Hemogram Rates as Prognostic Markers of ICU Admission in COVID-19

Background Since rst cases of SARS COV-2 were identied, the number of affected and dead people make necessary to identify factors related to worse evolution. Endothelial injury has been proposed as the main pathophysiological mechanism in the illness development, which provokes a hyper inammation and prothrombotic state. Leukocytes and platelets play a role in inammation and thrombogenesis so we propose to study if neutrophil-to-lymphocyte ratio (NLR), platelets-to-lymphocyte ratio (PLR), the novel neutrophil-to-platelet ratio (NPR), and the systemic immune-inammation index (SII), could be useful to identify patients who will need admission at Intensive Care Units. Methods A retrospective observational study was performed at HM Hospitales including 2245 patients with COVID-19 from March 1 to June 10, 2020. Patients were divided into two groups, admitted, and not admitted to ICU. health report system at hospital admission. All were assessed in the emergency department where a blood sample was drawn. Laboratory assessments consisted of complete blood count (including white blood cell count, leukocyte subtypes, hemoglobin count and platelet count), biochemical parameters and blood coagulation tests (D-dimer, prothrombin time and activated partial prothrombin time). Counts of neutrophils (x 9 cells/L), lymphocytes (x 10 9 cells/L) and platelets (x 10 11 cells/L) were used to dene four different inammatory rates: NLR neutrophils and lymphocytes, platelets lymphocytes, NPR the neutrophils and platelets and, nally, SII is dened neutrophils multiplied by platelets divided lymphocytes. suggests that a part of patients presenting with severe COVID-19 may present an underlying hyperinammatory response driving a cytokine release storm resulting in multiorgan failure and death (33). This novel form of microvascular obstructive thromboinammatory syndrome has been proposed as the pathophysiology underlying this hyperinammatory response. These hematological ndings are in consonance with other published data, however, our study emphasize the utility of combined white cell count ratios in evolution of hospitalized patient. This fact reects the complexity and heterogeneity of SARS-CoV-2 infection response: not only the total number of white cell, lymphocyte neutrophil or platelet recruited, but the balance between blood cell subtypes provide an adequate or ineffective immune response to SARS-CoV-2 infection. Our retrospective study shows the utility of four combined hemogram values as useful hospital predictors of unfavorable clinical evolution in a large number of COVID-19 infected patients. Specically, NRL, PRL, SII and PLR could be useful combined indicators of the inammatory and immunological host status.


Background
In December 2019, the rst cases of pneumonia of unknown cause emerged in the city of Wuhan, China. On January 7th a new coronavirus was identi ed at the Chinese Center for Disease Control and Prevention as the causative agent of pneumonia. A betacoronavirus (1) was named by the World Health Organization (WHO) as 2019-n-CoV (2). The syndrome developed is known as severe acute respiratory syndrome coronavirus 2 (SARS.CoV-2) (3-5) and on March 11 , 2020, with 108.000 cases reported in 114 countries around the world and a total of 4.291 people died, the WHO declared that the outbreak of this disease was considered as a pandemic (6).
Approximately 80% of infected people have mild to moderate symptoms while the remaining 20% have a severe clinical presentation. The Intensive Care Unit (ICU) admission differs from 5-32% (7-9) due to an acute hypoxemic respiratory failure.
The cause of respiratory failure is the development of Acute Respiratory Distress Syndrome (ARDS). It is a diffuse pulmonary in ammatory injury that leads to an increase in pulmonary vascular permeability and makes gas exchange di cult (10). Pathophysiological theories suggest a crucial role of endothelial dysfunction related to evolution of this infectious disease. Endothelial cells are involved in keeping homeostasis, vasomotor tone regulation, vascular permeability, vascular barrier function, and blood rheology (11,12). Endothelial injury may activate, through a cytokine release, a hyperin ammatory and procoagulant state. White blood cells, neutrophils, lymphocytes and monocytes, are directly involved in this systemic in ammatory response while platelets are the main mediators of hemostasis.
It has been described as common laboratory abnormalities in patients with COVID-19 infection, such as increased white blood cells count and neutrophil count and decreased lymphocyte count (13). Hemogram rates such as neutrophil-to-lymphocyte ratio (NLR) and platelet-tolymphocyte ratio (PLR) have been proposed to assess the extension of systemic in ammation.
NLR has been proposed as a prognostic marker of severity in oncological processes (14)(15)(16), appears to be an indicator of endothelial dysfunction and an important predictor of cardiovascular mortality (17,18) as well as a mortality predictor in patients with the acute respiratory syndrome (19,20). NLR is associated with worse outcome in patients with SARS-Cov-2 infection (21)(22)(23).
