Clinical Features and Outcomes of Critically Ill Patients with Coronavirus Disease 2019 (COVID19): A Multicenter Cohort Study

Background A novel coronavirus, named Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) causing coronavirus disease-19 (COVID-19) manifested by a broad spectrum of symptoms, ranging from asymptomatic manifestations to severe illness and death. The purpose of the study was to extensively describe the clinical features and outcomes in critically ill patients with COVID19 in Saudi Arabia. Method A multi-center, non-interventional, observational study for all critically ill patients aged 18 years or older who are admitted to intensive care units (ICUs) between March 1 st to August 31 st , 2020 with an objectively conrmed diagnosis of COVID19. The diagnosis of COVID19 was conrmed by Reverse Transcriptase – Polymerase Chain Reaction (RT-PCR) on nasopharyngeal and/or throat swabs. Multivariate logistic regression and generalized linear regression were used. We considered a P value of < 0.05 statistically signicant.


Introduction
A novel coronavirus, named Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) causing coronavirus disease-19 (COVID-19) emerged in China, in late 2019 1 . Shortly afterward, and due to the virus's extensive spread to nearly all countries, the WHO announced the COVID-19 outbreak as a pandemic on March 11, 2020 2 . COVID-19 is manifested by a broad spectrum of symptoms, ranging from asymptomatic manifestations to severe illness and death. In order to relieve these symptoms, COVID-19 is currently managed by certain antiviral medications, and in extreme cases, supportive treatments, including supplemental oxygen and mechanical ventilation 3 .
On March 2, 2020, the rst con rmed case of COVID-19 was announced in Saudi Arabia, and by 28 October 2020, the MOH had reported a total of 345,631 con rmed cases with a case fatality rate of 0.86% 4 . Owing to the implementation of successful healthcare policies, the epidemiological COVID19 curve in Kingdom of Saudi Arabia (KSA) reached a steady level two months from the beginning of the pandemic.
Moreover, the rates of critical cases and mortality in KSA are low due to the younger population in Saudi Arabia compared to European, North American, and Asian countries and the government's e cient precautionary measures 5 .
The in-depth clinical and laboratory characteristics of COVID-19 has been reported among COVID-19 Saudi Arabia patients; however, it is limited to case series and small sample size studies. Recently a total of 150 case series has been reported the clinical and therapeutic characteristics of hospitalized patients with con rmed COVID-19 in specialized hospital in Saudi Arabia. They found that 70% were mild cases 6 .
Although it is a limited number of cases, it highlights that reporting and assessing the characteristics of patients with con rmed COVID-19 are important to plan and implement policy interventions. There are limited reports demonstrating the variability in features of the disease between populations, taking into consideration the comorbidities, severity of the disease and immune system responses 7 .
The Saudi MOH has been driving the national COVID19 management protocols and living guidance, which had been developed in accordance with the latest scienti c, evidence-based COVID-19 studies. Speci c institution either adopted the same protocol or modi ed it in accordance with their internal expert committees 8 .
To the best of our knowledge, there is a lack of multicenter studies that examined the clinical course for patients with COVID-19 admitted to the ICU in Saudi Arabia. We therefore conducted this study to extensively examine the clinical characteristics, outcomes, and off label use of medications in critically ill patients with COVID19 in Saudi Arabia.

Study design
A multi-center, non-interventional, observational study of critically ill patients admitted to intensive care units (ICUs) with con rmed diagnosis of COVID19 in KSA. The diagnosis of COVID19 was con rmed objectively by Reverse Transcriptase -Polymerase Chain Reaction (RT-PCR) on nasopharyngeal and/or throat swabs. Retrospective part included de-identi ed data of COVID-19 PCR positive patients admitted prior to the date of IRB approval (March-April 2020). The prospective part was conducted between May 1 st and August 31 st ,2020. Patients were followed daily during ICU Length of stay (LOS) until in-hospital death or discharge whichever occurred rst.

Eligibility criteria
Patients were enrolled in the study if they were critically ill, aged 18 years or older and admitted to ICU with a positive PCR COVID-19. Patients with ICU LOS 1 day or 60 days, and/or labeled as "Do-Not-Resuscitate" status within the rst 24 hours of ICU admission were excluded.

Setting
This study was conducted in two large, tertiary governmental hospitals. The rst hospital was King Abdulaziz Medical city -Central Region (KAMC-CR) located in Riyadh and the second was King Abdulaziz University Hospital (KAUH) located in Jeddah. The distribution of total enrolled patients were 81 %, and 19 % in KAMC-CR and KAUH respectively. The primary site for this multicenter, prospective cohort study was King Abdulaziz Medical city (Riyadh).
King Abdulaziz Medical City is a tertiary-care academic referral hospital in Riyadh, Saudi Arabia. The ICUs admits medical, surgical, trauma, burn and transplant patients, and operates as a closed unit with 24/7 onsite coverage by critical care board-certi ed intensivists 36 . King Abdulaziz University Hospital is a tertiary care academic hospital located in Jeddah, Saudi Arabia. The hospital has a bed capacity of 1067 beds and ICUs admit medical, surgical and cardiac patients, and operates as a closed unit with 24/7 onsite coverage by critical care intensivists.

