As of December 2019, the first cases of COVID-19, a disease caused by the new coronavirus SARS-CoV-2, have been reported to the World Health Organization (WHO).1 In March 2020, WHO declared a pandemic2 and since then the world is facing an unprecedented challenge. Several regions have been forced to increase intensive care unit (ICU) capacity in a short period of time to deal with an expressive number of critically ill patients.3,4

Latin America followed this trend, and the populous city of São Paulo began to represent one epicenter of the disease in the world, with 90,293 suspected COVID-19 cases (18,438 confirmed)5 and 3622 deaths (1665 confirmed)6 reported by the end of April, 2020.

During a hospital surge, huge increases of ICU capacity may lead to degraded capability and possible modification in the standard of care.7 Studies investigating different strategies to deal with ICU capacity increase are scarce, but important to guide policy makers.

This study aimed to evaluate the impact of staff preparedness by comparing patients outcomes between two ICUs from the same hospital, with the same facilities and similar attended population, one previously well-established and another recently assembled to deal with a markedly increased volume of patients brought by COVID-19 pandemic.

Data was collected through a standardized case record form (CRF) developed by the International Severe Acute Respiratory and Emerging Infection Consortium (ISARIC) and WHO in Research Electronic Data Capture software (REDCap, Vanderbilt University).

We obtained demographic data, comorbidities, vital signs, radiologic findings, laboratory results and instituted therapy within 24 hours of admission. Results of viral respiratory pathogens tested were recorded.

Severity was evaluated using the Pneumonia Severity Index (PSI).11 Data was collected retrospectively through review of hospital charts and the epidemiology service database.

The assessed outcome was in-hospital mortality. Follow up time was censored on June 13, 2020.

Continuous variables were reported as medians and interquartile range (IQR). Categorical variables were summarized as counts and percentages. Missing data were not imputed.

Differences between the two ICUs were assessed using the chi-squared or Fisher's exact or Mann-Whitney test, as appropriate.

Outcome was assessed with unadjusted and multivariate models, using logistic regression. Our model included variables that presented p<0.20 on unadjusted analysis and had no significant correlation with other variables. The model was generated by backward stepwise regression. The fitness of our model was evaluated by the Hosmer-Lemeshow test. Results are presented as odds ratios (OR).

The level of significance was set at 0.05 (two-tailed). Analyses were performed with Statistical Package for Social Science (version 25; SPSS, Chicago, IL).

In this study, 114 cases with severe acute respiratory syndrome admitted to one of the two ICUs were included, 54 (47%) admitted to the well-established ICU-1 and 60 (53%) to the recently assembled ICU-2.

The clinical characteristics of those patients are summarized in Table 1. The median age was 57 years, with a male predominance. More than half of patients had at least one comorbidity (hypertension, diabetes, or obesity).

Compared with the well-established ICU-1, patients in the recently assembled ICU-2 were older (median age of 61.5 versus 54 in well-established ICU-1, p=0.045).

According to PSI, 47 (41%) patients were considered at high risk for death and other adverse outcomes, without significant difference between the two ICU populations (23 patients, 43% in the well-established ICU-1 versus 24 patients, 40% in the recently assembled ICU-2, p=0.779).

Data on respiratory pathogens were available for 112 patients, of whom 86 (77%) had SARS-CoV-2 confirmed by polymerase chain reaction (PCR) and 4 (4%) had influenza detected by PCR. There was no confirmed mixed infection by these two viruses. One patient had confirmed influenza B and a Streptococcus pneumoniae bacteraemia. 21 (19%) patients had no pathogen identification.

Within 24 hours of admission, most patients received an antibacterial agent (105, 92%) and a neuraminidase inhibitor (91, 80%). Experimental agents chloroquine or hydroxychloroquine were administered to 19 (18%) patients. Corticosteroids were prescribed to 7 (6%) patients.

