A brief description of our sample (as reported in Table 1)
Due to the exiguity of our sample and the descriptive nature of our work, we intentionally did not test p-values for the clinical features reported in Table 1. It has already been described, and appears strikingly from our dataset, how men are more severely affected than women by Covid-19: they have been 81% of our admissions despite younger age, lower body-mass index (BMI) and lower PFmed. All of our patients were affected by one or more of the following co-pathologies: obesity, hypertension or diabetes (also one case of type 1 diabetes in a 52 year-old man). The interval between symptoms onset to ICU admission is 7-9 days, in the lower range of existing data reporting a median of 7 to 12 days[19,20]; the interval between in-hospital admission to ICU-admission is just 2 days. Our sample is likely to refer to a more severe population than those reported from China: 100% of patients transferred from other ICUs to relieve their workload arrived intubated, while 77% of local males and 100% of local females underwent intubation (other ICU reports remain below 30%)[1,19]. This is likely caused by the availability of sub-intensive units in our hospital dedicated to non-invasive ventilation.
We experienced 3 thromboembolic events (1 cerebral stroke, 1 massive and lethal pulmonary embolisms and 1 other pulmonary embolism of a patient within 48 hours from ICU discharge)[21–23], relatively high rates of ventilator-acquired pneumonias (VAP), pneumothorax and continuous renal replacement therapies (CRRT). All VAP occurred after 96h of mechanical ventilation.
We acted as a backup centre for other overwhelmed hospitals of our region, so 50% of our patients were primarily admitted in a different ICU and then transferred to us for logistic reasons. We experience a mortality rate of 30% to date (literature ranges from 20 to 45%)[19,24].
A narrative discussion of our demographic data and clinical experience
At the beginning of this pandemic in Italy, patients were admitted to hospitals only when ingravescent symptoms meant a severe progression and deterioration of their respiratory condition. This is testified by the elevated rate of High-elastance (H) pulmonary phenotypes observed at CT scans at the moment of hospital admission: severe and frequently irreversible parenchymal damage already happened. During the last weeks, the strategy has been modified: hydroxychloroquine and enoxaparin treatments have been frequently provided early at home, tocilizumab or other available cytokine inhibitors have been administered at the moment of hospital admission. While waiting for scientific validation, it is reasonable to expect a significant improvement in Covid-19 management by these strategies.
Lacking strong evidences, we have adopted for our ICU patients the following treatment protocol: 7 days of Lopinavir/Ritonavir and Hydroxychloroquine; Valsartan 160 mg per day in patients not requiring aminic pressure support [26–28], Atorvastatin 40 mg per day  and 4 days of 50 mg/kg of vitamin C four times per day . Methylprednisolone was only used as a rescue therapy in shocked patients facing MOF . The active substances chosen depended mostly on what was made available from our hospital pharmacy.
In our centre, we have adopted a percutaneous tracheostomy strategy, explaining the high rate of tracheostomies performed. Knowing the median ICU LOS of about 2 weeks, we performed tracheostomy in 81% of our patients (median 7 days after intubation; IQR 5 days). We are aware of the debate about this choice  because of possible complications, risk of droplets diffusion, rehabilitation times and costs expected for these patients. Anyway, shorter cannuale allow lower ventilatory pressures (following Poiseuille law) ; more importantly, we had the chance to safely discharge to sub-intensive units “borderline” patients, still needing frequent aspirations or some hours-per-day of pressure support for lung recruitment, any time our ICU was close to exhaustion.
After the 2 dramatic pro-coagulative complications reported above, and following increasing literature evidences on the topic [34,35], we started an internal protocol of therapeutic anticoagulation with enoxaparin (twice daily, adjusted on body weight, renal and hepatic function), together with fibrinogen dosage at least twice per week. Since that moment, neither ischemic nor haemorrhagic clinically significant events have been observed. When minor bleeding was observed (usually minor epistaxis), a RotemⓇ-driven approach was successfully applied. All 31 patients were evaluated with compression ultrasounds (CUS) every 48 hours.
After careful consideration of the ongoing guidelines and recommendations[36,37], we decided not to apply prophylactic antibiotic regimens to newly admitted patients and to suspend the prophylaxis in those arriving from other centres. As reported by early autoptic reports, very few not intubated Covid-19 patients have signs of bacterial superinfections and indiscriminate use can cause bacterial resistance (with long-lasting effects on the ICUs), lower sensibility of diagnostic cultures and lower efficacy of empiric treatments. The elevated number of VAPs experienced is likely due to both the damaged pulmonary parenchyma and the difficulty of strictly respecting hygienic procedures while wearing personal protective equipment (PPE) for many hours. Our antibiotic-sparing approach, together with genotypic microarrays for rapid molecular diagnosis from bronchoaspirates or bronchoalveolar lavages (within 4-6 hours) and, if necessary, CT scans, lead to rapid and sensible VAP diagnosis and empiric treatments waiting for anti microbiogram-targeted therapies. VAPs are a major cause of clinical deterioration in Covid-19 patients, frequently leading to a dramatic reduction in pulmonary compliance and gas exchange. Thus, extreme alertness should be posed on their detection.
