Ethics approval and consent to participate

The study protocol was approved by the ethics committee of the Notre-Dame des Secours University Hospital and Bellevue Medical Center. The approval done by Notre-Dame des Secours University Hospital ethics committee was also used in Bhannes Medical Center. Each patient was contacted to get his written informed consent prior to enrolling him/her in the study. Consent was obtained from immediate family members in case of death.

Patients and data collection

This study is a retrospective cohort study on 416 enrolled adults with PCR-confirmed COVID-19 who were consecutively admitted to the three institutions for a COVID-19 infection, from the opening of the COVID units until their closure (period extending from March 2020 to June 2021). The COVID units included a COVID floor and an ICU. Data on COVID-19 cases and deaths were extracted from the computerized system or charts according to the availability of the information. Patients with incomplete information, patients younger than 18 years, and patients with a negative COVID-19 RT-PCR or an unverified COVID-19 result were excluded. Patients were classified into (0) asymptomatic, (1) mild, (2) moderate, or (3) severe disease depending upon their symptoms at presentation. Asymptomatic patients or people without any reported clinical symptoms were hospitalized at the beginning of the pandemic in Lebanon to decrease COVD-19 fear and to decrease the disease contagion. Mild symptoms include patients with a documented COVID-19 infection without the suggestion of viral pneumonia or hypoxia. Patients with moderate disease of pneumonia include patients with no evidence of severe pneumonia, and SpO2≥90%; Severe and critically ill patients include patients with severe pneumonia,SpO2<90%, acute respiratory distress syndrome (ARDS), sepsis, and septic choc [25]. Data collection was done according to the sheet found in S1 Appendix.

Statistical analysis

The SPSS software v.22 was used for the statistical analysis. The Chi-square test was used to compare two categorical variables, whereas the Student t test was used to compare two means. A logistic regression was then conducted, taking mortality in the COVID-19 unit as the dependent variable. All variables that showed a p<0.25 in the bivariate analysis were entered as independent variables. Significance was set at p<0.05.

Sociodemographic characteristics

A total of 415 patients accepted to participate in the study, with a mean age of 61.58 ± 16.34 years and 32.0% females; 93 (22.4%) patients died. Other characteristics of the sample are displayed in Table 1.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Sociodemographic characteristics of the patients (N = 415).

https://doi.org/10.1371/journal.pone.0278393.t001

Bivariate analysis

A significantly higher percentage of males, of those who were admitted to the ICU, of those who had severe COVID-19 infection, who had hypertension, congestive heart failure, atrial fibrillation, dementia, peripheral vascular or liver disease, who were oxygen-dependent at home due to other diseases, who had thrombotic complications during hospitalizations, endotracheal intubation during hospitalizations, dialysis during hospitalization, who were vented, and were not treated with enoxaparin at their admission day as a prophylactic or therapeutic measure. Moreover, a higher mean age, leukocyte count, CRP upon admission, total length of hospital stay, vented length of stay, and PCO2 worst value were significantly associated with higher odds of death; furthermore, a lower mean platelet count, D-dimer count, PO2 upon admission, PO2 worst value, and pH upon admission were also associated with higher mortality (Table 2).

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Bivariate analysis of death as the dependent variable.

https://doi.org/10.1371/journal.pone.0278393.t002

Multivariable analysis

The results of a first model, taking variables related to baseline characteristics entered as independent variables, showed that having acute kidney injury (aOR = 4.057) and older age (aOR = 1.053) were significantly associated with higher odds of death (Table 3, Model 1).

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Multivariable analysis: Logistic regression taking the presence vs absence of death as the dependent variable (N = 316).

https://doi.org/10.1371/journal.pone.0278393.t003

In the second model, taking baselines characteristics and factors related to the hospital admission entered as independent variables, the results showed that enoxaparin intake (aOR = 0.435) was significantly associated with lower odds of death, whereas older age (aOR = 1.049) and ventilation (aOR = 1.2) were significantly associated with higher odds of death (Table 3, Model 2).

When entering all the variables that showed significance in the bivariate analysis in a third model, the results showed that older age (aOR = 1.243) and a higher PCO2 worst value during hospitalization (aOR = 1.192) were significantly associated with higher odds of death (Table 4). Atrial fibrillation, having COPD and higher leukocyte count on admission were significantly associated with higher odds of death, but this effect size was very weak.

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. Multivariable analysis: Logistic regression taking the presence vs absence of death as the dependent variable and all variables that showed significance in the bivariate analysis as independent ones (N = 88).

https://doi.org/10.1371/journal.pone.0278393.t004

Enoxaparin use and mortality

Although our paper did not differentiate between therapeutic and prophylactic enoxaparin dosages, analytical studies demonstrated a protective relation between its use and mortality in hospitalized COVID-19-infected patients [20,21,26]. In addition, all patients included in our study benefited from anticoagulation administration from their day of admission. Since early anticoagulation was found to be beneficial in reducing mortality in COVID-19 infections [26], this can be thought to have had a positive impact on patients’ chances of survival. In the bivariate analysis, the increase in D-dimer level on admission was clinically significant, however, in the multivariable model no major association was found. Perhaps this statistical difference could be explained by the early administration of enoxaparin and its beneficial effects on decreasing thrombus formation. The D-dimer level was lower in the mortality group; this might be explained by a difference in enoxaparin dosages, not included in our study. Depending on the initial presentation, patients had either a therapeutic or prophylactic enoxaparin dosage. Patients with a higher d-dimer count probably took a therapeutic dose that is associated with lower thromboembolism and mortality rates [27,28].

