A retrospective cohort study of 238,000 COVID-19 hospitalizations and deaths in Brazil

The coronavirus disease (COVID-19) pandemic has overwhelmed health care systems in many countries and bed availability has become a concern. In this context, the present study aimed to analyze the hospitalization and intensive care unit (ICU) times in patients diagnosed with COVID-19. The study covered 55,563 ICU admissions and 238,075 hospitalizations in Brazilian Health System units from February 22, 2020, to June 7, 2021. All the patients had a positive COVID-19 diagnosis. The symptoms analyzed included: fever, dyspnea, low oxygen saturation (SpO2 < 95%), cough, respiratory distress, fatigue, sore throat, diarrhea, vomiting, loss of taste, loss of smell, and abdominal pain. We performed Cox regression in two models (ICU and hospitalization times). Hazard ratios (HRs) and survival curves were calculated by age group. The average stay was 14.4 days for hospitalized patients and 12.4 days for ICU patients. For hospitalized cases, the highest hazard mean values, with a positive correlation, were for symptoms of dyspnea (HR = 1.249; 95% confidence interval [CI], 1.225–1.273) and low oxygen saturation (HR = 1.157; 95% CI 1.137–1.178). In the ICU, the highest hazard mean values were for respiratory discomfort (HR = 1.194; 95% CI 1.161–1.227) and abdominal pain (HR = 1.100; 95% CI 1.047–1.156). Survival decreased by an average of 2.27% per day for hospitalization and 3.27% per day for ICU stay. Survival by age group curves indicated that younger patients were more resistant to prolonged hospital stay than older patients. Hospitalization was also lower in younger patients. The mortality rate was higher in males than females. Symptoms related to the respiratory tract were associated with longer hospital stay. This is the first study carried out with a sample of 238,000 COVID-19 positive participants, covering the main symptoms and evaluating the hospitalization and ICU times.

On December 31, 2019, the World Health Organization (WHO) was informed of several cases of pneumonia of unknown etiology, detected in several people in the city of Wuhan, China. In February 2020, the virus responsible for this condition was identified and named severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), which causes a disease known as coronavirus disease 2019 (COVID-19) 1 . On March 11, 2020, the WHO declared a public health emergency of international interest due to the high transmissibility of the virus 2 .
The outbreak likely started from consumption of contaminated animals or human contact with the Wuhan Seafood Market 3 . However, some researchers refute this theory because individuals from different continents who were not involved in the above situations were also infected with the virus 4 . A very recent investigation ordered by the US administration in May 2021, which could bring us closer to a definitive conclusion on the origins of the virus that has killed more than 4 million people globally and wrecked national economies, has reported inconclusive results on whether the virus jumped from animals to humans as part of a natural process or might have accidentally escaped from a Wuhan laboratory in central China 5,6 . Despite these doubts about the origin of the novel coronavirus, the most accepted explanation of transmissibility is that this virus spreads www.nature.com/scientificreports/ in the Brazilian SUS. We covered the parameters of age group (eight groups covering age subgroups from 0 to > 85 years), main symptoms, and hospitalization or ICU time. Therefore, this study aimed to analyze the relationship between hospitalization or ICU time and COVID-19 deaths in Brazil. This was a retrospective study based on the secondary database available from the Brazilian Ministry of Health between February 22, 2020, and June 7, 2021. The database covers national centers and multicenter records that link with the Brazilian SUS.

Methods
Ethical approval. Our study used information from a public database made available by the Brazilian government through the DATASUS platform. All data in this database were processed by the Brazilian government, removing all identifiers. The research was not sent to the Research Ethics Committee as it was covered by Article 1, Sole Paragraph, subparagraph II and III of RESOLUTION No. 510 of APRIL 7, 2016, of the Brazilian National Health Council: "Sole paragraph: They will not be registered or evaluated by the research ethics committee system: II-research using publicly accessible information, according to Law No. 12,527 of November 18, 2011; III-research using public domain information. The resolution is available at http:// conse lho. saude. gov. br/ resol ucoes/ 2016/ Reso5 10. pdf. All methods were performed in accordance with the relevant guidelines and regulations.
