of Rio de Janeiro (18) .
A bivariate analysis was performed to measure the association between the exposure variable of interest and outcome, using Spearman's correlation because the distribution was not normal. This test considered the null hypothesis that two variables are independent, considered to be statistically significant at 5% (type 1 error).
The effect of the exposure variable on the outcome was estimated using a multiple quantile regression analysis, using the neighborhood as the unit of analysis (19) . This regression model was chosen  Table 1

Discussion
Relating income with the incidence of infectious diseases is not new, as this relationship has been reported by other studies addressing infectious diseases (7)(8)(9)20) . Since the beginning of the pandemic, health authorities have expressed concern with the behavior of the epidemic in the world's most impoverished regions (6,(21)(22) . From what is known about the social structures in the city of Rio de Janeiro, considering a higher population density on the outskirts and more impoverished neighborhoods, the natural hypothesis is that these regions would experience a greater spread of the disease, as it occurs in the epidemiological chain of other respiratory diseases (7,9,20) . Considering that the largest populations with the worst access to services are concentrated in the most impoverished regions, we conclude that the disease is spreading more rapidly in the regions with the lowest per capita income, even though it was initiated in the areas with the highest per capita income (27) . For these reasons, it seems unlikely that the relationship between income and incidence is a reflex of the social isolation adopted in the city's wealthiest regions. On the contrary, the explanation likely lies in the second hypothesis raised in the UFRRJ report, that is, high underreporting rates in the poorest regions are the origin of this phenomenon.
The key to solving this theoretical imbroglio may be linked to three interrelated aspects: classification, testing, and unequal access to health services. indicates a considerable and fallacious decrease in the incidence of the disease, preventing the planning of more effective public policies (3) .
There is a consensus on the expressive underreporting of the disease. A mathematical model produced by the Imperial College London indicates a mean detection rate well below the actual context of the pandemic (28) . The number of cases in Brazil is estimated to be 12 to 15 times greater than that reported by the Ministry of Health (29)(30) .
One study based on the Chinese experience in Wuhan estimates that 86% of the infectious cases were not documented before the country's health authorities imposed restricted mobility, and these infectious cases were the source of contamination of approximately 80% of new cases (31) .  (8) .
Therefore, it is expected that approximately 5% of the infected people will need intensive care, 2.3% will require mechanical ventilation, while Brazil presents problems in the number and distribution of intensive care beds. Thus, it is urgent to expand COVID-19 testing in the population, decreasing regional inequalities in Brazilian cities (34) . It is worth mentioning that even in China, where the health system was minimally organized to receive COVID-19 patients, the influence of disparities in the access to health services was observed (35) . After all, capital-