Risk factors associated with mortality in patients hospitalized for coronavirus disease 2019 in Rio de Janeiro, Brazil

Abstract INTRODUCTION: Understanding the mortality-associated risk factors of coronavirus disease 2019 will impact clinical decisions. METHODS: This retrospective longitudinal study included patients hospitalized for coronavirus disease in Rio de Janeiro, Brazil. The Kaplan-Meier method and multivariate Cox regression analysis were used. RESULTS: Sequential Organ Failure Assessment score of ≥2 (hazard ratio 4.614; 95% confidence interval =2.210-9.634; p<0.001) and neutrophil/lymphocyte ratio of >5 (hazard ratio=2.616; 95% confidence interval=1.303-5.252; p=0.007) were independently associated with mortality. CONCLUSIONS: Sequential Organ Failure Assessment score and neutrophil/lymphocyte ratio on admission can identify coronavirus disease patients at increased risk of death and guide subsequent clinical decisions.

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is a viral pathogen that rapidly caused a devastating pandemic of coronavirus disease 2019 (COVID-19). As of December 12, 2020, SARS-CoV-2 has caused 71,704,885 infections and 1,604,978 deaths worldwide 1 . On February 26, 2020, the Ministry of Health confirmed the first case of COVID-19 in Brazil. Almost a year after the pandemic, Brazil has the third-highest number of confirmed cases and the second-most deaths in the world 1-2 .
Even though the majority of patients who develop COVID-19 have mild symptoms 3 , about 20% of hospitalized patients are admitted to an intensive care unit (ICU), 15% require mechanical ventilation (MV), and up to 20% of hospitalized patients die 4,5 . Studies on risk factors for intubation and mortality in patients hospitalized for COVID-19 have largely focused on patients from China, Europe, and the United States of America 6 .
In contrast, in Latin America, the hospitalized patients' characteristics, comorbidities, presenting symptoms, laboratory parameters, and clinical outcomes have not been thoroughly investigated. Considering that it has not yet been clarified whether phenotypic and genotypic characteristics of the population cause differences in the response to SARS-CoV-2 infection 7 , understanding the factors associated with mortality in patients with COVID-19 could impact clinical decisions and guide public health policies in Latin American countries.
We conducted a retrospective longitudinal study that included adult patients admitted to Hospital Casa Rio-Botafogo with suspected or confirmed COVID-19, according to the clinical, radiological, and laboratory criteria of the Ministry of Health-Brazil 8 , and those who died or were discharged between March 4, 2020 and June 21, 2020. This reference hospital was a private one that was located in the city of Rio de Janeiro, Brazil.
Using a form prepared for the study that was validated with the hospital file, general information was obtained from the reports of the Emergency Unit, and variables recorded during the physical examinations and laboratory tests performed on admission were also collected. In addition, information on hospitalization was obtained, including length of stay, admission to the ICU, treatment, use of MV, complications, and outcome (discharge/death).
The final number of patients included in the study was 98 after excluding five patients due to incomplete information, and the missing data were not imputed. For the descriptive analysis, continuous variables are presented as means with standard deviations (SDs) or as medians with interquartile ranges (IQRs), as appropriate. Univariate analyses to identify variables associated with the outcome (discharge/death) were performed using the Chi square test, Fisher exact test, student's t-test, or Mann-Whitney U test, as appropriate.
The Kaplan-Meier method was used for the survival analysis. The time calculated from the date of hospital admission to the date of death or last day of hospitalization was considered as a dependent variable. The log-rank test was used to compare the survival curves, and the Cox proportional-hazards model was used to identify the baseline variables that were independently associated with death at hospital admission after adjustments for age, sex, and comorbidities. Schoenfeld residue analysis was performed to verify the proportionality of the risks, and Cox-Snell residue analysis was used to adjust the models. The analyses were conducted using STATA version 15.0 (StataCorp, College Station, TX, USA) and statistical significance was set at p<0.05.
The study was approved by the Ethics Research Committee of Hospital Casa Rio-Botafogo in July 2020 and the CONEP Brazil Platform under CAAE number 47885515.8.0000.5279. All procedures were performed in accordance with the principles of Declaration of Helsinki, 1964Helsinki, , as revised in 1975Helsinki, , 1983Helsinki, , 1989Helsinki, , 1996Helsinki, , and 2000 More than half of the patients (51%) were men aged >70 years, and the majority (91%) had at least one comorbidity, with the most frequent being hypertension (72%), diabetes mellitus (30%), obesity (21%), and stroke (19%). A considerable proportion lived in nursing homes (20%) and had been previously hospitalized in the last 3 months (21%) ( Table 1).
A higher proportion of statistically significant deaths was found in older patients (p = 0.039), those with a higher Charlson comorbidity index (p = 0.018), and those with a history of stroke (p = 0.022). However, we also found a higher proportion of deaths in patients who lived in nursing homes (60%, 12/20).
SARS-CoV-2 infection was confirmed in the laboratory by real-time reverse transcription polymerase chain reaction (RT-PCR) in 58 patients (59%). Co-infection with influenza A (H1N1) virus was confirmed in two 75-year-old patients who died. In addition, a 53-year-old woman with human immunodeficiency virus (HIV) infection, but no other comorbidities, died.
Rio de Janeiro state has the fourth-highest number of COVID-19-confirmed cases in Brazil, behind only the states of São Paulo, Bahia, and Minas Gerais. However, in terms of absolute number of deaths, Rio de Janeiro ranks second, below the state of São Paulo 2 . To the best of our knowledge, this is the first study to evaluate the factors associated with mortality in patients hospitalized for COVID-19 in Rio de Janeiro, Brazil.
The baseline characteristics associated with mortality observed in this study were in line with the findings of other studies, such as a higher proportion of deaths in older patients, higher Charlson comorbidity index value, and history of stroke 3,9-10 . We also found a higher proportion of deaths in patients who lived in nursing homes (60%), in accordance with studies carried out in Brazil and other countries, which highlighted the high vulnerability of patients residing in nursing homes 11 .
All the clinical characteristics and laboratory parameters assessed have already been analyzed in several studies, although there is still no consensus on which are the most important [12][13] . We found that both leukocyte and neutrophil counts were significantly higher in the most severely affected patients who died, which may have been related to a secondary infection or infection-induced cytokine storm. In addition, these patients also had lower lymphocyte counts, although this was not significant. Lymphopenia is a common feature of many viral infections and may result from the direct infection of lymphocytes or cell apoptosis; therefore, monitoring the NLR in patients with COVID-19 has been suggested 12 .
We also found significantly higher levels of biomarkers of tissue and organ damage, such as LDH, AST, and urea. It has been postulated that this association could be explained by the virus causing direct damage to the organs by binding to angiotensin-converting enzyme 2 receptors; this leads to systemic hyperinflammation caused by a cytokine storm or hypoxia that results from respiratory failure 12 . In addition, as expected, due to the severity of the disease, higher APACHE II and SOFA scores, and lower values on the Glasgow scale were found in patients who died when compared to survivors.
Regarding the identified coinfections, in the literature, a higher frequency of complications and deaths have been described in critically ill patients with COVID-19 who had coinfection with influenza virus 14 . However, the same has not been reported in the studies published to date on coinfection with HIV 15 .
Despite having a small sample of patients, we found no statistically significant differences regarding the use of hydroxychloroquine, macrolides, ivermectin, low-molecular-weight heparin, or corticoids between the patients who survived and those who did not. Some of these observations were inconsistent with the findings of other studies, such as the protective effect of anticoagulants on the outcomes of patients with COVID-19.