Study design, setting, data sources, and participants. We performed a cohort, retrospective, national, multicenter, analytical study with secondary data from the Brazilian Database of Severe Acute Respiratory Syndrome (SARS), which includes cases of COVID-19 32  Variables. The following 12 symptoms were the independent variables: fever, dyspnea, low oxygen saturation (SpO 2 < 95%), cough, respiratory distress, fatigue, sore throat, diarrhea, vomiting, loss of taste, loss of smell, and abdominal pain. A categorical variable of eight age groups was also added, following the pattern adopted by the Centers for Disease Control and Prevention 34 : 0-17, 18-29, 30-39, 40-49, 50-64, 65-74, 75-84, and ≥ 85 years. This variable was calculated based on a patient's age at admission. Study size and missing data. We used predictors for two outcomes: hospitalization and ICU time (days).
The database contained 639,405 records. Hospitalization time was calculated as the difference between the hospitalization and discharge dates. For this model, the sample size was 238,075 records, 401,330 records were excluded because of missing data points. This missing data included a lack of filled-in data, such as the patient's admission or discharge dates.
ICU time is the interval between ICU entrance and discharge. The sample size was 55,562 records, with 583,842 records excluded because of missing data and a lack of ICU admission. Statistical analysis methods. We analyzed hospitalization and ICU time in days. Thus, we performed two models, the first with hospitalization time and the second with ICU time.
We report continuous variables with means and standard deviations. We calculated the proportions of dichotomous variables. Thereafter, Cox regression was calculated (hospitalization and ICU time), using the enter method and IBM SPSS Statistics 23 (https:// www. ibm. com/ suppo rt/ pages/ downl oading-ibm-spss-stati stics-23). The observed event was death during hospitalization or ICU stay. All variables with P > 0.05 were excluded from the analysis. We calculated hazard and survival functions for each age group. The mean coefficients of the equation, survival, and hazard for each independent variable were analyzed. We also calculated the cumulatedhazard. It indicates how the death risk increases or decreases depending on hospitalization or ICU time. In this way, we measure how high or low the death risk is while the patient remains hospitalized. The cumulated hazard demonstrates that more days hospitalized can represent an increased or reduced risk of death. The study design is shown in Fig. 1. Table 1 shows the data summary of patients, recoveries, and deaths. This retrospective analytical cohort study identified factors and mortality risks that interfere with the length of hospital and ICU stay in individuals with COVID-19. Of the 238,075 hospitalized patients, 161,150 (68%) recovered, while 76,925 (32%) died.
Survival curves showed distinct patterns between age groups. In the ICU, the probability of death depends on the patient's age. Younger patients were more resistant to prolonged hospital stays than elderly patients. Despite this, the 30-39 years age group had better hazard and survival curves than the 18-29 years age group. The cumulative hazard and survival curves are shown in Fig. 3 The daily mean survival probability decreased by 3.27% per day during the first 15 days. Patients who stayed for more than 14 days in the ICU had less than 50% probability of survival. Survival declined more sharply depending on age group. On the tenth day, patients aged ≥ 85 years had a 50% chance of survival. The survival and cumulative hazard by age group and day are shown in Table 5.

Discussion
COVID-19 infects individuals from all age groups. The transmission and its consequences may vary according to individual biological factors, environmental and socioeconomic characteristics, public policies developed, and the health system's capacity 35 . Several studies have described factors contributing to transmission, clinical presentation, mortality, possible treatments, and short-term results 26 . Some signs and symptoms can be interpreted as risk factors, indicating the need for hospitalization or transfer to the ICU.  36 reported some prominent differences, namely, a higher rate and length of hospitalization, gender and age groups most affected, need for invasive mechanical ventilation, length of stay in the ICU, and mortality rate.