Data collection
We collected the following information, demographic data, Acute Physiology And Chronic Health Evaluation II (APACHE II), Sequential Organ Failure Assessment (SOFA) and Nutrition Risk in Critically ill (NUTRIC) scores, co-morbidities, pre-hospital (Home) medications, vital signs, laboratory tests and radiological nding within 24 hours of ICU admission, ICU support measures needed during ICU stay, off label use of medications for COVID19, COVID19 viral load. In addition, D-Dimer, Fibrinogen Level, Ddimer/ brinogen ratio, thrombosis during ICU stay, procalcitonin, iron study, radiological studies, and complication (s) during ICU stay were prospectively collected and followed.

Outcomes
The primary endpoint was extensively to describe the clinical and laboratory characteristics of critically ill patients with COVID 19 admitted to Intensive Care Units (ICUs) in Saudi Arabia. The secondary endpoints to determine the mean ICU LOS duration, mean MV duration, ICU mortality and risk factors for poor prognosis in Saudi Arabia.

Data management and Statistical analysis
We summarized categorical variables as number (percentage) and numerical variables (continuous variables) as mean and standard deviation (SD). The normality assumptions were assessed for all numerical variables using statistical test (i.e. Shapiro-Wilk test) and also using graphical representation (i.e. histograms and Q-Q plots). We compared categorical variables using the chi square or Fisher exact test, normally distributed numerical variables with the t test, and other quantitative variables with the Mann-Whitney U test. Baseline characteristics, baseline severity and outcome variables were compared with ICU mortality within 30 days and Thrombosis during ICU stay.
Multivariate logistic regression and generalized linear regression were used to nd out the relationship between ICU mortality within 30 days, thrombosis during ICU stay and the different outcomes considered in this study, adjusting for the patient's baseline severity scores (namely, APACHE II, NUTRIC and SOFA scores).
We assessed model t using the Hosmer-Lemeshow goodness-of-t test. Generalized linear regression was also used to nd out the relationship between study outcome and the different study parameters considered in this study, adjusting for baseline severity scores. The odds ratios (OR) and estimates with the 95% con dence intervals (CI) were reported for the associations. No imputation was made for missing data as the cohort of patients in our study was not derived from random selection. We considered a P value of < 0.05 statistically signi cant and used SAS version 9.4 for all statistical analyses. The multiplicity adjustment was not done.

Demographic and Clinical Characteristics
A total of 560 critically ill patients with COVID-19 who had been admitted in ICUs at the two governmental hospitals, clinical characteristics, laboratory tests and off label use medications were obtained. The average age of the patients was 60 years (SD 14.58); 224 (40%) of the patients were ≥ 65 years old. A total of 417 (74.5%) were male. Among the 560 patients, diabetes mellitus (57.7%) was the most common coexisting illness followed by hypertension (53.6%) and dyslipidemia (22.7%). On the other hand, 17.3% of the patients had no coexisting comorbid condition. The most common admission source was direct ICU admission (48.8%) followed by critical care response team (CCRT) activation from wards/ oor (44.3%) with a median of 3 days before ICU admission.