There were significant differences on therapy institution between the two ICUs: antibiotics were administered more frequently in the recently assembled ICU-2 (59 patients, 98% versus 50 patients, 93%, p=0.001), as well as chloroquine/hydroxychloroquine prescription (3 patients, 6% in ICU-1 versus 18 patients, 30% in ICU-2, p=0.001).

Supportive care with invasive mechanical ventilation was needed for 55 (48%) patients, with a higher proportion in the recently assembled ICU-2 (24 patients, 44% in ICU-1 versus 31 patients, 52% in ICU-2, p=0.008).

All patients had at least 30 days of follow-up. As of June 13, 2020, 63 (55%) patients had been discharged from hospital, 3 (3%) remained in hospital and 46 (40%) patients had died. The median length of hospitalization was 11 days (IQR 6 – 21), excluding the three patients remaining in hospital by the end of follow-up.

In-hospital mortality rate was 40%, greater in the recently assembled ICU-2 than in the well-established ICU-1 (30% in ICU-1 versus 50% in ICU-2, p=0.017).

Age ≥ 60 years, mechanical ventilation, and ICU type (recently assembled) were risk factors for in-hospital mortality (Table 2). Since elderly is a well described risk factor for mortality by COVID-19,12,13 we categorized age (cut-off: 60 years).

Comorbidities is a variable that presented p < 0.20 on unadjusted analysis and was tested in the multivariate model, but it was not independently associated with the outcome (p = 0.55) and acted as a confounding factors with age. Therefore, this variable was excluded from the final model.

As age is a component of the PSI calculation, we believe that if we had considered PSI and age in the same multivariate model, the PSI would act as a confounding factor, distorting the association between exposure (age) and outcome.

We decided to not include chloroquine or hydroxychloroquine and corticosteroid in the multivariate model, since they were prescribed for just a few patients.

In multivariate analysis, age ≥ 60 years (adjusted OR 2.33; 95% CI 1.02-5.31), mechanical ventilation need (adjusted OR 2.79; 95% CI 1.22-6.37), and ICU type (recently assembled) (adjusted OR 2.38; 95% CI 1.04-5.44) remained independently associated with in-hospital mortality.

In this study, age ≥ 60 years, need of invasive mechanical ventilation and ICU type (recently assembled) were independently associated with in-hospital mortality among patients with severe acute respiratory syndrome during the COVID-19 pandemic in São Paulo, Brazil.

São Paulo, Brazil's largest city, experienced a rapid and unprecedented growth in the need for ICU beds, leading to increased hospital capacity in a short period of time. ICUs of public hospitals of São Paulo reached 90% capacity, with risk of collapse of the health system.

In this study, the ICUs mortality rate was 40%. Coincidentally, a recent systematic review and meta-analysis of 24 observational studies carried-out in Asia, Europe and North America including 10,150 patients with COVID-19 who were admitted to ICUs reported a mortality rate of 41.6%, with no significant difference across geographic locations.14

There is scarce information about outcomes of patients admitted to ICUs in Latin America during COVID-19 pandemic. The Brazilian Intensive Medicine Association created a national registry to characterize the epidemiologic profile of the Brazilian ICUs. In this project, among 33,161 patients with COVID-19 admitted to Brazilian ICUs, between March 1 and July 22, 2020, the in-hospital mortality rate was 34.4% which was remarkably different in private (27.8%) and public (50.4%) hospitals.15 These differences were due in part to the fact that in public hospitals patients usually arrive in worse clinical condition and the threshold for admission to an ICU are usually stricter. Results of the present study suggest that better outcomes are possible in public hospitals.

The variables age ≥ 60 years and need of invasive mechanical ventilation were identified in our study as independent factors associated with mortality. A systematic review of 14 observational studies including 4,659 hospitalized patients with COVID-19 identified older age and baseline cardiometabolic diseases (particularly hypertension, diabetes and chronic heart disease/cardiovascular disease) as risk factors associated with mortality.12,13 As with other diseases, patients with COVID-19 admitted to ICUs are more likely to undergo invasive mechanical ventilation and be intubated and are more likely to die.

