key: cord-0844157-3308f8hm
authors: Souza, William Marciel de; Buss, Lewis Fletcher; da Silva Candido, Darlan; Carrera, Jean Paul; Li, Sabrina; Zarebski, Alexander; Vincenti-Gonzalez, Maria; Messina, Janey; Sales, Flavia Cristina da Silva; Andrade, Pamela dos Santos; Prete, Carlos A; Nascimento, Vitor Heloiz; Ghilardi, Fabio; Pereira, Rafael Henrique Moraes; Santos, Andreza Aruska de Souza; Abade, Leandro; Gutierrez, Bernardo; Kraemer, Moritz U. G.; Aguiar, Renato Santana; Alexander, Neal; Mayaud, Philippe; Brady, Oliver J; Souza, Izabel Oliva Marcilio de; Gouveia, Nelson; Li, Guangdi; Tami, Adriana; Oliveira, Silvano Barbosa; Porto, Victor Bertollo Gomes; Ganem, Fabiana; Almeida, Walquiria Ferreira; Fantinato, Francieli Fontana Sutile Tardetti; Macario, Eduardo Marques; Oliveira, Wanderson Kleber; Pybus, Oliver; Wu, Chieh-Hsi; Croda, Julio; Sabino, Ester Cerdeira; Faria, Nuno R.
title: Epidemiological and clinical characteristics of the early phase of the COVID-19 epidemic in Brazil
date: 2020-04-29
journal: nan
DOI: 10.1101/2020.04.25.20077396
sha: f788670155076432a587bfa60cc1c3e0cf087e5a
doc_id: 844157
cord_uid: 3308f8hm

Background: The first case of COVID-19 was detected in Brazil on February 25, 2020. We report the epidemiological, demographic, and clinical findings for confirmed COVID-19 cases during the first month of the epidemic in Brazil. Methods: Individual-level and aggregated COVID-19 data were analysed to investigate demographic profiles, socioeconomic drivers and age-sex structure of COVID-19 tested cases. Basic reproduction numbers (R0) were investigated for Sao Paulo and Rio de Janeiro. Multivariate logistic regression analyses were used to identify symptoms associated with confirmed cases and risk factors associated with hospitalization. Laboratory diagnosis for eight respiratory viruses were obtained for 2,429 cases. Findings: By March 25, 1,468 confirmed cases were notified in Brazil, of whom 10% (147 of 1,468) were hospitalised. Of the cases acquired locally (77.8%), two thirds (66.9% of 5,746) were confirmed in private laboratories. Overall, positive association between higher per capita income and COVID-19 diagnosis was identified. The median age of detected cases was 39 years (IQR 30-53). The median R0 was 2.9 for Sao Paulo and Rio de Janeiro. Cardiovascular disease/hypertension were associated with hospitalization. Co-circulation of six respiratory viruses, including influenza A and B and human rhinovirus was detected in low levels. Interpretation: Socioeconomic disparity determines access to SARS-CoV-2 testing in Brazil. The lower median age of infection and hospitalization compared to other countries is expected due to a younger population structure. Enhanced surveillance of respiratory pathogens across socioeconomic statuses is essential to better understand and halt SARS-CoV-2 transmission.

The first case of COVID-19 was detected in Brazil on February 25, 2020. We report the epidemiological, demographic, and clinical findings for confirmed COVID-19 cases during the first month of the epidemic in Brazil.

Individual-level and aggregated COVID-19 data were analysed to investigate demographic profiles, socioeconomic drivers and age-sex structure of COVID-19 tested cases. Basic reproduction numbers (R 0 ) were investigated for São Paulo and Rio de Janeiro. Multivariate logistic regression analyses were used to identify symptoms associated with confirmed cases and risk factors associated with hospitalization. Laboratory diagnosis for eight respiratory viruses were obtained for 2,429 cases.

By March 25, 1,468 confirmed cases were notified in Brazil, of whom 10% (147 of 1,468) were hospitalised. Of the cases acquired locally (77·8%), two thirds (66·9% of 5,746) were confirmed in private laboratories. Overall, positive association between higher per capita income and COVID-19 diagnosis was identified. The median age of detected cases was 39 years (IQR 30-53). The median R 0 was 2·9 for São Paulo and Rio de Janeiro. Cardiovascular disease/hypertension were associated with hospitalization. Co-circulation of six respiratory viruses, including influenza A and B and human rhinovirus was detected in low levels.