Overall Outcomes
Overall survival rate was 52.6% (295 patients). Whereas, the overall ICU mortality rate within 30 days was 42.3% (237 patients). The median ICU LOS, hospital LOS and Mechanical ventilation duration were 10 days (Interquartile range (IQR): 6.00 to 17.50), 17 days (IQR: 11 days to 25 days) and 9 days (IQR:3 days to 17 days) respectively. The rate of ICU readmission for survival within 3 months was 9.7% (Table 1). (Additional le 2) shows the clinical and laboratory ndings on admission. Higher baseline severity scores (APACHE II & SOFA score) and nutritional risk (NUTRIC score) were associated with higher 30 days ICU mortality (P < 0.0001). The most common blood group types were O+, A + and B + with proportion of 22.5%, 18.2% and 14.3% respectively (Additional le 1). Among common blood group types, A + was associated with higher 30 days-ICU mortality (57.8%) (P = 0.0361) as well as thrombosis during ICU stay (P = 0.0113) ( Table 5).
Among 560 patients, 370 (66.1%) require mechanical ventilation (MV) during the rst 24 hours. Requiring MV on ICU admission was associated with higher ICU mortality within 30 days (P = < 0.0001). Mixed Acid-Base Disorder was the most frequent acid-base disorder (21.61%) within 24 hours of ICU admission. Among patients who had ICU mortality within 30 days, metabolic acidosis was the most frequent (P = < 0.0001). Higher alveolar-arterial gradient with median of 430.7 mmHg (IQR: 292.7-582.8 mmHg), (P < .0001), and lower base excess with median of -1.9 (IQR: -3.95 to 0.45) were associated with higher rates of 30 days ICU mortality.
Within 24 hours of ICU admission, higher body temperature, white blood cells (WBCs), blood glucose level (BGL), serum creatinine, brinogen, procalcitonin, creatine phosphokinase (CPK), aspartate aminotransferase (AST) and Total iron-binding capacity (TIBC) were associated with higher 30-days ICU mortality. On the other hand, platelets count, serum iron, lymphocytes, Glasgow Coma Scale (GCS) and mean arterial pressure (MAP) were higher among the survivors within 30 days of ICU admission.
Patients who developed Acute kidney injury (AKI) within 24 hours of ICU admission, were associated with higher ICU mortality within 30 days (p-value < .0001). In addition, positive cumulative uids balance within 24 hours of ICU admission was associated with higher ICU mortality but not was statistically signi cant (Additional le 2).
On ICU admission, Bilateral Patchy Shadowing followed with Local Patchy Shadowing with a proportion of 60.8% and 7.32 respectively were the most common radiological nding on chest X-ray. Patients with radiological nding of Bilateral Interstitial Abnormalities were associated with higher 30 days-ICU mortality. No radiographic abnormality was found in 37 patients (6.61%), and was associated with lower 30-days ICU mortality (p-value 0.0001) (Additional le 2).

