Northern Italy, and particularly Lombardy, is one of the regions of the world mostly affected by COVID-19, caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection. To investigate the still largely unknown pathophysiology of this disease, we have built a consortium of Italian Hospitals to include a large cohort of COVID-19 patients from mild out-patients managed by GPs to inpatients developing mild, moderate or severe disease assessed both in hospital and at a 3-6 month follow-up visit). Consortium partners have a wide expertise to allow for 1) comprehensive assessment of risk factors for severe COVID-19 syndrome; 2) study the pathophysiology of its cardio-respiratory manifestations; 3) estimate risk scores also with artificial intelligence and 4) assess its clinical immunoinflammatory and cardiorespiratory sequelae in discharged patients at short term follow-up. To this aim, we will

1. Enroll around 5500 COVID-19 patients (1000 outpatients and 4500 in-patients), which will allow to:

   1.1 Phenotype patients with COVID-19 of variable severity 1.2 Assess the prevalence of COVID-19 among GPs in relation with their use of PPE 1.3 Evaluate the impact of patients' demographic and clinical characteristics COVID-19 severity

2. Use an electronic CRF (on RedCap) to record clinical, biohumoral and imaging data of inpatients with COVID-19 of various severity to explore the prognostic and pathophysiological role of immunologic factors, activation of blood coagulation, endothelial dysfunction, inflammatory response, genetic (ni particular X-linked), hormonal and metabolic factors, comorbidities and acute cardiac damage. Blood samples will be collected. We will also use machine learning techniques to develop multivariable models for patients' risk stratification

3. A follow-up visit at 3-6 months after discharge will be performed to identify residual clinical consequences that might affect long-term prognosis.

Background

COVID-19 has shown a lower case-fatality rate compared to other major viral outbreaks in contemporary history, including severe acute respiratory syndrome (SARS) of 2002-2003. However, the relative susceptibility to symptomatic infection and the case fatality risk increase substantially after 60 years of age, in men, and in overweight patients, raising questions about the underlying biology of host responses. This includes possible genetic derterminants of sex bias. Cardiac involvement, as characterized, by elevation of cardiac Troponin I and brain-type, natriuretic peptide, is frequent in COVID-19 and it is associated with worse prognosis. Myocardial injury and heart failure accounted for 40% of deaths in a Wuhan cohort, either exclusively or in conjunction with respiratory failure. Thus, it seems that cardiac involvement is both prevalent and of prognostic significance in COVID-19. However, both the actual incidence of myocardial injury (biomarkers elevation may simply reflect systemic illness in critically-ill patients) and the pathophysiology of cardiac involvement remain to be clarified. The SARS-CoV-2 virus interacts through the structural glycopeptides of the "crown" spikes with its cellular target that, in humans, is the angiotensin2 (ACE2) converting enzyme, expressed in particular in the heart and lungs. ACE2 is used by SARS-CoV-2 to be internalized by alveolar epithelial cells. Therefore, chronic intake of ACE inhibitors, or sartans, may influence the course of the COVID-19 disease because an increased expression of ACE2 (such as that induced by ACEi therapy) could facilitate the internalization of the virus and the progression of infection. However, the infection by the virus leads to the down-regulation of ACE2. The imbalance between ACE and ACE2 leads to an increase in angiotensin II, which binds AT1R, which increases pulmonary vascular permeability and lung damage. Thus, the role ACEi and ARBs on the susceptibility to SARS-COv-2 infection remain to be clarified.