Interestingly, even though patients in the recently assembled ICU-2 were older and had an increased need for mechanical ventilation when compared with patients in the well-established ICU-1, ICU type was independently associated with in-hospital mortality, more than doubling the risk for those individuals admitted in the recently assembled ICU.

As there is no validated severity score for patients with COVID-19, we have chosen PSI for its broadened evaluation. Although patients in ICU-2 were older, had a major male predominance and had more comorbidities, the rate of patients with high risk for death and other adverse outcomes according to PSI did not show significant differences between ICUs, indicating that probably patients in ICU-1 had worst clinical condition, laboratory and radiologic findings than patients in ICU-2. As the patient allocation to each ICU was random, according to the availability of ICU beds, we expected the population to be similar, as well as the treatment received in the ER.

Our study has the differential of being able to explore the impact of staff preparedness on clinical outcome, since in our institution the strategy to deal with the COVID-19 emergency was to hire an external ICU medical team to take care of patients in the ICU beds that had been expanded.

Although we could expect that an experienced staff ought to have better treatment skills, our results shed light on the importance to take this variable into account when studying the possible determinants of death during an emergency.

For hospitals to effectively respond to the anticipated surge in critical care patients, the balance between space (ICU beds), staff (nurses, physicians, and respiratory therapists), and stuff (i.e., ventilators) needs to be maximized.16

Significant increases in the normal capacity of an ICU may occur, but at the cost of reducing performance and a possible modification in the standard of care.7

There are reports of increasing ICU capacity by utilizing alternate hospital sites and non-ICU staff under the supervision of trained critical care personnel.7

Our findings can inform policy makers that the capacity of trainning and supervising recently assembled teams must be strengthened, possibly by establishing mentorship and evaluation programs, besides developing detailed and clear treatment protocols.

In our institution, a clinical management protocol was released with weeks of delay from the first COVID-19 case, what could have contributed to discrepancies between medication usage between the ICUs.

Although this study was carried out before evidence showing benefit of using corticosteroids in COVID-19 cases who require supplemental oxygen, our CRF had a section for collecting information on the use of medications and already included a specific field for corticosteroids.

This study has some limitations inherent to observational monocentric design. However, we included consecutives cases, patients were rigorously selected, and data collection used an ISARIC tool. Despite having been carried-out in a single center, this study conducted a strict comparative assessment of the ICU type under similar hospital conditions.

Since no randomization tool was applied, there was a potential for selection bias which could have lead to the baseline characteristics differences between patients assigned to the two ICUs. The multivariate analysis adjusted for the most significant differences, but certainly randomized and multicenter studies could bring more strength to our findings.

Regardless of the absence of more detailed analysis of severity between the groups, we consider that the results shown may be useful, especially for scenarios like ours. Finally, in this study we evaluated an initial period of the pandemic and it is possible that better outcomes would be obtained over time,14 particularly in the recently assembled ICU, with standardization of clinical protocol, in addition to continuous education of the professionals.

In conclusion, we found that age ≥ 60 years, need of invasive mechanical ventilation, and ICU type (recently assembled) were independently associated with in-hospital mortality among patients with severe acute respiratory syndrome during the COVID-19 pandemic. This finding highlights the importance of developing support strategies for improve preparedness of staff recently assembled to deal with emergencies.

Dr. Figueiredo-Mello, Dr. Sztajnbok, Dr. Freitas Ribeiro, Dr. Malaque, and Dr. Vidal has contributed to the conception and conduction of the work.

Dr. Freitas Ribeiro has conducted the data analysis and interpretation.

Dr. Figueiredo-Mello, Dr. Sztajnbok, Dr. Freitas Ribeiro, Dr. Malaque, and Dr. Vidal has written this manuscript.

Dr. Figueiredo-Mello, Dr. Cavalin, Dr. Lanza, Dr. Castanheira, Dr. Rego, Dr. Custodio, and Dr. Siqueira has contributed with data collection

Dr. Figueiredo-Mello has been the main advisor of the work, from conception of the study to revision of this manuscript.