In the first reports of COVID-19, age group was an identified risk factor. Older people were more affected, had more extended hospital stays, greater clinical severity, and high mortality 18,37 . Our results also indicated that older age groups were associated with extended hospital and ICU times. This may be related to the vulnerability of this population 38 , which includes immunosenescence and changes in the respiratory system, such as decreased respiratory capacity and production of surfactants 39 . Our results also converge with some studies 35,37,40 by showing that the population most affected by COVID-19 is between 50 and 64 years of age. Recently, a worldwide survey identified that approximately 72% of confirmed cases of SARS-CoV-2 infection were aged ≥ 40 years 41 . In addition to age, comorbidities are another factor associated with hospitalization and ICU time, such as cardiovascular diseases, hypertension, diabetes, chronic respiratory diseases 42 , and smoking status 43 . We did not include comorbidities in our study, but it is noteworthy that after age 50, 67.8% of people have multimorbidity 44 . During the initial epidemic in Hubei, Gao 45 reported that 58% of patients admitted to the ICU had at least one comorbidity, which increased the probability of progression to severe forms of the disease.  www.nature.com/scientificreports/ In addition to high age group, many studies identified that affected individuals were predominantly male 45 and that the mortality rate was also higher in this population 46 . Research has identified that males were affected most in the SARS and MERS epidemics 47,48 . Thus, the results presented in this study converge with these findings, considering the predominance of males in hospitalization and ICU admission and mortality. It has been reported that differences in the levels and types of male and female sex hormones influence the susceptibility to infection by COVID-19 because sex hormones modulate adaptive and innate immune responses 49 . Thus, the reduced susceptibility of females to viral infections can be attributed to the protection from the X chromosome and sex hormones 50 , especially estrogen 46 .
In terms of identifying percentage changes in mortality rate due to specific comorbidities, Wu and McGoogan 51 identified a 10.5% increased mortality rate in individuals with cardiovascular diseases, 7.3% for those with diabetes, 6.3% for those with chronic respiratory diseases, and 6% for those with hypertension. In concordance with the information prior, we identified that respiratory symptoms, such as respiratory discomfort, dyspnea, and SpO2 < 95%, correlated positively with length of stay, abdominal pain, and vomiting.
In Lombardy (Italy), the epicenter of the first outbreak of COVID-19 in a Western country, respiratory syndrome was present in 40% to 96% of ICU patients 31 . However, a study 52 highlighted that the most common and persistent symptoms in cases of long-term COVID-19 in patients aged 5 to 17 years were fatigue, headache, and loss of smell. Our study identified cough, fever, and respiratory discomfort/decompensation as the most common symptoms in cases of both hospitalization and ICU admission.
Concerning ICU time and survival, Blake et al. 53 found that the mean time interval from the first report of symptom onset to admission to the ICU was 10 days and, with respiratory decompensation, invasive mechanical ventilation was required in almost 75% of cases, resulting in intubation in 76% of these individuals after the first 24 h after admission to the ICU. Research in China identified that 39% of patients hospitalized for 28 days died and 97% of individuals who needed mechanical ventilation had the same outcome 31 .
Our study identified that survivors spent 10 days hospitalized, while death occurred 14 days after admission. In addition, we found that patient survival decreased by 2.27% per day of hospitalization. The larger chance of death during hospitalizations occurred after 10-14 days in the ICU. The most significant resistance to death was observed in younger patients. Our results agree with the findings in the literature that indicates that survival for older adults reaches 50% after the thirteenth day of hospitalization.
Some researchers have suggested that COVID-19 inhibits the body's cellular immune function, causing damage to T lymphocytes 54,55 . Thus, the low absolute value of lymphocytes could be used as an evaluative index or a reference score in the diagnosis of new COVID-19 infections in clinic settings.
