key: cord-0700755-424zy24c
authors: Heili-Frades, S.; Minguez, P.; Mahillo-Fernandez, I.; Prieto-Rumeau, T.; Herrero Gonzalez, A.; de la Fuente, L.; Rodriguez Nieto, M. J.; Peces-Barba Romero, G.; Peces-Barba, M.; Carballosa de Miguel, M. d. P.; Fernandez Ormaechea, I.; Naya Prieto, A.; Ezzine de Blas, F.; Jimenez Hiscock, L.; Perez Calvo, C.; Santos, A.; Munoz Alameda, L. E.; Romero Bueno, F.; Hernandez-Mora, M. G.; Cabello Ubeda, A.; Alvarez Alvarez, B.; Petkova, E.; Carrasco, N.; Martin Rios, D.; Gonzalez Mangado, N.; Sanchez Pernaute, O.
title: COVID-19 Outcomes in 4712 consecutively confirmed SARS-CoV2 cases in the city of Madrid.
date: 2020-05-25
journal: nan
DOI: 10.1101/2020.05.22.20109850
sha: 16f7e69f1dc6823620bf0950bb6fee14dcb039d0
doc_id: 700755
cord_uid: 424zy24c

There is limited information describing features and outcomes of patients requiring hospitalization for COVID19 disease and still no treatments have clearly demonstrated efficacy. Demographics and clinical variables on admission, as well as laboratory markers and therapeutic interventions were extracted from electronic Clinical Records (eCR) in 4712 SARS-CoV2 infected patients attending 4 public Hospitals in Madrid. Patients were stratified according to age and stage of severity. Using multivariate logistic regression analysis, cut-off points that best discriminated mortality were obtained for each of the studied variables. Principal components analysis and a neural network (NN) algorithm were applied. A high mortality incidence associated to age >70, comorbidities (hypertension, neurological disorders and diabetes), altered vitals such as fever, heart rhythm disturbances or elevated systolic blood pressure, and alterations in several laboratory tests. Remarkably, analysis of therapeutic options either taken individually or in combination drew a universal relationship between the use of Cyclosporine A and better outcomes as also a benefit of tocilizumab and/or corticosteroids in critically ill patients. We present a large Spanish population-based study addressing factors influencing survival in current SARS CoV2 pandemic, with particular emphasis on the effectivity of treatments. In addition, we have generated an NN capable of identifying severity predictors of SARS CoV2. A rapid extraction and management of data protocol from eCR and artificial intelligence in-house implementations allowed us to perform almost real time monitoring of the outbreak evolution.

Demographics and clinical variables on admission, as well as laboratory markers and therapeutic interventions were extracted from electronic Clinical Records (eCR) in 4712 SARS-CoV2 infected patients attending 4 public Hospitals in Madrid. Patients were stratified according to age and stage of severity. Using multivariate logistic regression analysis, cut-off points that best discriminated mortality were obtained for each of the studied variables. Principal components analysis and a neural network (NN) algorithm were applied.

A high mortality incidence associated to age >70, comorbidities (hypertension, neurological disorders and diabetes), altered vitals such as fever, heart rhythm disturbances or elevated systolic blood pressure, and alterations in several laboratory tests. Remarkably, analysis of therapeutic options either taken individually or in combination drew a universal relationship between the use of Cyclosporine A and better outcomes as also a benefit of tocilizumab and/or corticosteroids in critically ill patients.

We present a large Spanish population-based study addressing factors influencing survival in current SARS CoV2 pandemic, with particular emphasis on the effectivity of treatments. In addition, we have generated an NN capable of identifying severity predictors of SARS CoV2. A rapid extraction and management of data protocol from eCR and artificial intelligence in-house implementations allowed us to perform almost real time monitoring of the outbreak evolution.

The new coronavirus pandemic was confirmed to have spread to Spain on 31 st January 2020, when a German tourist tested positive for SARS-CoV-2 in La Gomera, Canary Islands. Post-hoc genetic analysis showed that at least 15 strains of the virus had been imported and community transmission begun by mid-February 1 . By 13th March, cases had been confirmed in all of the state 50 provinces.