Socioeconomic disparity determines access to SARS-CoV-2 testing in Brazil. The lower median age of infection and hospitalization compared to other countries is expected due to a younger population structure. Enhanced surveillance of respiratory pathogens across socioeconomic statuses is essential to better understand and halt SARS-CoV-2 transmission. 

To investigate individual-level diagnostic, demographic, self-reported travel history, place of residence and likely place of infection, differential diagnosis for other respiratory pathogens, as well as clinical details, including comorbidities, we collected case data notified to the REDCap database 8 from February 25 to March 25, 2020. Data was contributed by public health and private laboratories.

Diagnosis and case definitions (see Appendix, pp.1) were based on World Health Organization (WHO) interim guidance. To explore the time-lag between the number of imported cases and of local cases we used the Granger causality test 9 .

Geospatial analysis of COVID-19 cases, demographic and socio-economic data Based on data from the first COVID-19 reports in Brazil 10 , we hypothesized that rates of incidence and testing for COVID-19 are higher in areas of higher per capita income. For the Greater Metropolitan Region of São Paulo (GMRSP), per capita income at the GMRSP neighbourhood level (517 zones) were retrieved from the 2017 Pesquisa Origem e Destino survey (www.metro.sp.gov.br/pesquisa-od/). 13,913 notified cases (COVID-19 confirmed, ruled out, and without final diagnosis) resident in the GMRSP were geocoded based on self-reported address using the Galileo algorithm and verified using Google API. Per capita income for each zone was linked to each notified case based on residential address. We compare per capita income for all notified cases between those tested (positive and negative) and untested, and for confirmed cases by RT-PCR. Full details on the statistical analysis can be found in the Appendix, pp.1.

To quantify transmission potential of COVID-19 in Brazil, an exponential model was used to represent the incidence of COVID-19 at the national level and in São Paulo and Rio de Janeiro states.

Time series of confirmed cases were modelled as samples from a negative binomial distribution with a mean equal to a fixed portion of the incidence. The analysis was carried out in a Bayesian framework with uninformative priors on all parameters apart from the removal rate, which was given an informative prior. The informative prior ensured that the average duration for which an . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted April 29, 2020. 

To investigate which factors are associated with a confirmed COVID-19 result and with hospitalization summary statistics were calculated for continuous variables and for categorical variables and summarized as medians (range and interquartile range, IQR), as appropriate. Missing data were removed (assumed missing at random) (see Appendix Table S1 and Fig. S3 ). Uni-and multivariate analysis included only cases with complete information for the relevant variables. These analyses compared demographics, symptoms, clinical signs and comorbidities between confirmed COVID-19 cases (RT-PCR positive) and ruled-out COVID-19 cases (RT-PCR negative). Additionally, separate multivariate logistic regression models were built to predict hospitalisation (binary variable: hospitalised vs. not hospitalised) based on symptoms, clinical signals and comorbidities, and to predict testing status (positive or negative for RT-PCR SARS-CoV-2). The associations between the outcome and independent variables were reported as adjusted odds ratios (AOR) with 95% confidence intervals and likelihood ratio test (LRT) using the univariate and multivariate logistic regression models. Model diagnostics were performed to check for model specification errors, multicollinearity and influential observations. A 0·05 significance level was applied.

The funder had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all data in the study and had final responsibility for the decision to submit for publication.

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted April 29, 2020. . https://doi.org/10.1101/2020.04.25.20077396 doi: medRxiv preprint

By March 25, 2020, four weeks after the first report of COVID-19 in Brazil, 67,344 COVID-19 cases had been notified as COVID-19 suspected infections from 172 cities across all five administrative regions of Brazil. Of these, 1,468 cases were confirmed (2·18% of all notified cases) and notified through the REDCap system (Fig. 1A) (Fig. 1B) . The epidemic curves of locally-acquired cases followed the curves from imported cases with a lag of two days (Granger causality test) (Fig. 1A) . is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 29, 2020. . https://doi.org/10.1101/2020.04.25.20077396 doi: medRxiv preprint these locations used a Bayesian approach to fit an exponential growth model to COVID-19 aggregated incidence data. Consistent with previous studies in China and overseas 12 , we find that epidemic spread in São Paulo and Rio de Janeiro states is characterized by similar R 0 values of 2·9 (95% CI 2·1-4·4) and 2·9 (95% CI 2·2-4·5). The R 0 for Brazil was slightly higher with median of 3·2 (95% CI 2·4-5·4) (Fig. 2) . Analysis of the age-sex structure of confirmed and notified cases compared to the Brazilian demographic structure revealed a disproportionately lower proportion of confirmed COVID-19

