Clinical Management of Adult Patients with COVID-19 Outside Intensive Care Units: Guidelines from the Italian Society of Anti-Infective Therapy (SITA) and the Italian Society of Pulmonology (SIP)

The need for hospital admission is an essential component of the initial evaluation of all patients with COVID-19 presenting at the emergency department or receiving home visits by family doctors. There is no general consensus on the optimal hospitalization criteria (either in general or in specific subgroups), and on how they should be applied in daily clinical practice [12‐21].

Factors such as older age, male sex, presence of comorbidities, severe obesity, and shortness of breath have been consistently associated with hospitalization (as endpoint) in patients with COVID-19 across observational studies [12‐28]. These factors likely reflect two different, non-mutually exclusive components: (1) an increased risk to progress to moderate/severe disease (e.g., comorbidities, male sex, severe obesity); (2) the presence of an already severe/advanced disease (e.g., shortness of breath). Independent of what they reflect, the prognostic impact of these factors is the key determinant, since the major interest is to understand which patients may be safely followed/cured at home without a consequent, unfavorable prognostic effect. Nonetheless, this is seldom assessed directly, for two possible, different reasons: (1) lack of follow-up in non-hospitalized patients; (2) the endpoint is correctly prognostic (e.g., mortality, admission to ICU, recovery or improvement of clinical status), but the population is composed only of hospitalized patients; therefore, it may remain unclear as to whether patients with a favorable outcome would have experienced the same outcome without hospital care. In other words, there is no assessment of the efficacy of hospitalization (considered as an intervention) with respect to clinical outcomes. Together with the lack of evidence from RCTs, this is a major limitation when assessing the available evidence for replying to the present question.

Notwithstanding the aforementioned limitations, the available literature provides some information about the prognostic performance in patients with COVID-19 of known severity scores for community-acquired pneumonia, such as the pneumonia severity index (PSI) and the CURB-65 score, that may provide some rationale on their possible use for deciding in favor or against hospitalization (although with the relevant bias of being assessed in inpatients and not in outpatients) [29‐32]. In a recent single-center prospective study including 249 patients with COVID-19 in Spain [32], PSI assessed at hospital admission showed overall good accuracy for predicting case fatality (area under the receiver operating characteristic curve [AUC ROC] of 0.87, with 95% confidence intervals [CI] 0.81–0.94). However, up to 37.7% patients with low PSI scores (I–III) showed progression to severe disease, likely making this tool unreliable for decisions on hospitalization. Similar results were observed for the CURB-65 score, with 26.6% of patients with a low CURB-65 score (0–2) progressing to severe disease. In another study of 208 patients with COVID-19, all with a low CURB-65 score of 0–2, 40 (19.2%) patients progressed to severe diseases [31].

Other scoring systems have been proposed for stratifying patients with COVID-19 at high or low risk for disease progression and/or poor outcome at the time of hospital admission [31‐34]. The MuLBSTA score is a fatality prediction score for viral pneumonia that includes six variables, namely multilobal infiltration, lymphopenia, bacterial co-infection, smoking history, age, and hypertension [32, 34, 35]. The MuLBSTA score has also been proposed to avoid hospitalization in patients with COVID-19, since none of the patients with less than five points showed disease progression. However, the cohort was composed of hospitalized patients and the sample size was small (n = 72) [34]. Another prognostic score, named CALL and developed in a cohort of 208 hospitalized patients with COVID-19, is based on comorbidities, age, blood lymphocyte count, and serum lactate dehydrogenase (LDH) value [31]. Overall, patients with a low CALL score of 4–6 points had less than 10% probability of progression to severe disease [31], but a formal assessment of this score in terms of internal and external validity is needed [36]. The National Early Warning Score-2 (NEWS-2) based on six physiological parameters was also evaluated for its prognostic performance in hospitalized patients with COVID-19, but any possible impact on hospitalization decisions remains to be assessed [33, 37, 38]. Besides the aforementioned, also many other prognostic scores/models have been proposed that, for the purpose of the present question, suffer from the same limitation, i.e., it is not possible to directly extrapolate their performance to hospitalization decisions [39‐53].