A state of alarm and national lockdown was imposed on 14th March. On 29th March it was announced that, beginning the following day, all non-essential workers were to stay at home for the following 14 days. By late March, Madrid had recorded the highest number of cases and deaths in the country. On 25th March 2020, the death toll in Spain surpassed the one reported in Mainland China, while only deaths in Italy remained higher. As of 9th May 2020, there have been 223,578 confirmed cases with 133,952 recoveries and 26,478 deaths in Spain. Regarding the course and outcomes of SARS-CoV-2 pneumonia, there are already population-based data from China 2 , Singapore 3 , Italy 4 and recently the United States of America 5 . This work aims to present a comprehensive analysis of COVID-19 risk factors including vital signs, comorbidities, and biomarkers, as well as treatment effectiveness using machine-learning and classifying mortality predictors of SARS CoV-2 based on a large population cohort from 4 public Hospitals in Madrid.

A total of 4712 patients attending any of the 4 centers participating in the study, from January 

Demographics, baseline comorbidities, vital signs, biomarkers, use of treatments and outcomes were extracted from clinical charts. Tobacco consumption was analysed and multivariate logistic regression models were performed testing potential interactions between tobacco and comorbidities.

All data were adjusted for age, taken as a continuous variable. In addition, patients were stratified into 5 age subgroups (0-16, 17-30, 31-50, 51-70 and 71-100) and in the range 31-100 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 25, 2020. . https://doi.org/10.1101/2020.05.22.20109850 doi: medRxiv preprint separately. Biomarkers with > 50% of non-missing data at first day of hospitalization were included in the analysis. The effect of treatments was studied in those patients receiving at least one drug from the clinical protocol (N=2776). Untreated patients were mild cases discharged from the emergency department. Patients were classified according to severity of the condition at the time of data collection (17/04/2020), specifically recording fatal outcomes and ICU admission. Thus, analysis and predictions were made using the following classification frameworks: 1) deceased patients and those admitted to ICU versus rest, 2) deceased versus rest, and 3) deceased versus ICU admitted.

Parameters with a high percentage (>50%) of missing values were excluded from the analysis.

Missing data for biomarkers and vital signs were imputed using average value within age ranges considered in each analysis.

The neural network (NN) model was built using the caret R package (https://CRAN.Rproject.org/package=caret). We used "nnet" method, and implemented an automatic selection of: i) the optimal number of units per hidden layer (1 to 5), and ii) the optimal value for the regularization parameter to avoid over-fitting (0.1 to 0.5 in increments of 0.1 units). We used "twoClassSummary" method to compute sensitivity, specificity and the area under the ROC curve. This NN algorithm was performed 10 times over a matrix of re-sampled patients. Resampled matrices were composed of patients with severe status (outcome = 1) and a random selection of the same number of patients with less severe status (outcome = 0). For each resampled matrix, we performed a 10-fold cross validation where randomly used 90% of patients as the training dataset and the remaining (10%) as the testing dataset. A total of 100 NN models were performed and a ROC curve is calculated for the set of predictions and real values using the 10-fold cross validation. Our final AUC, accuracy, sensitivity and specificity were calculated as the mean of the 100 NN models performed in total. Our R script is available free at: https://github.com/pminguez/MachineLearning4UnbalancedData/blob/master/nn4covid19.R.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 25, 2020. 

Principal component analysis (PCA) was performed on the matrices with missing data imputation, centered and scaled, using the prcomp and factoextra R libraries. Elipses represent 95% of the data. PCA analyses were performed over the same scenarios chosen to perform the NN (see above section).

For binary type of variables, which included comorbidities, treatments and gender, odds ratios for each of the two severity classes in each comparison were calculated, contingency 2x2 tables were obtained and significance was calculated with 2-tailed Fisher test. P values were adjusted using FDR and adjusted p values <0.05 were taken as significant. Graphics were built sorting variables by relative odds ratios (1/odds ratio when <1).

For non-binary variables, comprising vital signs and biomarkers, a non-parametric Wilcoxon rank sum test was performed over the set of values of each variable for the two severity states.

P-values were adjusted by FDR and FDR<0.05 were considered as significant. Means of the two sets of values for each variable was calculated to assign a direction of the test.