Discussion
In this multicenter prospective study, majority of the included patients were male (74.5%), and 40% of the patients were ≥ 65 years, which was consistent with previously published studies 9,10,11,12 . Diabetes mellitus was the most prevalent comorbid condition in our cohort affecting more than half of the patients admitted to the ICU (57.7%), followed by hypertension (53.6%) respectively, while in most reports hypertension was the most prevalent comorbid condition 13,14,15,16,17 (19) , and 67% (20) . This variation in mortality rate might be attributed different factors such as baseline patient characteristics, different duration of follow up among studies, and different ICU bed availability among different hospitals as a study reported that patient who were admitted to hospitals with lower ICU bed capacity had a higher risk of death 21 . In a previous multicenter cohort study that included 2215 critically ill adults with con rmed COVID-19 from US, the reported median ICU LOS was 9 days (IQR, 5-14 days) and the median hospital LOS was 16 days (IQR, 11-22 days) which was comparable to our data 21 .
The median APACHE II score on admission to ICU was 12. In comparison, this score was 10 in a study in Wuhan 17 . The median NUTRIC score and SOFA score were 3 and 5, respectively. Shock, AKI, and mixed acid-base disorders were experienced by around a quarter of patients. Two-thirds of patients received mechanical ventilation, and these population had the greatest bene t of corticosteroids 46 . The severity of ARDs in our patients seems less than other studies. The median PaO2/FiO2 ratio was 98.3, while it was 118.5 in the ICNARC report 44 , 132 in Atlanta study 47 , 160 in a study in Lombardy 48 , and 194 in the Singaporean study 43 . Several in ammatory biomarkers were elevated in our study such as ferritin, CRP, D-dimer, CPK, and brinogen, which are key biomarkers in ARDs and cytokine release syndrome associated with COVID-19 48 .
For the off label use of medications, as expected systemic corticosteroids were used in the majority of patients since a randomized controlled trial (Recovery trial) found a signi cant mortality bene t in COVID-19 patients who received a respiratory support in the dexamethasone arm compared to the standard of care arm 46 . Despite the current controversial e cacy of tocilizumab, its adverse reactions and expensive price, it was used in more than a third of patients 49,50 . Although treatment dose of heparin is not currently recommended, around a third of our patients received this dose 49 . For oseltamivir, it is recommended as empiric anti-in uenza treatment in hospitalized patients while awaiting in uenza testing results 49 . In our study it was used in around a third of included patients. Azithromycin was used in a quarter of patients. It is unclear whether it was used for community-acquired pneumonia or the prescribers believed in its e cacy in COVID-19, which is not currently supported 49,50 . Vitamin C and thiamine were used in around 20% of patients but there are insu cient data on these supplements 49 .
Hydroxychloroquine was rarely used in our study as it is no longer recommended in hospitalized COVID-19 patients 49,50 .
In our study of critically ill COVID-19 patients we found that the median time for viral load to be undetected was 13 days. Median time to viral clearance was ranging from 7 to 12 days and up to 28 days in some reports 22 . In a previous study conducted in the capital of Hunan, it took 17 days for patients to test negative 23 . Our report of median 13 days for viral clearance is in alignment with reported duration of average two weeks, we expect that critically ill patients to have prolonged virus release. Data from a large cohort of COVID-19 patient from a single University Hospital in Milano (Italy), reported a viral clearance rate within 14 and 28 days were 32% and 54%, respectively 22 .
We also observed a signi cant association between patients with persistence positive COVID-19 PCR and ICU mortality within 30 days, to date little is known about the association between timing of clearance and disease severity or mortality. One large cohort study of COVID-19 patient reported that neither viral clearance rate at 14 and 28 days nor time to negative viral RNA load were predictor of mortality rate 22 . Our observation is different than this report, also our result is important in daily clinical practice, and it could be potentially used to guide patient care. Similar association was reported in an earlier retrospective cohort study, they concluded that SARS-CoV-2 viral load among hospitalized patients with COVID-19 independently correlates with the risk of intubation and in-hospital mortality 24 . The hypothesis of persistence viral load and its relationship with disease severity and risk of mortality warrant further investigation.
The most common complication during ICU stay was respiratory failure followed by acute kidney injury and thrombosis. Since our COVID-19 patients are all critically ill patients, it is not surprisingly to report higher respiratory failure percentage. Several mechanisms have been proposed in literatures as the cause for substantial respiratory failure seen in COVID-19 patients. These include diffuse alveolar damage which account for ARDs, pulmonary edema, vascular occlusion and ventilation and perfusion mismatch are other possible mechanism for hypoxemic respiratory failure 25  COVID-19 may predispose both venous and arterial thromboembolic disease due to excessive in ammation, hypoxia, immobilization and DIC 37,38,39 . We have evidence of high thrombosis rate of 11.4% in critically ill COVID-19 patients despite the use of prophylactic anticoagulation, which is in line with 16% reported in an earlier prospective cohort study 40 . Another study conducted in three Dutch hospitals found a remarkable higher rate of composite thrombosis outcomes in ICU COVID-19 patients (31%) which is higher than our reported rate 41 . Nonetheless, most experts agree that the signal for increased thrombotic risk is enough to recommend pharmacologic venous thromboembolism (VTE) prophylaxis in all hospitalized COVID-19 patients as long as there is no contraindication. Our nding reinforces thrombosis risk in critically ill COVID-19 patients and strongly suggest using pharmacological VTE prophylaxis in all COVID-19 patients admitted to the ICU.
Interestingly, among our population who developed thrombosis during ICU, blood group A + and B-were the most signi cant group types to develop thrombosis. A large retrospective review showed no signi cant connection between blood type and worsening of the disease, between blood type and the need for hospitalization, positioning requirements for patients during intubation, or any in ammatory markers 42 . However, they did not assess the association between blood group type and thrombosis risk, this nding needs to be further evaluated. After controlling for the confounding factors, we are planning to conduct a further research study to assess the ICU mortality in relation to the pharmacotherapeutic regimen (s) received during the patient's hospital stay.
While dexamethasone has shown some promising results in severe COVID-19 patients 46 , efforts still ongoing to nd effective treatments for COVID-19 and several clinical trials are taking place to test the e cacy and safety of various drugs in critically ill patients. Finally, there is no strong evidence of signi cant clinical outcomes improvement in critically ill COVID-19 patients.
The uniqueness of our study is the extensive list of variables and outcomes we were able to capture throughout the study period. These variables and outcomes could be used for benchmarking between different countries and healthcare settings. Our study may have been affected by several limitations.
During the study period, there were several changes in the national treatment protocols for COVID-19 and due to the nature of our study it was very hard to control for these changes. The data was collected for critically ill patients with COVID-19 so the results of our study cannot be generalized to mild or moderate COVID-19 patients. The nonintervention nature of the study allows for treatment decision based on the treating physicians bias toward using one treatment regimen versus other cannot be ruled out. We encountered many confounding factors which could affect the external validity and the interpretation of the mortality outcome. However, we conducted several analyses to control for these variables.

Conclusion
Several variables were associated with increasing the risk of ICU death at 30 days, and the incidence of Availability of data and material The datasets used and/or analyzed during the current study are available from corresponding author on reasonable request.
Ethics approval and consent to participate The study was approved in April 29 th , 2020 by King Abdullah International Medical Research Center Institutional Review Board, Riyadh, Saudi Arabia (Reference No: RC20/192/R). Also been approved by King Abdualziz University, faculty of medicine, unit of biomedical ethics research committee, Jeddah, Saudi Arabia (Reference No: 231-20). Participants' con dentiality was strictly observed throughout the study by using anonymous unique serial number for each subject and restricting data only to the investigators. Informed consent was not required due to the research's method as per the policy of the governmental and local research center.

Consent for publication
Not applicable.