infections reported in younger categories (0-9, 10-19 years of age) and a slightly higher proportion in middle-age categories (20-29 and 30-39 years of age) (Fig. 3) . Specifically, compared to the proportion of the total Brazilian population per age category, the proportion of confirmed COVID-19

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 29, 2020. . https://doi.org/10.1101/2020.04.25.20077396 doi: medRxiv preprint infections in the 0-9 and 10-19 years of age categories are 16·4-and 5·3-fold lower compared to Brazilian demographic structure (Fig. 3A) .

We found that most confirmed cases were in males (776 [54·7%] of 1,420 -46 confirmed cases had missing information for sex and/or age) (Fig. 3A) . The median age of cases was 39 years (IQR, 30-53, range: newborn-93 years). Nearly half (695 [48·9%] of 1,420) of the confirmed cases were in the age range of 20 to 39 years of age (Fig. 3A) . Similarly, 51·6% (2,288 of 4,438) of cases tested for SARS-CoV-2 belonged to this age-group (Fig. 3B) , which is substantially higher than the corresponding fraction of the Brazilian population (68,451,093 [32%] of 211,755,692). 9·5% (133) of cases were health care workers. Overall, only four newborns, three infants (6 to 8 month-old), ten children (1 to 12 years old), and twelve adolescents (12 to 17 years old) were diagnosed with COVID-19. In addition, nine patients were pregnant, one in the first trimester, one in the second trimester, four in the third trimester and 3 had missing information). Six cases were HIV-positive. Proportion (%) of the country's population in each age-sex class is shown as faded bars.

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 29, 2020. (Fig. 4A) . To test whether notified tested cases were associated with socioeconomic status, we evaluated the association between COVID-19 diagnosis and socioeconomic status in the subset of cases in the Greater Metropolitan Region of São Paulo (GMRSP) region with geocoded residential information using an ordinal probit model. We found that the proportion of tested cases in GMRSP increased as income per capita increases (z-score = 0.19, likelihood ratio test P-value <0.01) (Fig. 4C, Table S2 ).

Moreover, the increase in the proportion of tested cases for a unit-increase in income is higher in weeks 2, 3 and 4 compared to week 1. For the range of income per capita observed, given the same amount of income per capita, the proportions of tested cases were lower in weeks 2, 3 and 4 than week 1. Overall, there was a noticeable upwards trend in the association between testing rate and per capita income uncovering a widening socioeconomic disparity in testing practice as the number of cases expands. The income distribution of the untested fraction increasingly approximates the average for GMSP, whereas the tested and confirmed cases (both laboratory and clinical epidemiological) are consistently higher over the study period.

We also analysed the results for other respiratory pathogens tested in Brazil as part of the differential diagnosis by Central Public Health Laboratories and National Influenza Centres (Brazilian , Fig. S4 ). . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 29, 2020. In a univariate analysis of4,387 cases with a final classification as confirmed (n=1,101) or discarded (COVID-19 ruled out) (n= 3,286), we found that increasing age, symptoms (cough, difficulty breathing, dyspnoea/tachypnea, sputum production, nasal congestion, nasal flaring, nausea/vomiting, headache, irritability/confusion, difficulty swallowing, intercostal retraction and Alteration on chest auscultation) and clinical signs (fever and conjunctival congestion) were higher associated with a negative SARS-CoV-2 results (see Appendix Table S4 ). Overall, a total of 12·5% (184/1,468) of confirmed COVID-19 cases had at least one comorbidity. Most common comorbidities were heart disease, hypertension, diabetes, and chronic respiratory disease. (Figure 5B) . Interestingly, age was not significantly associated with hospitalization after accounting for co-morbidities.