Logistic Regression models were used to obtain the vital signs cut-off values and to discriminate those, which were related to an unfavorable evolution. Using ROC curves and Youden's J statistic we obtained the cut-off values. We defined a binary classification for each parameter based on the cut-off points and selected parameters (p-value<0.2) to feed a multivariate logistic regression model. ROC curves for the model were calculated on 10-fold cross validation, and AUC, accuracy, sensitivity and specificity were averaged across the 10 ROC curves. The goodness of fit of the model was evaluated by the Hosmer and Lemeshow test.

Two models were used for this approach.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 25, 2020. 

Supplementary Figure 1 represents smoking habits and associated mortality. In spite of the fact that tobacco . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 25, 2020. . https://doi.org/10.1101/2020.05.22.20109850 doi: medRxiv preprint seemed to have a protective effect, as could be inferred from the results of the NN and PCA (Figure 1) , significance was lost after performing a multivariate model with all relevant variables ( Table 1) . As shown in the table, all comorbidities were found to increase mortality risk.

We performed PCA and NN analyses over the 4712 patients, classified into two statesdeceased and alive-at the time of data collection. Figure 1A shows principal component (PC or dimension) 1 against PC2. Arterial hypertension and cardiovascular affections were the conditions which most contributed to separation between states. Of note, the NN was able to predict the status with an AUC=0.74 (accuracy=0.7, sensitivity=0.4, specificity=0.81), as can be observed in Figure 1B . The comorbidities which principally contributed to the NN model prediction power are shown in Figure 1C . As can be observed in the figure, also using this approach arterial hypertension as well as neurological and cardiovascular diseases were strongly associated to a poor outcome. Odds ratios of the different comorbidities further revealed that seven disease families were significantly enriched in patients with a fatal outcome (FDR<0.05) ( Figure 1D ). On the other hand, although smoking appeared to be more frequent between survivors, the difference was not significant.

In relationship with abnormal vital signs in the emergency department, blood pressure and temperature were found to be the variables which better discriminated between deceased patients and survivors, according to PCA analysis (Figure 2A) . The NN was able to classify patients with an AUC=0.87 (accuracy=0.8, sensitivity=0.7, specificity=0.69, Figure 2B ), further reaching a value of 0.91 when classification was made between deceased and ICU-admitted patients (data not shown). In addition, temperature, age and maximal heart rate were associated with death when the two latter conditions were compared ( Figure 2C ). On the other hand, female sex, the absence of fever, adequate saturations at admission, as well as lower diastolic arterial pressures associated with survival ( Figure 2C ). In order to identify predictive cut-off values of these variables, univariate and multivariate linear regression models were carried out (Supplementary Table 4 ). Using these thresholds, we performed a multivariate . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 25, 2020. . https://doi.org/10.1101/2020.05.22.20109850 doi: medRxiv preprint logistic regression analysis, in which age >70 years, a minimal oxygen saturation <86%, heart rate >100 bpm or fever >37.8º indicated lower survival ( Table 2) Therapeutic modalities and patient survival.

The results are displayed in Table 3 :

For the usual confidence levels, the significantly negative coefficients corresponding to a higher survival probability were shown by treatment with erithromycin and Cyclosporine A, labeled (**) in the table, while tocilizumab, labeled (*), exhibited an intermediate significance level around 15%. It should be mentioned, however, that erithromycin was only administered to 82 patientsmostly in intensive care as prokinetic drug-, which is a relatively small number considering the sample size.

In order to provide a complementary decision tool for treatment prediction of outcomes, we In the whole population, Cyclosporine, hydroxycloroquine, bemiparin, dexamethasone and anakinra associated to survival, albeit only the first three drugs showed a significant (FDR<0.05) enrichment in survivors ( Figure 3A) . In the range of 31-50 years old (N=393 patients) PCA

showed not consistent ROC curves and no drugs associated to any of the outcomes with significant support. As regards ages between 51 and 70 years (1029 patients) (Figure 3C) , . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 25, 2020. . https://doi.org/10.1101/2020.05. 22.20109850 doi: medRxiv preprint Cyclosporinee was the only drug associated with survival (FDR<0.05). We further explored the effect of drugs on survival in the most severe group of patients, as defined either by admission to the ICU or by a fatal outcome. Interestingly, PCA showed a clear separation between these categories and revealed a positive contribution of all drugs to survival ( Figure 4A ). NN showed an AUC=0.9 (accuracy=0.84, sensitivity=0.67, specificity=0.73, Figure 4B ) with enoxoparin, tocilizumab and dexamethasone as the compounds with higher capacity to classify patients ( Figure 6C) . The enrichment analysis showed the benefit of those three drugs -with high odds ratios and FDR<0.05-along with an effect of lopinavir, betaferon, Cyclosporine and methylprednisolone ( Figure 4D ).

