Risk factors associated with delay in diagnosis and mortality in patients with COVID-19 in the city of Rio de Janeiro, Brazil

Resumo Investigamos os preditores de atraso no diagnóstico e mortalidade de pacientes com COVID-19 no Rio de Janeiro, Brasil. Uma coorte de 3.656 pacientes foi avaliada (fevereiro-abril de 2020) e as características sociodemográficas dos pacientes, o bairro e o índice de desenvolvimento social (IDS) foram usados como fatores determinantes dos atrasos no diagnóstico e da mortalidade. Foram realizadas análises de sobrevivência de Kaplan-Meier, modelos de regressão Cox dependentes do tempo e análises de regressão logística multivariada. O tempo mediano desde o início dos sintomas até o diagnóstico foi de oito dias (intervalo interquartil [IQR] 7,23-8,99 dias). Metade dos pacientes se recuperou no período avaliado e 8,3% faleceram. As taxas de mortalidade foram maiores nos homens. Atrasos no diagnóstico foram associados ao sexo masculino (p = 0,015) e pacientes que moravam em áreas com baixo IDS (p < 0,001). As faixas etárias estatisticamente associadas à morte foram: 70-79 anos, 80-89 anos e 90-99 anos. Atrasos no diagnóstico superiores a oito dias também foram fatores de risco para óbito. Atrasos no diagnóstico e fatores de risco para morte por COVID-19 foram associados ao sexo masculino, idade abaixo de 60 anos e pacientes que vivem em regiões com menor IDS. Atrasos superiores a oito dias no diagnóstico aumentam as taxas de mortalidade. Palavras-chave COVID-19, Análise multivariada, Mortalidade Abstract We investigated the predictors of delay in the diagnosis and mortality of patients with COVID-19 in Rio de Janeiro, Brazil. A cohort of 3,656 patients were evaluated (Feb-Apr 2020) and patients’ sociodemographic characteristics, and social development index (SDI) were used as determinant factors of diagnosis delays and mortality. Kaplan-Meier survival analyses, time-dependent Cox regression models, and multivariate logistic regression analyses were conducted. The median time from symptoms onset to diagnosis was eight days (interquartile range [IQR] 7.238.99 days). Half of the patients recovered during the evaluated period, and 8.3% died. Mortality rates were higher in men. Delays in diagnosis were associated with male gender (p = 0.015) and patients living in low SDI areas (p < 0.001). The age groups statistically associated with death were: 70-79 years, 80-89 years, and 90-99 years. Delays to diagnosis greater than eight days were also risk factors for death. Delays in diagnosis and risk factors for death from COVID-19 were associated with male gender, age under 60 years, and patients living in regions with lower SDI. Delays superior to eight days to diagnosis increased mortality rates.


Introduction
The emergence of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) in China at the end of 2019 has caused a sizeable global outbreak. It is a major public health issue responsible for more than 603,697 deaths and more than five million infected people worldwide 1 . The first case of this disease, called COVID-19, in South America, was from a 61-year-old Brazilian diagnosed at the Albert Einstein Hospital in São Paulo (Brazil) after returning from a trip to northern Italy in February 2020 2,3 . Since this episode, more than 2,098,389 cases and 79,488 deaths have been confirmed in Brazil, framing the country in a public health state of emergency of international interest, according to the World Health Organization (WHO) 1 .
The virus is transmitted human-to-human via droplets or direct contact, with a mean incubation period of infection of 6.4 days 4 . Among patients with pneumonia associated with COVID-19, fever is the most common symptom, followed by cough and difficulty breathing. Bilateral lung involvement with ground-glass opacity is the most common finding from computed tomography images of the chest. However, most patients may be asymptomatic and still transmit the virus even before the onset of symptoms 1,4 . Studies demonstrated that the virus is detectable for some time on smooth surfaces, aerosols, and feces 4 .
Currently, the gold standard for confirming suspected cases of the COVID-19 is reverse transcription-polymerase chain reaction (RT-PCR). This test detects viral nucleic acid for laboratory diagnosis 5 . Additional methods, such as serological tests, in which the virus antigens or antibodies produced are investigated, can also confirm the diagnosis 6,7 . Large-scale testing, rapid diagnosis, and immediate case isolation, associated with rigorous screening and preventive measures of social distance, and basic hygiene, are essential procedures to reduce the spread of COVID-19 1,8 .
Rio de Janeiro, the second-largest city in Brazil with important industrial and touristic flows, is one of the areas with the highest number of cases (n > 66,909) and deaths (n > 7,703) by COVID-19 [9][10][11] . In this city, much of the population is concentrated into clusters known as favelas, where the susceptibility to infections is high. This occurs due to the absence or scarcity of basic sanitation (i.e., lack of drinking water, toilets, sewage, garbage collection, and safe housing), as well as space restrictions, overcrowding, and vio-lence, which inhabitants from properly following the COVID-19 preventive measures 4,9,12 . Thus, considering the current pandemic situation worldwide with an important emphasis on underdeveloped regions, our aim was to investigate the risk factors associated of delay in the diagnosis and mortality of patients with COVID-19 in Rio de Janeiro, Brazil.