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 29, 2020. Of the four fatal cases, one had cardiovascular disease/hypertension, one had both cardiovascular disease/hypertension and renal disease, and two fatal cases had no reported comorbidities. Only one case had reported close contact with a confirmed COVID-19 case reinforcing that local transmission was already well established in Brazil by March 25, 2020.

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 29, 2020. . https://doi.org/10.1101/2020.04.25.20077396 doi: medRxiv preprint

These findings provide evidence that SARS-CoV-2 transmission in Brazil shifted rapidly from a scenario of imported to local transmission. We found that the proportion of tested cases is higher in zones with higher per capita income. We showed that during the first month of COVID-19 in Brazil, only 33·1% of the reported confirmed cases were conducted in public health laboratories. Our results support similar transmission potential (R 0 ) of SARS-CoV-2 in Brazil to other geographic regions. Overall, our clinical findings demonstrate that chest X-ray abnormalities and O 2 saturation <95% are strongly associated with hospitalization. The combination of universal access to diagnostic and the success of interventions will dictate the fate of COVID-19 in Brazil. Overall, these findings 14 . Secondly, our retrospective study has focused predominantly on symptomatic patients (92%) that presented themselves to health services for testing. Therefore, we cannot describe the full spectrum of disease. Population-based serologic surveys are urgently needed to properly determine the asymptomatic and oligosymptomatic fraction. Finally, many patients remained hospitalized when the dataset was extracted, and, we were unable to estimate clinical outcomes given the long duration of infection.

Together with changes in surveillance guidelines, socioeconomic bias in testing suggests that the number of confirmed case counts may substantially underestimate the true number of cases in the population. Additional reasons for underreporting include (i) a significant proportion of asymptomatic infections 15 , (ii) people with mild and even moderate disease are unlikely to present to health services for testing, (iii) limited testing capacity in public health service in Brazil in face of the large number of cases due to delays in importing reagents and kits used in molecular testing.

Close monitoring of state-and municipality-level data will further help to inform mitigation strategies.

Our results suggest that approximately 50% of the COVID-19 cases in Brazil were skewed towards age groups between 20 to 39 years with substantially fewer cases in younger age groups. This pattern could be explained by (i) a higher risk of exposure of this group due to more frequent international travel (travel bans were only implemented on March 23, 2020), and (ii) younger age . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 29, 2020. . https://doi.org/10.1101/2020.04.25.20077396 doi: medRxiv preprint groups being less likely to acquire an infection and/or less likely to acquire significant symptoms upon being infected 16 .

COVID-19 infections were reported in paediatric and pregnant patients [17] [18] [19] . Paediatric infection appears to typically be of mild or moderate severity; we observed a similar proportion of asymptomatic infections compared to reports in 36 children in China (24% vs. 28%) 19 . Also, the onset symptoms of pregnant women were similar to those reported in non-pregnant adults with COVID-19

infection. On the other hand, proportion of hospitalisation of paediatric patients in Brazil was lower than those observed for children in China (3.3% vs. 38.9%) 19 . Also, in contrast to China, none of the pregnant women that tested positive for COVID-19 in Brazil had pneumonia or were hospitalized 17, 18 .

However, the absence/lower number of hospitalisations could be explained by resource availability and local clinical practice guidelines. Despite the small sample size, our findings in pregnant and paediatric patients in the early-phase COVID-19 pandemic in Brazil require further understanding of SARS-CoV-2 infection in these groups.

Although clinical features in Brazil are similar to those recently reported in other countries 1,4,5 , we observed that 8% of confirmed cases reported no symptoms. This should not be considered as an estimate of the asymptomatic fraction. Firstly, it is not possible to distinguish true asymptomatic infections from cases in the pre-symptomatic phase. Secondly, routinely collected data tends to be incomplete. Thirdly, these cases were tested because they were in contact with a known confirmed case. Lastly, there is an ascertainment bias towards symptomatic infections due to the case definition used for notification (Appendix). Other estimates of the asymptomatic fraction have varied widely, including 18% on the Diamond Princess ship 15 , 50-75% in the Italian village of Vo'Euganeo 20 and 31% based on repatriation flight screening 21 .