PLR has been described as a maker in acute in ammatory and prothrombotic states as well as PLR may re ect the degree of cytokine release, for this reason this ratio might provide useful information about clinical evolution of COVID-19 patients (24).
Systemic immune-in ammation-index (SII), is also a hemogram ratio studied as a prognostic indicator in the follow-up of sepsis (25) and in cancer patients (26,27).
In this study we propose to include the new neutrophil-to-platelet ratio (NPR). It incorporates the number of neutrophils, implicated in the in ammation response to infection and also involved in the thrombotic mechanism, and a number of platelets whose role in hemostasis is essential as well as his immunomodulatory properties (28).
According to our results that have been published previously (21)(22)(23), the main objective of this research was to clear out the role of hemogram rates as prognostic markers of severe COVID-19 de ned as the need for ICU admission and as secondary outcomes to nd other variables and parameters related to the risk of ICU admission in patients affected by COVID-19.

Materials And Methods
COVID-19 patients that were hospitalized at any of the 10 hospitals of the HM Hospitales group between March 1 and June 10, 2020, were retrospectively included in the study. Last follow up of the clinical or laboratory data of the patients included was on June 24, 2020.
A total of 2543 COVID-19 patients were admitted during the study period. Patients whose age was under 18 years old (n = 5), having missing laboratory data in the rst 24 hours of admission (n = 258) or died at hospital admission (n = 26) were excluded. Therefore, from the 2543 patients, 2254 (88.6%) were included in the nal analyses as shown in the ow diagram in Figure S5.
During period previously described, due to the dramatic pandemic situation with a multitude of admitted patients and a shortage of PCR tests, there were changes in the diagnostic protocol by the Spanish Ministry of Health. For several weeks, the diagnosis of Covid-19 was based solely on clinical characteristics and radiological criteria. Infection by SARS-CoV-2 was con rmed by PCR in 2114 (93.8%) patients. The remaining 140 patients included presented clinical and/or radiological signs compatible with COVID-19, as per protocol.
The study protocol was approved by the HM Hospitales Ethics Committee on March 25, 2020 (approval number 20.03.1573-GHM).
Demographic data, comorbidities, epidemiological characteristics and laboratory ndings from each patient were collected from the electronic health report system at hospital admission. All patients were assessed in the emergency department where a blood sample was drawn. Laboratory assessments consisted of complete blood count (including white blood cell count, leukocyte subtypes, hemoglobin count and platelet count), biochemical parameters and blood coagulation tests (D-dimer, prothrombin time and activated partial prothrombin time).
Counts of neutrophils (x 10 9 cells/L), lymphocytes (x 10 9 cells/L) and platelets (x 10 11 cells/L) were used to de ne four different in ammatory rates: NLR is the ratio between neutrophils and lymphocytes, PLR is the ratio between platelets and lymphocytes, NPR is the ratio between neutrophils and platelets and, nally, SII is de ned as neutrophils multiplied by platelets and divided by lymphocytes.
Moreover, we propose to study the in ammatory rates and their rate of change, getting data from up to four consecutive blood laboratory tests since the patient was admitted to the hospital. The rate of change compares the evolution of NLR, PLR, NPR and SII with the value obtained at admission and was de ned as the slope of the linear t of the relative rates versus time from hospital entry in days. It is considered positive change if the value of the ratio is increased more than 10% per day, negative if the value decreased at least 10% per day and in the cases were the change was between − 10% and 10% per day were classi ed as null.
Summary statistics were made for the entire cohort and the patients who needed to be admitted to the ICU and those who were not were grouped together. Continuous variables were summarized as median (interquartile range) and categorical variables as absolute frequency (relative frequency, %).
Differences between groups were evaluated applying Mann-Whitney U test for quantitative variables and Χ 2 test or Fisher's exact test for categorical variables. Spearman's rho test was used to evaluate correlation between continuous variables. Correlation plots were constructed using the R package GGally.
Variables that demonstrated differences between both groups, admitted and not admitted to ICU, with p value < 0.2 were selected for univariable logistic regression.
To implement the bivariable logistic regression models the in ammatory rates were con ated to other variables. Those parameters that added to the model, got modify the in ammatory rates estimate by at least 10 per cent, were considered to build the multivariable adjusted models. Model A included age, heart rate, temperature > 38ºC, systolic and diastolic blood pressure, NLR rate of change > 10 % per day, AST, D-Dimer and glucose. Model B-D included previous model and oxygen saturation (> 94, 90-94 or < 90 %), LDH and CRP respectively (Table 5).