Finally, it should be noted that the characteristics of patients admitted to hospitals are highly variable across different healthcare systems [30] and among the same healthcare system the threshold for hospitalization may vary according to the burden of disease and resource availability. In addition, clinical severity is not the only variable guiding hospital admission. Indeed, some patients may have social contraindications to outpatient management, such as inability to maintain oral intake, fiduciary isolation, or impaired functional status. Finally, there is still no evidence that implementation of severity tools into clinical practice results in better prognosis, lower costs, and lower mortality in patients with COVID-19. None of the available tools has been currently fully validated and is ready for widespread implementation in clinical practice [30‐35].

The currently available evidence does not allow us to provide recommendations based on the GRADE system regarding the criteria of hospitalization for patients with COVID-19, owing to the lack of studies adequately assessing the impact of hospitalization as an intervention (overall and in specific subgroups). While further research efforts certainly remain necessary in this area, in the opinion of the panel it is nonetheless reasonable to set temporary hospitalization criteria (taking into account the unfavorable prognostic effect of several parameters), which are provided below as a best practice recommendation, and that follow the criteria recently released by the Italian Ministry of Health, based on expert consensus [54]. As an expression of expert consensus, we acknowledge that these criteria may be subject to modification should further evidence become available in the near future.

*This does not strictly apply to patients with chronic obstructive pulmonary disease or other chronic respiratory disease, in whom similar values may be well tolerated, but who nonetheless need a careful personalized evaluation for hospitalization considering the presence of a baseline respiratory disease besides COVID-19.

Most literature on the pharmacological treatment of patients with COVID-19 pertains to hospitalized patients. This can be attributed to many reasons, including (1) frequent unavailability of several data (laboratory results, clinical charts) for outpatients with mild disease; (2) a more clear definition of subgroups in hospitalized patients (i.e., of different phenotypes based on laboratory/radiology data); (3) an expected higher frequency of loss at follow-up in outpatients; (4) the low frequency of critical outcomes such as death in outpatients with COVID-19 (in whom mild forms are expected to predominate). For this last reason, in this review, although mortality remained an important outcome, we focused also on other relevant outcome measures (hospitalization, worsening of clinical conditions). It is also important to note that the population of interest for this question was patients with mild (or in some cases moderate) disease not requiring hospitalization. In this regard, we acknowledge that some patients requiring hospitalization may be not hospitalized (or not hospitalized promptly) in the case of exponential increases in the number of cases and overcrowded hospitals. However, this should not be the rule, and efforts should be directed towards preventing further excessive increases in the number of severe cases through efficient vaccination campaigns.

Despite the smaller amount of literature on outpatients with COVID-19 as compared to inpatients, for the present question the guidelines panel decided to rely on high-level evidence from RCTs, after having verified the absence of homogenous directions of effect combined with large effect sizes in large observational studies. There was indeed a large agreement across panel members that therapeutic decisions in outpatients (in a large portion of whom COVID-19 is expected to be a self-limiting disease) should be guided by clear evidence of a possible benefit, in line with the principle of “first, do not harm”. More details regarding the RCTs discussed in this section are available in the Supplementary Material.

It should also be noted that our literature selection was initially focused on RCTs published in peer-review journals. However, based on the possibility of impacting clinical practice, we also discussed the results of a large RCT on the efficacy and safety of colchicine in outpatients with COVID-19, which was available only on a pre-print server at the time of revision but that has been subsequently published in its final peer-reviewed form (see below).

While preliminary observational, retrospective experiences including outpatients with COVID-19 suggested a possible beneficial effect of hydroxychloroquine in terms of mortality or reduced rates of hospital admission [55, 56], these hypothesis-generating findings were not confirmed in subsequent randomized studies. The efficacy of hydroxychloroquine in outpatients with COVID-19 was assessed in a multicenter, open-label RCT trial conducted in Spain [57]. Enrolled patients were non-hospitalized adults with COVID-19 and less than 5 days of symptoms. No differences between the two arms were observed in the primary efficacy endpoint (reduction of viral load in nasopharyngeal swabs). As regards secondary endpoints, hydroxychloroquine did not reduce the risk of hospitalization, nor did it shorten the median time to complete resolution of symptoms. In a double-blind RCT, conducted in symptomatic, non-hospitalized adults with COVID-19, no differences were observed between hydroxychloroquine and placebo with respect to the primary endpoint, an ordinal 10-point severity scale [58]. Results of further RCTs published in 2021 are in line with the absence of a favorable effect of hydroxychloroquine in outpatients with mild COVID-19 [59, 60].