Age and mortality

Our findings reveal that being of older age is associated with a greater chance of mortality. This isn’t surprising knowing that it’s widely accepted that advanced age is a major risk factor for COVID mortality [29,30]. A meta-analysis that included more than half a million participants from many countries worldwide highlighted the significant effect of age on COVID mortality with people over 60 years old being the most vulnerable to the virus [30].

Acute kidney injury and mortality

Our results showed that patients who presented with Acute Kidney Injury displayed an increased likelihood of death. A retrospective cohort study revealed that people with Covid-19 frequently face complications like AKI. It has been linked to a bad prognosis and develops early along with respiratory failure [31]. Furthermore, it was also found that the primary outcome of patients with both COVID-19 and AKI was in-hospital death, and if the patient survived, dialysis to restore kidney function was imperative[32]. The positive association between AKI and older age is also supported by other studies that linked those previous factors to a higher risk of death [33,34]. All the patients in our study who came with AKI to the ER and died had severe disease at presentation (grade 3). This can be thought to be related to a delayed presentation. In fact, taking into account that Lebanon has been through a severe economic crisis [35], many citizens might have delayed their hospital admission and self-medicated due to the lack of financial resources and the inability of a deteriorating healthcare system to cover their needs. In addition, at that time, hospitals were overloaded which further delayed hospital admission for some. Therefore, it can be thought that higher mortality rates in patients with AKI, a proven complication, could be explained by their delayed presentation in the disease course.

Atrial fibrillation and mortality

Having atrial fibrillation was also correlated with higher odds of death, however, this outcome extent was very weak, which could be due to our small sample size. Nonetheless, this effect was also seen in other studies [36]. Atrial fibrillation is a condition that increases the risk of thromboembolic events which may increase the risk of death. It has been shown that the risk of developing a new-onset AF is increased and that an aggravation of pre-existing AF is probable during the acute phase of COVID-19 infection [37]. This aggravation can be due to several mechanism including direct myocardial damage, electrolytes imbalance and a cytokine storm -through activation of the system inflammatory system- which have profound cardiovascular consequences on patients mainly in the form of AF [38]. It’s worth noting that Atrial Fibrillation is not specific to COVID-19 but is rather a widespread reaction to the systemic inflammation caused by viral illnesses such as Influenza [2].

PaCO2 levels and mortality

Furthermore, and in synergism with other studies [39], a higher PaCO2 worst value during hospitalization showed to be associated with higher death rates. Even though other ABGs components (i.e. PaO2, pH) did not display any statistical difference in our study, Gupta and Al [39] showed a significant association between PaO2 and poorer outcomes. In addition, it has been shown that hypercapnia is common in invasively ventilated COVID-19 patients [40]. That same study also revealed that the main differences between hypercapnic and normocapnic patients were that people with hypercapnia had greater severity of ARDS, higher incidence of venous thromboembolic events, and increased ventilation rates. However, no association with mortality was proven [40].

Ventilation duration and mortality

While our study did not differentiate between the type of ventilation used, to our knowledge, few papers studied the association between length of ventilation (invasive or noninvasive) and death rates. Our research demonstrated that a higher ventilation duration was associated with a higher death rate. This may be related to the fact that an increase in ventilation rate or its duration was associated with a higher PCO2 and an increase in ARDS severity [40]. Likewise, another study found that patients with COVID who needed ventilation had a certain degree of lung damage that was linked to a greater mortality rate [41].

COPD and mortality

Having a history of COPD was associated with higher odds of death. Nevertheless, this effect was also seen in other studies[15]. COPD, especially those with CO2 retention, is associated with a higher PaCO2 rate. Since a higher PaCO2 was clinically significant in our study in increasing mortality rates, COPD and PaCO2 may be linked together.

Leukocyte count and mortality

A study done by Blomme S. and Al, [42], showed that the deceased COVID-19 group had significantly higher total leucocyte counts in the first 20 days after admission compared to the recovered COVID-19 group, telling that higher degrees of inflammation lead to a more severe outcome. In our study, we found a similar but weak effect because of the small sample size. Higher leucocyte counts on admission showed a statistical diference in patients’ prognosis. This can suggest an increase in inflammation even at presentation could be linked to a poorer outcome.

Limitations

Our limitations are mainly related to retrospective cohort limitations; the absence of some laboratory data for some patients (i.e. levels of procalcitonin, D-dimer, IL-6, ferritin, ABGs…), which may have affected their analysis in the multivariate analysis. Also, the lack of standardized care in the early COVID-19 pandemic may have affected our study since the three centers had different approaches to COVID-19 management. A specific study that allows considering management discrepancies would be necessary to confirm our findings. The number of patients who had a moderate/severe condition was 326; however, only 122 patients were admitted to the ICU, which might be a source of bias in terms of mortality. In fact, the lack of hospital and ICU beds for proper care may have affected our study, since many patients were ventilated at home before admission, and many self-medicated themselves prior to admission. Furthermore, the study’s timeline did not include vaccinated patients and newer strains (i.e. Delta, Omicron), which can have a different outcome. In addition, not differentiating between enoxaparin dosages is one of the limitations of our study. Further studies are recommended to confirm our findings.