Interaction and strati ed analyses were performed for each in ammatory rate adjusted to model A and conducted for age (< 70 and > 70 years), sex, hypertension, diabetes mellitus, oxygen saturation (< 90 and > 90 %) and LDH and CRP both categorized through their respective median values (Figs. 1A, S1B, S1C, S1C).
Statistical inference was performed using two-tailed test and with type I error rate of 0.05. All statistical analyses were done using R (version 4.0.0).

Results
Clinical or laboratory data of 2254 patients admitted to Group HM Hospitales due to COVID-19 infection during the study period, from March 1 to June 10, 2020, are shown. Clinical characteristics are summarized in Table 1 and laboratory ndings in Tables 2 and 3. The median age of patients was 69 [57-80] and 59.5% were men. At the time of admission, all patients were assessed in the emergency room and a blood sample was taken for current analysis and subsequently all patients received standard approved treatment for COVID-19. Infection by SARS-CoV-2 was con rmed in one thousand nine hundred fty four (93.8%) patients.
One hundred eighty ve (8.29%) patients experienced severe acute respiratory failure and were admitted to the ICU and were compared with those who did not require ICU admission. Three hundred thirty one (14.7%) patients passed away, sixty one (33%) from the ICU group and two hundred seventy (13%) of non-ICU group (p < 0.0001).
At the time of hospital admission, clinical differences were observed between patients who were admitted to ICU and those who did not, including sex ( (Table 3).   Figure S2 and optimal cut-off values are shown in Table 4).   Figure S1C).
PLR showed no signi cant association with ICU entry in the strati ed analysis ( Figure S1D).

Discussion
As the number of patients with COVID-19 infection is rapidly increasing worldwide, early identi cation of patients at risk of progression to a critical state is essential both to consider an aggressive speci c intervention and to identify differential clinical risk groups.
Several studies have published many different clinical scores trying to predict patients at risk of ICU admission or die (22,(29)(30)(31). A recent systematic review by Wynants et al. showed that the vast majority of the published studies presented some relevant limitations, such as a nonrepresentative selection of control patients, lack of su cient complete follow-up or poor selection of the overall study sample (32).
A large number of parameters recorded have been widely used as prognostic factors in the inpatient setting. In our study, male gender and arterial hypertension were risk factors for ICU admission. These ndings are in consonance with previous reported studies. Especially arterial hypertension could re ect a clinical predisposition to vascular damage, typical of some severe cases of COVID-19. However, we did not nd an incremented risk in patients presenting other comorbidities in relationship with potential vascular damage as Diabetes Mellitus or Cardiovascular Disease, even though these comorbidities were prevalent in our study population.
Blood test values have been widely used in early evaluation and monitoring of COVID-19 patients. In our study many differences were found in patients whose were admitted at ICU compared to non-ICU admitted including higher levels of LDH, CRP, APTT and D-Dimer. They also presented higher white blood cell count and neutrophils, while the inverse relationship was seen in non-neutrophil white blood cell series (low lymphocyte, monocyte and eosinophil ranges) and platelet levels ( Table 2).
Many of these parameters, speci cally derived from white blood cell or NLR has been used as in ammatory marker of COVID-19 (23,24,(34)(35)(36)(37), and its prognostic value stands out among our ndings (21,22). NLR could show the importance of the balance between neutrophil and lymphocytes and it seems to be an independent, robust and reproducible parameter in routine clinical practice.
On the other hand, the new NPR emphasizes the importance of the relationship between immune response and homeostasis (21,22). An endothelial thrombo-in ammatory syndrome has been hypothesized following alveolar viral damage by SARS-CoV-2 (38) and other virus agents (39). This would be caused by infection of endothelial cells using the cell surface angiotensin converting enzyme 2 receptor expressed in these cells (40,41) and it has been demonstrated endothelial cell involvement across vascular beds of different organs in necropsies of patients with COVID-19 (42,43). We hypothesize that a damaged and activated endothelium would increase the permeability and release of cytokines that would increase the chemotaxis of in ammatory cells and also send signals to the blood cells to favor the repair. In this context, after this aggression, platelets and neutrophils are activated by soluble agonists and adhesive proteins via their surface receptors playing a determining role in microvascular occlusion during thromboin ammatory disease (44).