We found no large RCTs about the use of oral or intravenous steroids in outpatients addressing the outcomes of interests. Nonetheless, we think that some indirect evidence can be extrapolated from results of RCTs conducted in hospitalized patients with mild forms of COVID-19. More specifically, in a controlled, open-label trial, hospitalized patients with COVID-19 were randomly assigned to receive dexamethasone orally or intravenously at the dosage of 6 mg once daily for up to 10 days or to receive usual care alone [61]. The primary outcome measure was death within 28 days after randomization. In this study, mortality was lower in the dexamethasone group than in the usual care group among patients receiving oxygen (see question 4), but not among patients with milder forms not receiving oxygen. In our opinion, although indirectly, this finding, i.e., a lack of effect in patients with mild forms not requiring oxygen supplementation, can be extrapolated to outpatients, who usually share these features. Regarding inhaled steroids, in a recent open-label, phase 2 RCT including 146 participants, budesonide was compared to standard care alone in adults within 7 days of mild COVID-19 symptoms onset [62]. The primary endpoint was COVID-19-related urgent care visit, with a lower proportion of urgent visits being observed in the budesonide arm (1% vs. 14%). Keeping in mind the limitations connected to the limited sample size (which in part remain although related to early termination), this favorable result should prompt further dedicated investigation and validation [63].

Azithromycin was one of the most common antibiotics administered to patients with COVID-19 in the first months of the pandemic, based on some preliminary exploratory studies reporting a possible beneficial effect. According to the currently available evidence, the use of azithromycin does not improve clinical status in patients with mild COVID-19. At the end of our first literature review, as for steroids, the evidence supporting this statement could only be indirectly extrapolated to outpatients from RCTs conducted in inpatients, showing no beneficial effect [64]. However, the results of an open-label RCT conducted in outpatients with suspected COVID-19 aged 65 years or older (or at least 50 years with at least one comorbidity) and comparing azithromycin in addition to usual care vs. usual care alone recently became available [65]. Azithromycin was not associated with better time to first reported recovery compared with usual care alone (hazard ratio [HR] 1.08, with 95% Bayesian credibility interval from 0.95 to 1.23). Sixteen out of 500 outpatients (3%) in the azithromycin arm and 28 out of 823 outpatients (3%) in the usual care alone arm were hospitalized (absolute benefit in percentage 0.3%, with 95% Bayesian credible intervals from − 1.7 to 2.2). Overall, it should be noted that only 1148/1388 subjects (83%) had a SARS-CoV-2 molecular result available, and only 434 subjects (31% of the entire population) had a positive result. Non-superiority of azithromycin was also observed in a subgroup analysis in confirmed COVID-19 cases, which is consistent with the primary study results, although a larger imprecision of estimates should be taken into account.

The efficacy and safety of colchicine for the treatment of outpatients aged 40 years or older, with suspected/proven COVID-19, and with a least one risk factor for disease progression (see the Supplementary Material for details) was assessed in a double-blind RCT [66]. Although not reaching superiority in the entire cohort (suspected or proven infections), in the subgroup of outpatients with proven COVID-19 colchicine (administered for 1 month) was associated with a reduction in the risk of death or hospitalization, which was the primary composite efficacy endpoint (occurring in 4.6% and 6.0% of patients in colchicine and placebo arms, respectively). While no significant difference was observed in the overall study population and the small number of events in the cohort advocate for caution and against claims of large therapeutic effects, it is opinion of the panel that colchicine may be considered for the treatment high-risk outpatients, provided the results of this trial are confirmed in other RCTs of outpatients with proven COVID-19. Of note, in a previous small RCT, colchicine was associated with improved time to clinical deterioration in hospitalized patients not requiring oxygen supplementation [67]. However, the small sample of the trial does not allow either generalization to hospitalized patients or extrapolation to outpatients (based on the fact that colchicine was administered to hospitalized patients with mild disease).