Blood tests including levels of relevant analytes at first day of admission were recorded in all patients with < 50% missing data, reaching a total of 2,682 patients included in the analysis.

According to PCA, three groups of biomarkers could syntethise class differences, a. monocyte and lymphocyte counts, b. urea and creatinine and c. neutrophils, segmented cell percentage and LDH ( Figure 5A) . The specific NN model yielded an AUC=0.73 (accuracy=0.69, sensitivity=0.62, specificity=0.66, Figure 5B ). The variables with higher weights in class predictions were lymphocytes, hemoglobine and urea ( Figure 5C ). Wilcoxon rank sum test was calculated comparing the values of each variable in deceased and living patients ( Figure 5D ).

While higher counts of lymphocytes, monocytes, platelets and red cells, as also levels of alanine aminitransferase and hemoglobin were found to associated with survival in this analysis, additional parameters were increased in the deceased patients. Out of them, absolute numbers of lymphocytes showed the higher discrimination capacity between outcomes.

The present study shows performance of a NN capable of identifying and classifying mortality predictors in SARS CoV-2 infected patients. The NN code is offered to the scientific community, since the algorithm has the ability to adapt and self-learn, accounting for a valuable predictive tool in a pandemic which is still evolving. A local clinical practice-based guidance for inpatient . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 25, 2020. Regarding comorbidities, in agreement with previous data, a past history of arterial hypertension, neurological disorders and diabetes weighed negatively in the outcome of our patients. These comorbidities have been reported in other series 6, 7 which suggest that the virus preferentially targets patients with the so-called metabolic syndrome. In relation to neurological disorders, many authors have argued for a possible tropism of the virus to neural tissue, as indicated by the high prevalence of anosmia and ageusia 8 , but also by frequent cases presenting with encephalitis 9 and strokes 10 . Intriguingly, smoking has not been found to increase risk or severity. Moreover, there are some data pointing to its possible protective role, an effect that has being claimed to rely on nicotine 11 . However, our results could not consistently support this potential effect. Targeted studies are warranted to definitively clarify this aspect, which can significantly affect public health.

Regarding clinical signs, fever>37,8º, age>70 and heart rhythm disturbances, were predictors of poor prognosis in our cohort. With respect to heart rhythm, there is certain evidence pointing to an arrhythmogenic potential of the virus 12 . Older age has repeatedly been regarded as a decisive factor in the development of severe forms of COVID19, a fact that should be taken into consideration in the distribution of resources 13 . Along with the weight of age, our study also replicated the protective effect of female gender found in other studies. In addition, we could define a range of "safe" oxygen saturations between values of 86% to 97,5%, suggesting that not only hypoxia but also hyperoxia 14 could be detrimental. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 25, 2020. . https://doi.org/10.1101/2020.05.22.20109850 doi: medRxiv preprint respiratory failure appears to be the consequence of a vigorous innate immune response and does not exactly correlate with viral load 16 . Indeed, severe cases of SARS-CoV2 pneumonia can be distinguished for the increase in circulating levels of IL10, IL6 and TNF , which in  association to the rise in ferritin serum concentration point to macrophage activation syndrome as the driver of acute respiratory distress and sepsis. 17 . This insight has led to the incorporation of anti-cytokine drugs, such as tocilizumab, and Janus kinases inhibitors to clinical studies and treatment guidelines, while conversely the use of glucocorticoids remains controversial 18 .

Cyclosporine A is an immunomodulatory drug included in the WHO Essential Medicines List.

The rationale for selecting this compound has been recently discussed 19 . Briefly, the presence of betacoronavirus inside immunocompetent cells is thought to induce the unfolded protein response jeopardizing peptide quality control activities and mitochondrial function 20 .