COVID-19 is characterized by changes in heart rate and cardiac autonomic modulation, systemic activation of inflammatory processes, with endothelial damage and involvement of cardiovascular (CV) and respiratory systems. Although most patients remain asymptomatic or mildly symptomatic, in a subset of them the host inflammatory response continues to amplify with progressive lymphocytopenia, high white blood cells and neutrophil counts, to end-up with a systemic inflammation characterized by multiple organ failure and elevation of key inflammation markers (e.g. interleukin, tumor necrosis factor, interferon-y inducible protein, etc.). These biomarkers are not just indicators of inflammation, but are also associated with prognosis. Patients who died of COVID-19 showed higher levels of IL-6, ferritin and CRP. Moreover, biomarkers of myocardial injury and ECG abnormalities were associated with elevated inflammatory markers suggesting an indirect mechanism of cardiac injury. However, recent data have demonstrated the presence of the virus within the myocardium of some COVID-19 pts, implicating also direct myocardial injury. Also low Vit.D, with immunomodulating action, is associated with poor outcome. Another interesting aspect of the complex pathophysiology of COVID-19 is the finding that 71.4% of nonsurvivors and 0.6% of survivors in a Wuhan hospital showed overt disseminated intravascular coagulation (DIC). It is well known that sepsis is a common cause of DIC and inflammatory cytokines can promote the activation of blood coagulation in many ways. However, whether SARS-Cov-2 is more prone to DIC development and the role of anticoagulation in determining the prognosis in COVID-19 need to be established. Finally, no data is available on the short-term sequelae in COVID-19 pts after discharge, in terms of residual structural and functional cardiorespiratory damage and its determinants (viral, inflammatory, metabolic and pro-thrombotic factors).

Hyphotesis and Significance

We hypothesize that COVID-19 could represent a "new" CV risk factor inducing acute and chronic CV changes able to affect clinical evolution and long term prognosis. Suggested important mechanisms of COVID-19 severity related to injury of CV and respiratory systems include: 1) a pro-inflammatory cytokine storm, with endothelial damage and DIC; 2) patients' demographic and clinical features (age, sex, body mass index, genetic factors, autonomic cardiac modulation, medical history in particular diabetes and CV diseases, sleep disordered breathing, low vitamin D levels, thyroid dysfunction, and current drug treatment); 3) evidence of cardiac damage during course of the disease. All these possible determinants of COVID-19 severity need to be systematically evaluated according to an integrated approach in a large number of patients developing COVID-19 of different severity, including inpatients and outpatients. Given the complexity of the hypothesized multifold pathogenetic mechanisms, also approaches to data analysis through artificial intelligence (machine learning algorithms) may allow to develop multivariable models to 1) effectively risk stratify patients to identify those at highest risk requiring more intensive support; 2) Promptly recognize patients most vulnerable for adverse outcomes, to prioritize palliative care and improve cost/effectiveness of healthcare resources deployed. Finally, no information is yet available on the short term residual structural and functional consequences on the immune, CV and respiratory systems following discharge of COVID-19 patients who have recovered from acute disease.

Preliminary Data

In spite of the widespread use of mechanical ventilation in patients with severe COVID-19 and hypoxemia, hypoventilation is uncommon in these patients. Conversely, hypoxemia is usually accompanied by an increased alveolar-to arterial O2 gradient, signifying either ventilation-perfusion mismatch or intra-pulmonary shunting. The presence of a significant ventilation-perfusion mismatch is further supported in COVID-19 patients by the increase of PaO2 with supplemental oxygen. Whereas, when PaO2 does not increase with supplemental oxygen, presence of intra-pulmonary shunt is the most likely cause of hypoxemia. Moreover, preliminary data from China indicate that 71.4% of nonsurvivors and 0.6% of survivors in a Wuhan hospital showed evidence of DIC. One critical mediator of DIC is the release of tissue factor (TF), a glycoprotein activator of blood coagulation cascade present on surface of many activated cell types, and of circulating microvesicles (MV). COVID-19 appears characterized by predominantly pro-thrombotic DIC with high venous thromboembolism rates, elevated D-dimer and fibrinogen levels in concert with low anti-thrombin levels, and pulmonary congestion with microvascular thrombosis and occlusion on pathology and evidence of ischemic limbs, stroke, myocardial infarction in critically ill patients. D-dimer is a biomarker of coagulation activation triggered by TF but it does not identify per se the molecular mechanisms (venous or arterial) and/or the dysfunctional cell population involved. MVs have received increasing attention as novel players in CV disease (CVD). A subgroup of procoagulant MVs express also TF, predict CV events and identify patients at high recurrence risk. COVID-19 clinical manifestations are also similar to those of other autoimmune/inflammatory disorders in which a thrombophilic vasculopathy is sustained by systemic inflammation, with activation of the complement cascade. Also low levels of Vit D and thyroid dysfunction seem to characterize more severe disease. However, the cross-link between inflammation and coagulation, as well as the role of host biology, previous treatments and clinical history in modulating the clinical course of COVID-19 remain to be clarified.