Overall, 10% of COVID-19 cases in Brazil were hospitalized compared to 19% in the USA 22 . As mentioned above, these differences may reflect factors other than disease severity, for example, resource availability, local clinical practice guidelines and testing availability. On the other hand, they may also reflect right censoring, whereby cases that were notified towards the end of the period studied had not yet been hospitalized. This would be expected given the median lag of four days between symptom onset and hospitalization observed in Brazil. Although age was not a risk factor for hospitalization after controlling for comorbidities, is should be noted that the age distribution among patients who were hospitalized differed from that reported in China, with a higher proportion of younger (<39 years: Brazil, 24.5% vs. China, 10%) and older patients (>70 years: Brazil, 21.8% vs. China, 15%) 18 . However, such comparisons need to be taken cautiously due to different testing and notification practises in the two countries.

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 29, 2020. We showed that patients with pre-existing cardiovascular diseases/hypertension were at increased risk of hospitalization. The prevalence of at least one comorbid condition among infected individuals in Brazil was similar to that reported in China (12.5% vs. 10.5%) 23 . Previous studies suggest that persons with underlying health conditions, such as cardiovascular, diabetes and chronic lung diseases, appear to be at higher risk for severe COVID-19 infection than persons without these conditions 22,24 . Pre-existing cardiovascular disease appears to be particularly important, potentially due to the involvement of the renin angiotensin system signalling pathway 25 .

This study provides new information on co-circulation and co-detection of other respiratory pathogens in the early phase of the COVID-19 epidemic in Brazil. Particularly, we found cocirculation of eight other respiratory viruses, the most common respiratory infections were influenza A and B, and human rhinovirus (HRV). Co-detection of SARS-CoV-2 with influenza A and human metapneumovirus (hMPV) have also been reported in China 26,27 . Here we found co-detection of SARS-CoV-2 with influenza A and hMPV, and we expanded the description of the other multiple codetection scenarios of SARS-CoV-2 with other respiratory viruses, including HRV, influenza B, human respiratory syncytial virus, and other coronaviruses (i.e. coronavirus 229E/NL63, hCoV OC43/HKU1).

Although, viral co-infection has been reported with many other respiratory viruses, no difference in clinical disease severity between viral co-infection and single infection has been reported 28 .

In conclusion, we provide the first description of COVID-19 in Brazil. Our study provides crucial information for diagnostic screening and health-care planning, and for future studies investigating 

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted April 29, 2020. . https://doi.org/10.1101/2020.04.25.20077396 doi: medRxiv preprint

Clinical features of patients infected with 2019 novel coronavirus in Wuhan, China

Coronaviridae Study Group of the International Committee on Taxonomy of V. The species Severe acute respiratory syndrome-related coronavirus: classifying 2019-nCoV and naming it SARS-CoV-2

Genomic characterisation and epidemiology of 2019 novel coronavirus: implications for virus origins and receptor binding

Clinical Characteristics of Coronavirus Disease 2019 in China

First Case of 2019 Novel Coronavirus in the United States

Coronavirus Disease 2019 (COVID-19) in Italy

World-Health-Organization. Coronavirus disease (COVID-2019) situation reports

Research electronic data capture (REDCap)--a metadata-driven methodology and workflow process for providing translational research informatics support

Investigating Causal Relations by Econometric Models and Cross-spectral

Virological assessment of hospitalized patients with COVID-2019

The reproductive number of COVID-19 is higher compared to SARS coronavirus

Percepção do estado de saúde, estilos de vida e doenças crônicas

The Lancet Infectious

Estimating the asymptomatic proportion of coronavirus disease 2019 (COVID-19) cases on board the Diamond Princess cruise ship

Early Transmission Dynamics in Wuhan, China, of Novel Coronavirus-Infected Pneumonia

Clinical features and obstetric and neonatal outcomes of pregnant patients with COVID-19 in Wuhan, China: a retrospective, single-centre, descriptive study. The Lancet Infectious

Clinical characteristics and intrauterine vertical transmission potential of COVID-19 infection in nine pregnant women: a retrospective review of medical records

children with coronavirus disease 2019 (COVID-19) in Zhejiang, China: an observational cohort study. The Lancet infectious

Covid-19: identifying and isolating asymptomatic people helped eliminate virus in Italian village

International journal of infectious diseases : IJID : official publication of the International Society for Infectious Diseases

United States

Characteristics of and Important Lessons From the Coronavirus

The authors thank the clinicians and epidemiologist for technical support. This work was supported