Cyclosporine is able to modulate the unfolded protein response and rescue cells from irreversible damage under hypoxia and stress conditions 21 . In addition, synthetic activities of this family of virus have been shown to rely on their binding of cyclophilins, which are selectively blocked by Cyclosporine 22 , a mechanism accounting for the broad antiviral activities displayed by the drug 23 .

Remarkably, when we analysed the drugs used by our group, survival always appeared to be related to the exposure to Cyclosporinee both at the individual level and in combination. The same tendency was found using logistic regression or NN in the different age strata, both in the global cohort and more firmly in the inpatient subgroup. For all these reasons and in view of these results and some other unpublished observations we set up a randomized clinical trial assessing efficacy of Cyclosporine as add-on therapy with standard of care, which is now ongoing 24 .

As shown in Supplementary Table 1 Corticosteroids alone did not show benefit in our series, a finding that at first sight could argue against their use in COVID-19 pneumonia in accordance with current WHO guidance 25 .

However, in agreement with previous observations 26 , we found that their combination with other drugs was consistently associated to a higher survival of our critically ill cases, as can be concluded from the results of the NN in Figure 4 where they stand as second therapy after tocilizumab in increasing chance of survival. Tocilizumab is currently undergoing a clinical trial, but preliminary results published as a press release 27 and also data from a pilot study 28 suggest that it could have a beneficial effect in this viral pneumonia.

Notably, we were able to observe that most of the specific therapies in our protocol could improve survival. In particular, given that it is the ICU admitted subgroup of patients the one exposed to a larger combination of drugs, in relationship with severity and use of maximum therapeutic effort, we explored the impact of medications in these patients. Figure 4 specifically shows patients who were admitted to ICU and survived, albeit 40 of these patients have not

been discharged yet. In order of importance, the 5 drugs that were able to lower mortality in critically ill cases were enoxoparin, dexamethasone, tocilizumab, Cyclosporine and bemiparin.

In relationship with analytical markers we analysed matrix yields data, which are in agreement with previous publications. In summary, the following alterations were more likely associated to death or ICU admission in order of importance: uremia, neutrophilia, hiponatremia, raised Creactive protein, leukocytosis, elevated levels of LDH, hypoglycemia, high levels of GOT and GGT, lymphopenia, renal failure, increased ferritin, bilirubin and lengthening of clotting times.

Conversely, a normal or high percentage of lymphocytes, absence of anemia or of thrombocytopenia and and normal liver tests were protective traits.

This study has several limitations. Firstly, data was collected from the electronic health record database. This precluded the level of detail yielded by a manual medical record review.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 25, 2020. Thirdly, due to the rapid increase in ICU beds in at least one of the 4 hospitals, it was not possible to adequately assign some of the beds as ICU.

We wish to thank all the subjects included in this study for their courage in fighting this virus, each and every one of the doctors at these four hospitals for their titanic effort and also all our statisticians, mathematicians, and the entire leadership team of the participating hospitals. The work is hosted by ISCIII project COV20/00181.

Sarah Heili-Frades MD 1# contributed to the conception and design of the work, did the analysis, interpretation of data and wrote the manuscrip , Pablo Minguez PHD 2# designed the NN algorithm, performed the analysis and wrote the manuscript, Ignacio Mahillo Fernández PHD 3 , Tomás Prieto-Rumeau PHD 4 , Antonio Herrero González 5 , Lorena de la Fuente PHD 6 

The survey is a retrospective study with de-identified medical record data. No patient management protocols have been altered due to the study. The study was approved by our Institutional Ethics Committee.

The datasets analyzed during the current study are available from the corresponding author on request.

The authors declare that they have no competing interests

The study has no specific funding. AS has a Marie Sklodowska-Curie grant (#796721). LF is supported by ISCIII (CA18/00017). PM has a Miguel Servet contract funded by the ISCIII (CP16/00116).