The high incidence of “H-lungs”  suffering reduced compliance might have caused the unusually high rate of pneumothoraces. On the other hand, the initial attempt to relief lungs from fluid overload (with frequent echographic B-line findings) and the presence of ACE2-viral receptor in the kidney might have caused renal damage explaining the frequent resort to CRRT.
Our mortality rate (30%) is in range with those reported by European literature. We just want to recall how, in our view, Covid-19 management is based on two key milestones: prepare for mass-casualty scenarios and work as a network. Despite our province has not been severely affected, since early March we have doubled the number of ICU beds, stopped ordinary operating room schedule and relocated workforce; contemporarily we acted as a backup hospital for other centres in our region (Emilia-Romagna). To date, 40% of our Covid-19 patients have been transferred from other overwhelmed ICUs. Recommended nurse-to-patient ratio was always respected. We believe that mortality rates should not be assessed per single centres, but per region: this is an estimate of network efficiency. With current empiric SARS-Cov-2 treatments still lacking strong scientific evidence, it is essential to apply normal standards-of-care to these complex and fragile patients. Healthcare systems should implement strategies for rapid re-allocation of patients from the so-called “red zones” to every already existing ICU, trying to limit the recourse to extra-beds and the opening of brand-new emergencial ICUs.
Discussion about iron metabolism and lymphocyte counts in our sample:
Cytokines release hyper-express Hepcidin, leading to Ferroportin internalisation and reduced iron absorption and availability in body fluids [40,41]. Serum iron and TfSat are known to reduce early after infection, trying to block its onset by reducing iron availability to the pathogen, but then increase to almost-normal values within 7-10 days[42,43]. This is the same timing we observe in our patients: admitted to ICU around 7-9 days after symptoms onset still with extremely low levels of TfSat, then they present a statistically significant increase in its values. Serum iron does not show significant trends, but overall it follows the same distribution (it is part of the formula used to calculate TSat). Both TfSat and serum iron remain under the normal reference values during the whole infection.
Ferritin is a very early and non-specific indicator of inflammation. It resulted to be the first severely elevated biomarker together with lymphopenia: it is reasonable to think that their early dosage in at-home symptomatic patients might be extremely useful in individuating those who can benefit of early hospital admission. After its initial rise, ferritin can take longer than a month to normalise after an infection. Thus, it remains normally elevated in the ICU setting. Despite being apparently superficial, its dosage constitutes the key element to suspect sHLH. sHLH is a frequently misdiagnosed syndrome related to viral infections and thus of primary importance in this Covid-19 pandemics. We diagnosed at least one case in our centre.
Overall, despite unable of a more detailed description, we demonstrate that iron metabolism is deranged in Covid-19 and is likely to follow some already described patterns. We are not able to correlate it with immune response. These findings tell us that our actual ICU setting still focuses on very severe patients at an advanced state of disease. Referred to the early reports from Hubei, our patients possibly refer to the little subgroup of mechanically ventilated ones that experienced very poor survival rate. Being such a specific subset of patients, we have not observed significant differences between survivors and non. Anyway, our work might be of specific interest for researchers involved in iron and immunity and for clinicians working in ICU.
Lymphocytes are constantly reduced in ICU Covid-19 patients with respect to reference values. All the subsets are also dramatically reduced, more than reported by other recent publications referring to non-ICU populations [8,45]. We observe a conserved CD4/CD8 ratio. The nadir of lymphopenia is on day 2 of ICU stay; then, we observe a progressive tendency towards normalisation, more evident in patients experiencing positive outcome and ICU discharge. Similar timing of lymphocyte modifications were observed during Severe Acute Respiratory Disease (SARS) outbreak in 2004, despite with less dramatic reduction [46,47]. Referring to critically ill patients, this more severe reduction of all lymphocyte subgroups might be an indicator of severity: in fact, to date no other publications are available about ICU populations on lymphocytes subgroups. These data might be relevant to researchers for a better understanding of the altered immunologic response in severely affected patients.
The study is affected by some limitations. The sanitary emergency, the non-university nature of our hospital and the previously unknown characteristics of this disease led to many difficulties. Collect, analyse and communicate data has been complex and could only be performed during the scarce spare time left by healthcare assistance; during the initial phase, it has been difficult to create an effective protocol shared between the operators and some data have been missed. We could only rely on the commonly available tests present in our laboratories, without the chance to measure hepcidin and other crucial molecules that could have allowed a finer description of the investigated processes. Moreover, our centre is not located in the epicentre of the crisis and Covid-19 cases in our local community have been numerically limited; thus, we have frequently acted as a backup hospital to relieve other collapsed ICUs nearby: this means that, for a part of the reported cases, the first few days of ICU-stay occurred in a different unit and the dosage of relevant markers was missed during that initial phase. Being a monocentric study, the sample size remains limited and this could lead to the impossibility to detect and describe some more subtle physiologic processes. Finally, slightly different therapeutic approaches have been applied to patients following published findings and different complications (ventilator-acquired pneumonia, pro-coagulative states, renal and hepatic insufficiency) have affected patients’ evolution: it is not possible to quantify how these differences affected iron metabolism and lymphocyte count.