Recently, the results of RCTs assessing the efficacy and safety of neutralizing monoclonal antibodies in outpatients with COVID-19 were published [68‐70]. The main target of SARS-CoV-2 neutralizing monoclonal antibodies is the surface spike glycoprotein that mediates viral entry into host cells. The neutralizing monoclonal antibodies in late phases of clinical development and approved by the US Food and Drug Administration (FDA) for emergency use are (1) the combination of bamlanivimab (also known as LY-CoV555) and etesevimab (also known as LY-CoV016), which consists of antibodies directed against different portions SARS-CoV-2 spike protein and its receptor binding domain (RBD); (2) REGN-COV2, which is the combination of the two monoclonal antibodies casirivimab (REGN10933) and imdevimab (REGN10987), which are designed to bind different regions of the SARS-CoV-2 spike protein RBD. Overall, the results of published RCTs (detailed in the Supplementary Material) show an accelerating effect of both combinations on the natural decrease of SARS-CoV-2 viral load, as well as a possible favorable effect in reducing progression to hospitalization [68‐70]. Of note, the effect of bamlanivimab/etesevimab in reducing hospitalization was maximized in patients aged 65 years of age or older or with a body mass index of 35 or greater (although the latter results stem from post hoc subgroup analyses) [69]. The Italian Medicine Agency (AIFA) recently granted a conditional approval for the use of neutralizing monoclonal antibodies in outpatients at risk within at most 10 days after symptoms onset, considering the current absence of drugs of proven efficacy for this specific setting (outpatients with COVID-19 at risk of progression) [71]. The opinion of the guidelines panel is in line with such a conditional use provided that risk factors for progression are present. Nonetheless, it should be stressed (as also recognized by AIFA) that uncertainties remain about the true magnitude of the effect, and that the favorable efficacy data are still preliminary. For this reason, the current certainty of evidence was ultimately considered low.

We did not retrieve large, peer-reviewed RCTs on the possible favorable/unfavorable effects of antivirals (except for a large RCT showing no effect of lopinavir/ritonavir compared with placebo with respect to the risk of hospitalization [60]), convalescent plasma (one study was only partially conducted in outpatients, see question 8), and prophylactic antithrombotic agents in outpatients with COVID-19 (except for preliminary favorable results for sulodexide, to be further confirmed) [72]. Finally, the possible role of ivermectin, an antiparasitic drug showing some in vitro activity against SARS-CoV-2 [73], remains unclear, with conflicting evidence and still not convincing support for its use stemming from currently available RCTs [74‐78].

A strikingly small number of published RCTs that assess the pharmacological treatment of COVID-19 in outpatients are presently available. This prompted the panel not to currently support many pharmacological treatments for COVID-19 (in line with the principle of “first, do not harm” in patients with mild disease presentation), with the exception of neutralizing monoclonal antibodies in specific situations, the possible exception of colchicine (especially after a future replication of the favorable results in patients with positive COVID-19 molecular test in the COLCORONA RCT [66]), and the exception of symptomatic use of antipyretic and analgesic agents. The panel nonetheless remains very open to possible future modifications of the recommendations, provided high-quality results from large RCTs show a clear beneficial effect either in the entire population or in specific subgroups of outpatients with COVID-19.

* Of note, there was some agreement across the panel regarding the possibility to consider colchicine for the treatment of selected subgroups of outpatients with COVID-19, provided the favorable results in patients with positive COVID-19 molecular test in the COLCORONA RCT are replicated in other studies [66].

Thrombotic complications may contribute to the morbidity and mortality of hospitalized patients with COVID-19, through various non-mutually exclusive mechanisms such as venous thromboembolism, arterial thrombosis, and thrombotic microangiopathy [79]. Consequently, prophylaxis and treatment of thrombotic complications with anticoagulant agents are largely employed in the management of hospitalized patients with COVID-19.