6/7
Likewise, a greater use of antibiotics, although not statistically significant, was found in the group of patients who died. Ceftriaxone was used more frequently in both groups; however, carbapenems, linezolid, and polymyxin B were more frequently used in the group of patients who died. This may have been a reflection of the presence of secondary infections in this group of patients, especially in patients with ventilation-associated pneumonia.
This study showed that a SOFA score of ≥2 and an NLR of >5 on admission were independently associated with mortality in patients hospitalized for COVID-19, as has already been shown in studies from China, Europe, and the United States 12-13 . In future research, these parameters may serve to establish scores that will allow an initial assessment of patients with COVID-19 to provide timely care.
On the other hand, taking into account that the older adults made up almost the majority of the population in this study, we characterized a significant high-risk group that can benefit from stricter social distancing, especially when the restrictions due to COVID-19 are relaxed.
The present study had some limitations. First, this was a singlecenter study with a relatively small sample size, which was limited to the information recorded in the reports; therefore, some variables were not included in the analysis and their roles could have been underestimated. Second, patients who had incomplete information and who remained hospitalized until the end of the study period were excluded (data censored), which could have had an impact on the estimates. Third, the dynamic changes in laboratory parameters and their associations with mortality were not evaluated, and the clinical frailty of the patients was not evaluated. Finally, although the patient inclusion criteria were based on the clinical, radiological, and laboratory criteria of the Ministry of Health in Brazil, only 59% of the patients were confirmed by real-time RT-PCR, which could be linked to late hospitalization.

ACKNOWLEDGMENTS
This study used data from medical records that were judiciously updated by courageous COVID-19 front-line emergency doctors and intensive care staff in a moment of widespread panic over an unknown disease. We would like to thank all those involved for their great contribution. Finally also thank the unit's coordinators who supported and authorized this study.