NPR ampli es the value of this relation and we propose this rate as a new biomarker that could be useful to predict prognosis and severity of COVID-19 (14,15,16). This ratio has demonstrated to be very useful to predict in hospital mortality (21,22). Due to the relevant role of neutrophils and platelets in SARS-CoV-2 clinical evolution, we postulate that an increase in the value of neutrophils in relation to a decrease in platelets count could be a relevant prognostic factor, especially in early detection of potential severe cases of COVID-19. As we discussed, the role of neutrophils in response to infection and in ammation is well known, and as for platelets, their role in haemostasis is fundamental when vascular damage occurs. However, the role of neutrophils in thrombosis is increasingly recognized and more is known about the immunomodulatory properties of platelets in such a way that they interact with each other during infection, in ammation and thrombosis by modulating the functions of each (14).
On the other hand, the dynamic evolution of the immune response to SARS-CoV-2 infection could be crucial in the evolution of COVID-19 patients. Previous publications have shown the utility of velocity of change of hemogram rates to predict mortality in COVID-19 patients (21,22).
According to our results, the velocity of change of four hemogram rates during rst days of hospital staying would be signaling a greater in ammatory state in some patients who will later require admission to the ICU as these parameters together with rates at hospital admission have shown their usefulness as prognostic markers of in ammation in patients who ultimately required admission to ICU. In this sense, the hemogram is a tool within the reach of all hospitals and doctors who do not have the technical and material means to carry out complex immunological studies, which often produce late results. The analysis of the hemogram's rates would provide much more information than could be extracted a priori by evaluating the parameters in isolation. We now know that it is crucial to initiate early anti-in ammatory treatment when the patient deteriorates and the hemogram could be an indicator of that signal that could indicate which patients could potentially bene t from earlier anti-in ammatory therapy.
Dynamic changes during the evolution of the disease distort the assessment of every therapeutic intervention, especially in retrospective studies. An increase in the frequency of use of a drug due to clinical aggravation may not be distinguished from an effect of that drug. It is well known that the mechanisms of action of corticosteroids translate into several effects on the response of the immune system, classically producing lymphopenia, neutrophilia and also decreasing cytokine production. In theory, this mechanism may explain an increase in the hemogram rates. However, various studies regarding the prognostic value of NLR in in ammatory diseases (for example, in Bechet's disease or Alcoholic hepatitis), have shown a reduction in the ratio in patients under corticosteroid treatment (45,46 Our study presented several limitations. Diagnostic criteria during the rst months of the pandemic in Spain changed, from detection by PCR to clinical and radiological criteria as a consequence of changes in the diagnostic protocol set for by the Spanish Ministry of Health amidst the overworked health system. In fact, 6.2% of patients in our study COVID-19 was suspected. Additionally, the results of the strati ed analysis (Figs. 1A, S1B, S1C and S1D) suggest that age is a signi cant factor that in uences on ICU admission, affecting the rest of the variables. This fact could explain the association with mortality of age over 70 years with sex, hypertension, SatO2 > 90, LDH > 677 U/L and CRP > 131 mg/L.
During the study period, due to the pandemic situation in Spain, there was limited access to ICU beds and ventilators, which could have conditioned some results since patients could have been candidates for ICU based on age, comorbidities and survival chances. However, NPR is the only hemogram rate that maintains predictive capacity in the lower risk strata (< 70 years, male sex, Sat02 > 90, LDH > 677 U/L and CRP > 131 mg/L), which implies that their degree of independence with respect to these variables is greater (Fig. 1A). After reviewing our data, the same reason could explain the mortality rate of patients not admitted to the ICU. Although this number is small, it represents a percentage that may be relevant for the nal analysis. This data may derive from patients with a high number of comorbidities or extreme age, a priori without ICU admission criteria. The presence in our study of other patients without ICU criteria but with a favorable evolution is not well established, however, this fact reproduces the usual clinical practice at the present time. Similarly, the use of drugs such as corticosteroids, antivirals and other drugs were variable, and did not always respond to the same criteria during the study period. Finally, this is a retrospective study and lacks a control group, which limits the systematic adoption in routine clinical practice. Further comprehensive studies are needed to determine how useful are these blood tests and the rates at which future prognostic scores demonstrate usefulness in guiding treatment decisions

Conclusions
Our group is reporting the potential use of different blood test rates and their dynamic assessment as prognostic implications for ICU admission risk in patients affected by COVID-19.
Therefore, as the hemogram is a tool within the reach of all hospitals and doctors, we propose the use of the four hemogram rates for the assessment of the patient affected by COVID-19, especially the novelty NPR that could be very useful as a marker in the prognosis of this  Interactions and strati ed analyses for NPR (neutrophil-platelets ratio) adjusted to model A (