However, the hospitalization of patients, especially older adults, or people with comorbidities, who contract COVID-19 is one of the signs of disease severity as approximately 7.2% of hospitalized patients die. Similarly, the need to be admitted to the ICU has become a highly relevant means of assessing the impact of COVID-19 on public health worldwide.
The average length of hospitalization and ICU stay may have been influenced by the readiness of the public health service in Brazil. At the beginning of the pandemic period, especially during the first wave, there was reduced availability of beds and tests for the Brazilian population, which may have contributed to the delay in the treatment of patients and compromised the health system's effectiveness.

Conclusions
The heterogeneity of both the combination and severity of COVID-19 symptoms varies individually, hindering diagnosis and efficient clinical management. Identifying symptomatological patterns and clinical outcomes can guide more consistent approaches for reducing death and sequelae. The study presented a pattern between patient age, symptom groups, and hospital stay. The results showed that symptoms associated with the respiratory system, namely dyspnea and SpO2 < 95%, were positively associated with a more extended total hospital stay. There was a positive correlation between respiratory discomfort and abdominal pain with respect to a patient's time in the ICU. We emphasize that a patient's probability of survival decreases significantly with each day of hospitalization, with a high chance of death when patients spend more than 14 days in the ICU or 20 days in the hospital, with younger patients showing better resistance. These data bring together potential warning symptoms for developing severe or critical conditions due to COVID-19 infection during the pandemic. Signs or symptoms suggestive of respiratory distress are associated with more severe disease progression and a worse prognosis. These conditions become latent if associated with age. Another relevant clinical scenario is the association between respiratory and digestive symptoms, especially abdominal pain. Intestinal involvement has been identified in autopsies and biopsies of patients with COVID-19 and some clinical reports have described abdominal discomfort as a primary symptom. These data can serve as a warning for the Delta and Ômicron variants of COVID-19. That is, patients with digestive symptoms, even without respiratory symptoms, should receive greater attention, especially if they report a history of epidemiological exposure.
Information was collected from a database provided by the Ministry of Health, Brazil. Since February 2020, the health crisis caused by the pandemic has highlighted the problems arising from investment in public health, which has undergone successive cuts in recent years. The different clinical manifestations and outcomes emphasize the importance of adhering to hygienic and preventive principles and highlight the importance of finding and developing new diagnostic approaches and therapeutic options sensitive to the action of SARS-CoV-2. In the current context of a global pandemic and public health emergency caused by COVID-19, knowledge of the protective and risk factors for developing COVID-19 infection or developing severe forms of the disease are essential for the proper management of patients in the early stages of the disease. This study had limitations; the set of variables used did not encompass all variables associated with hospitalization and ICU time. Furthermore, the high proportion of missing data (63%) was a limitation of the routine national surveillance system. In addition, the research reflects the behavior of a specific country in COVID-19 treatment, with no statistical analysis performed to show a causal relationship between the variables of interest.
Future research must be conducted to broaden the discussion on the relationship between COVID-19 parameters and variables, such as prior comorbidities and sociodemographic characteristics of the population, to evaluate their impact on the patients' time in hospital. Moreover, researchers must discuss the formation of public policies that prioritize democratic and equal access to health care systems, especially in countries like Brazil that have been through substantial economic and political instability, which have been worsened by the present health crisis. Health equity remains a significant challenge for policymakers and academics in this area. A recent challenge is to reduce mortality arising from new COVID-19 variants that continue to emerge across the globe, making it necessary for surveillance to identify them and studies to determine their transmissibility and severity. Finally, it is necessary to adopt public policies aimed at flattening the COVID-19 curve, which requires a combination of strategies to slow the spread of COVID-19 and spread out the epidemic's peak, preventing hospitals from reaching capacity and mortality.

Availability of supporting data
The datasets used and analyzed during the current study are available from the corresponding author on reasonable request. The datasets generated and/or analyzed during the current study are available in the Open Datasus website repository, [https:// opend atasus. saude. gov. br/ organ izati on/ minis terio-da-saude].