Study design and data collection
This retrospective cohort study included patients diagnosed with COVID-19 in Rio de Janeiro, located in the southwest region of Brazil, which represents approximately 3.0% of the Brazilian population.
Data were obtained from the Painel Rio COVID-19 database 13 , maintained by the City Hall of Rio de Janeiro. We included all patients (irrespective of age or gender) with a positive diagnosis for COVID-19 from February 27, 2020 to April 26, 2020. Patients' social development index (SDI) were classified according to the addresses of Rio de Janeiro 14 into two groups: low social index and high social index. The time for diagnosis (in days) was estimated by the difference between the date of the onset of the first symptoms of COVID-19 and the date of diagnosis. For the analysis of predictors of diagnosis delays, the independent variables were age group, sex, SDI, and evolution of the disease. The dependent variable was the time to diagnosis.
In our analysis of risk factors associated with COVID-19 mortality, the independent variables were age group, sex, SDI, and time to diagnosis. In this case, the dependent variable, mortality, was binary and categorized as 'dead' or 'alive' . Patients with the active disease were considered alive.

Statistical analyses
A descriptive analysis was performed using the median and interquartile range (IQR) for continuous variables with a non-normal probability distribution, and the frequency for categorical variables.
The Kaplan-Meier method was adopted to estimate the delay in diagnosis. The log-rank test was used to compare the Kaplan-Meier survival curves of the independent variables. Both univariate and multivariate time-dependent Cox re-gression models were used to investigate the risk factors associated with the delay in diagnosis. The hazard ratio 7 and corresponding 95% confidence interval (CI) were used to quantify the size of the effects of risk factors.
In relation to mortality data, the initial objective was to investigate the prognostic factors for the time until death by COVID 19 using the Cox regression survival analysis method, however, given the lack of information on patient's follow-up time (e.g. date from disease diagnosis to death, cure, or loss of follow-up) in the original database, it was not possible to use the method of survival analysis. Therefore, to evaluate COVID-19 mortality risk factors, bivariate analysis was performed using the Custom Table function of the SPSS Software (IBM, USA) a preparatory phase of the multivariate analysis, which allowed for determining the correlation of possible independent variables that may influence the disease outcome. The multivariate analysis of logistic regression was used to investigate the risk factors related with mortality due to COVID-19. In addition, sensitivity analysis (univariate and multivariate) was performed, excluding patients with active disease. The odds ratio (OR) and corresponding 95% CI were used to quantify the size of the effects of risk factors. The chi-squared test was used to investigate associations between the independent and dependent variables. All statistical analyses were conducted in SPSS version 20, and the threshold for significance was set at 5.0% (p < 0.05).
We included all 3,656 patients in our analysis of the time to diagnosis (no censored nor truncated events). The median time to diagnosis was eight days (IQR, 7.236-8.997), with significant differences between males (8.0 days [IQR, 7.487-8.598]) and females (7.0 days [IQR, 6.557-7.595]) (log-rank, p = 0.027). The age of the patients had a significant effect on the time until diagnosis (log rank, p = 0.009). Young patients aged 10-19 years ( Table 2).
Regarding the disease outcome according to patients' gender, more men died from COVID-19 (5.0%; n = 182;) compared to women (3.3%; n = 122). The recovery rate was also higher in women (25.1%; n = 919) than in men (22.5%; n =   (Table 3). There was also a significant association between the disease outcome and age (p < 0.001). Higher death rates were found in older age groups: patients aged 20-29 years died less frequently (0.1%) compared to groups between 70-79 years (1.8 %) and 80-89 years (2.0%). The highest rates of recovery were recorded for patients between 30-39 years of age (11.4%), while these rates were extremely low (0.2%) for patients aged 90-99 years. The age group of 40-49 years had the highest number of patients with active disease (11.5%). A statistically significant association was found between the disease outcome and the patients' SDI (p < 0.001). Although patients living in neighborhoods with a lower SDI had with a higher number of deaths compared to those from a high SDI (3.3% vs. 1.6%), they recovered more often (25.1% vs. 14.5%). The time between the onset of symptoms until diagnosis was also significantly associated with the disease outcome (p < 0.001). A higher mortality rate was recorded for patients with a diagnosis delay greater than eight days (6.0%; n = 220) compared to delays of less than eight days (2.3%; n = 84). Patients diagnosed earlier (< 8 days) had higher recovery rates compared to those diagnosed later (41.1% vs. 7.4%) (  (Table 4).
After conducting the mortality analysis (Table 4), the sensitivity analysis of the logistic re-  (Table 4), where these age groups were not associated with mortality odds of the disease, but it is the age groups from 70 years onwards that have been associated with mortality. The results of the multivariate model of sensitivity analysis were similar with the univariate sensitivity analysis (Table 5).