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 25, 2020. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 25, 2020. . Table 3 . Univariate logistic regression of the relationship between each of the treatments and survival. We consider 26-predictor variables for each of the 26 treatments. We show the estimation of the coefficient for each of the treatments, and, in order to test statistical significance, we also show the p-value for testing against . A negative coefficient indicates that the corresponding treatment, according to the adjusted model, diminishes thus being associated with survival. For the usual confidence levels, the significantly negative coefficients corresponding to a better survival were the treatments Eritromycine and Cyclosporinee, labeled (**) in the . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 25, 2020. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 25, 2020. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 25, 2020. Significance is calculated using a Wilcoxon test with FDR correction over p-values, FDR<0.05 was taken as significant. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 25, 2020. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 25, 2020. Significance is calculated using a Wilcoxon test with FDR correction over p-values, FDR<0.05 was taken as significant.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 25, 2020. . https://doi.org/10.1101/2020.05.22.20109850 doi: medRxiv preprint Figure 3 . Treatments in dead and non-dead patients of three different age groups. A) PCA, Neural Network ROC and importance, and Fisher test results for patients of all ages (2776 patients). B) PCA, Neural Network ROC and importance, and Fisher test results for patients in 31-50 age range (393 patients). C) PCA, Neural Network ROC and importance, and Fisher test results for patients in 51-70 age range (1029 patients).

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 25, 2020. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 25, 2020. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 25, 2020. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 25, 2020. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 25, 2020. . https://doi.org/10.1101/2020.05.22.20109850 doi: medRxiv preprint

Phylodynamics of SARS-CoV-2 transmission in Spain

Clinical course and risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study

Epidemiologic Features and Clinical Course of Patients Infected With SARS-CoV-2 in Singapore

The response of Milan's Emergency Medical System to the COVID-19 outbreak in Italy

Presenting Characteristics, Comorbidities, and Outcomes Among 5700 Patients Hospitalized With COVID-19 in the New York City Area

Diabetes mellitus is associated with increased mortality and severity of disease in COVID-19 pneumonia -A systematic review, meta-analysis, and meta-regression

Association of Renin-Angiotensin System Inhibitors With Severity or Risk of Death in Patients With Hypertension Hospitalized for Coronavirus Disease 2019 (COVID-19) Infection in Wuhan, China

Anosmia and Ageusia: Common Findings in COVID-19 Patients

Encephalitis as a clinical manifestation of COVID-19

Neurological complications of coronavirus and COVID-19

Smoking Upregulates Angiotensin-Converting Enzyme-2 Receptor: A Potential Adhesion Site for Novel Coronavirus SARS-CoV-2 (Covid-19)

SARS-CoV-2, COVID-19 and inherited arrhythmia syndromes. Heart Rhythm

Is it wrong to prioritise younger patients with covid-19?

Nuevos datos de toxicidad, nuevas recomendaciones y soluciones innovadoras: sistemas automatizados de titulación y destete de oxigenoterapia Rev Patol Respir

Pharmacologic Treatments for Coronavirus Disease 2019 (COVID-19): A Review

Clinical progression and viral load in a community outbreak of coronavirus-associated SARS pneumonia: a prospective study

Complex Immune Dysregulation in COVID-19 Patients with Severe Respiratory Failure

Clinical evidence does not support corticosteroid treatment for 2019-nCoV lung injury

Why choose Cyclosporine A as first-line therapy in COVID-19 pneumonia

Modulation of the unfolded protein response by the severe acute respiratory syndrome coronavirus spike protein

Suppression of NF-kappaB by Cyclosporine a and tacrolimus (FK506) via induction of the C/EBP family: implication for unfolded protein response

The SARS-coronavirus-host interactome: identification of cyclophilins as target for pancoronavirus inhibitors

MERS-coronavirus replication induces severe in vitro cytopathology and is strongly inhibited by Cyclosporine A or interferon-α treatment

The SARS-CoV2 study group of HU Fundación Jimenez Díaz is currently conducting a clinical trial to test efficacy and safety of CsA as add-on therapy with standard of care in a 1:1 randomization with severity stratification

WHO Clinical management of severe acute respiratory infection when novel coronavirus (nCoV) infection is suspected

Potential benefits of precise corticosteroids therapy for severe 2019-nCoV pneumonia

The cytokine release syndrome (CRS) of severe COVID-19 and Interleukin-6 receptor (IL-6R) antagonist Tocilizumab may be the key to reduce the mortality