Aims

1. To investigate the epidemiological link of patients' clinical characteristics (gender, BMI, age, presence of CV risk factors, ongoing treatment, underlying CV diseases and myocardial injury) with outcomes.
2. To evaluate the pathophysiological role of: 1) activation of immune system and host inflammatory response; 2) activation of thrombotic and coagulation factors and endothelial damage (with possible DIC; 3) metabolic and endocrinologic factors, including thyroid dysfunction, low Vit.D (vs ARDS); 4) occurrence of sleep related breathing disorders and alterations in autonomic cardiac modulation; 5) genetic X-linked factors related to ACE2 expression and to gender bias 6) cardiac structural and functional changes as assessed by cardiac ultrasounds and MRI. Artificial intelligence methods will also be applied to risk-stratify patients affected by SARS-Cov-2
3. To identify the persistence of viral load, immunologic or coagulation alterations (plasma, cell or MV-related), and respiratory and cardiovascular consequences of COVID-19, by clinical/instrumental follow-up assessment at 3-6 months after discharge

Sample size

For Aim 1, in this epidemiologic survey we expect to include about 5500 patients: 4500 in-patients and 1000 out-patients. For Aim 2 and 3, because the context is underpinned by relatively sparse knowledge, ours will be considered as pilot assessments with no formal sample size calculation. For Aim 3 we will include roughly 3000 discharged patients Specific aim 1. Patients will be divided in two groups to identify outcome predictors. a) controls: individuals who did not develop severe COVID-19, b) cases: individuals who developed severe disease. The lack of enough knowledge in COVID- 19 patients about predictors of outcome limits the performance of standard regression models. Machine learning techniques can facilitate the objective interpretation of medical observations in building risk score. In particular, a combination of association rule mining with the Dempster-Shafer theory (DST) can compute probabilistic associations between clinical features and outcomes.

Statistical analysis

Specific aim 2.To identify the relationship between each potential group of predictors and in-patients prognosis, we will apply multivariate logistic regression models. All association estimates will be reported as Odds Ratio (OR) and relative 95% confidence intervals. To address the problem of variable selection in high dimensional data (numerous predictors and confounders), we will use a new statistical approach based on random forest. To overcome problems due to uncommon outcome we will consider alternative regression model as log-binomial and Poisson regression with robust variance The development of a machine-learning algorithm to identify a new score of prognosis will be based on the above results and conducted on a subsample of in-patients with all potential predictors and phenotype. The sample will be randomly divided into training (70%) and validation (30%) set. The training set will be used to build the score applying several machine learning algorithms. The score with the best predictive performance (C-index) on the validation set will be chosen by means of the two-tailed adequate hypothesis testing of equal predictive performance assuming I error type of 0.05 and power of 80%. When the null hypothesis will not be refused, the parsimony criterion will be applied.

Specific aim 3. To characterize patients at follow-up in terms of viral load and alterations of immune or coagulative systems and respiratory/cardiovascular consequences, we will apply generalized linear mixed models which take into account the correlated response during time of the same patient, modeling appropriately the variance-covariance matrix of repeated measurements.

Inclusion Criteria:

• Positivity to the test for COVID-19 and / or chest Xray or CT positive for interstitial pneumonia compatible with infection with this virus, regardless of the severity of the infection and the need or not for hospitalization
• Informed consent freely granted also verbally

Exclusion Criteria:

• Failure to satisfy the inclusion criteria

Outpatients and hospitalized patients with confirmed COVID-19 infection, recruited by General Practitioners or in Italian hospitals (mostly in Northern Italy), respectively

Sharing individual participants data that underlie the results included in published articles, after deidentification

Beginning 9 months and ending 36 months following article publication

Investigators whose proposed use of the data has been approved by an independent review committee identified for this purpose. Data will be available for individual participant data meta-analysis