Discussion
We evaluated data on more than 3,500 patients diagnosed with COVID-19 in Rio de Janeiro, Brazil, for a period of two months, with a resulting mortality rate of around 8.0%. Time from the onset of symptoms to diagnosis was approximately one week, with male patients, younger people, and those living in less developed regions Previous studies on the etiopathogenesis of viral infection and the clinical management of the disease demonstrated differences in the prevalence and severity of COVID-19 according to patients' gender, which may be an important risk factor for mortality [21][22][23] . Epidemiological data on gender are essential to understanding the distribution of risk, infection, and disease in the population, and the extent to which sex affects clinical outcomes comorbidities 21,24 . A report including 552 hospitals from 30 provinces in China, revealed that the majority (58.0%) of patients with COVID-19 were males and also shows that patients of that gender are more likely to contract the disease 22 . This may occur given the higher prevalence of previous chronic diseases in men that enable the virus to develop, and contribute to increased disease severity 24 . Additionally, studies evaluating other SARS caused by coronavirus suggest high levels of estrogen (more prevalent in women) as an important protective factor that may help to control the infection 21,24 . Moreover, women usually have a greater self-care in health when compared to men 25 , which may contribute to a more rapid perception of disease symptoms and, consequently, an earlier diagnosis. In this context, it is important to promote specific measures of prevention, surveillance, and greater intensive intervention for older men with comorbidities 21,24 . Failing to integrate gender differences in COVID-19 surveys can neglect a key risk factor and possibly create or increase inequities in healthcare.
We also found the patients' SDI in Rio de Janeiro was an important variable for delays in diagnosis and a risk factor for mortality. Previous studies showed extensive socioeconomic in-  26 . Individuals with low incomes may postpone the decision to seek healthcare services due to past negative experiences and the inability to miss work 27 . The different regions of Rio de Janeiro vary in demographic and infrastructure characteristics. Residents in regions with a lower SDI are, usually, less literate, younger, and have limited access to basic sanitary services when compared to residents in regions with a higher SDI 9,12 . These determinants are often associated with causing several differences in various traits (e.g., prevalence, severity) of diseases among the population 28 , now including COVID-19 4 . Given this, it is critical that the government introduces further emergence measures, such as improving the healthcare infrastructure, to prevent the virus from spreading in these fragile areas. The current scenario also highlights the need to strengthen access to the Brazilian Unified Healthcare System 6 , in addition to the importance of increasing investments in research, technology, and innovation to effectively combat the pandemic in the country [15][16][17][18][19][20] .
Finally, current data show that Brazil has the second-highest number of confirmed coronavirus cases in the world. At the same time, the country is ranked 19 th in diagnostic tests application for the population 29 . Until May 25, 2020, around 3,460 tests per million inhabitants were performed in the country, compared to 45,586 tests per million inhabitants conducted in the United States, and the 40,000 tests per million inhabitants in Germany and Italy. These data confirm the lack of tests carried out in Brazil, and that cases are underreported. It is estimated that one in 10 positive cases are reported 30,31 .
Our study has some limitations. Although we performed an analysis with a cohort of patients from one of the largest cities in Brazil, representing different socioeconomic levels, it may not reflect the reality from other regions. Some of these data may be underestimated, given the underreporting of cases. Other variables potentially associated with COVID-19 can be evaluated in further studies. The inclusion of patients with active disease in the analysis may represent a bias because some of these cases can die due the progression of the disease. Given some lim-itations from the original database (e.g. lack of information on patient's follow-up), further survival analysis were not possible.

Conclusion
Delays in diagnosis and risk factors for death from COVID-19 in Rio de Janeiro, Brazil were associated with male gender, age under 60 years, and patients living in regions with lower social development index, including favelas. Additionally, a delay between the onset of symptoms until diagnosis greater than eight days may increase mortality rates. These results highlight the continuous challenges to contain, mitigate, and control the disease in environments with scarce resources, including the lack of mass testing. Further preventive measures are still needed, especially for vulnerable groups.