The literature review for the present section was performed taking into account some key concepts. First, the population of interest was represented by hospitalized patients with COVID-19 outside ICU. Second, this population could be conceptually divided into two subpopulations: (1) hospitalized patients with COVID-19 who have already developed thrombotic complications, and thus already require anticoagulant agents at therapeutic dosage, and in some cases fibrinolysis [80]; (2) hospitalized patients with COVID-19 with no evidence of thrombotic complications, in whom anticoagulant agents might be useful to prevent the development of thrombotic complications. We decided to focus on this latter subpopulation, in an attempt to answer the following clinically relevant questions: (1) Do these patients require administration of anticoagulant agents? (2) If they require anticoagulant agents, should we administer them at prophylactic or at therapeutic dosages? (3) If they require anticoagulant agents, should a specific anticoagulant agent/class be preferred to others?

To answer the first question (do these patients require administration of anticoagulant agents?), we searched for studies in which the administration of anticoagulant agents was compared to no administration of any of these agents. To answer the second question (if they require anticoagulant agents, should we administer them at prophylactic or therapeutic dosages?), we searched for studies in which the intervention was receipt of anticoagulant agents at prophylactic dosage and the comparator was receipt of anticoagulant agents at therapeutic dosage, or vice versa. To answer the third question (if they require anticoagulant agents, should a specific anticoagulant agent be preferred over others?), we searched for studies in which the intervention was receipt of a given anticoagulant agent and the comparator was receipt of another/other anticoagulant agent/s. For all questions, we defined the clinically relevant efficacy endpoints of interest to be (1) mortality; (2) need for ICU admission; (3) development of thrombotic complications. An additional safety endpoint of clinical interest was development of hemorrhagic complications.

The ideal situation would have been to base our answers to these three questions on high-quality evidence from randomized studies. However, no peer-reviewed RCTs conducted outside ICU were initially retrieved from the literature search, although it is worth noting that the results of an interim analysis combining three different RCTs (ATTACC, ACTIV-4, REMAP-CAP) has recently become available as two pre-print manuscripts, showing an advantage, for anticoagulant prophylaxis, of anticoagulants at therapeutic dosages vs. standard prophylactic dosages in patients with moderate COVID-19, independent of d-dimer values, and, conversely, an apparent lack of benefit in patients with severe COVID-19 (ICU level of care) receiving therapeutic dosages of anticoagulants [81‐83]. Although still to be peer-reviewed (and thus not considered as high-certainty evidence) these results were taken into account for drafting recommendations. Of note, a published RCT was not considered for the present question as conducted exclusively in ICU patients [84]. Finally, the results of the ACTION RCT, which became available just before release of the present guidelines, showed no advantage of therapeutic vs. prophylactic dosages of anticoagulants (enoxaparin followed by oral rivaroxaban, or, mostly, rivaroxaban alone) for improving clinical outcomes of hospitalized patients with COVID-19 and elevated d-dimer concentrations [85]. This study will deserve global discussion in future versions of the present document when full peer-reviewed results of the other RCTs introduced above will be available, also with the aim of comparing different possible anticoagulant regimens.

The majority of studies retrieved in our literature search were observational. Most of them were retrospective (33/35, 94%) [86‐118]. Only two studies were prospective (6%) [119, 120]. A brief summary of the retrieved evidence is presented in the following paragraphs and in Table S1 in the Supplementary Material, whereas an extended summary of the characteristics and results of all included studies is available in the Supplementary Material.

Before proceeding with the summary and discussion of evidence, some difficulties we found in organizing the available evidence should be acknowledged: (1) it was not always possible to limit the study populations to non-ICU inpatients; (2) in many studies, it was unclear whether admission to ICU was a baseline condition (before initiation of anticoagulant agents) or an outcome (i.e., occurring after initiation of anticoagulant agents); (3) it was not always possible to limit the population to patients without already existing thrombotic events; (4) in light of the previous point, in many studies it was unclear whether thrombotic complications occurred in patients already receiving therapeutic anticoagulants or they were instead the reason for the administration of therapeutic anticoagulants; (5) composite endpoints were ultimately not included in our review because of their heterogeneity across studies; (6) the impact of pre-existing outpatient use of anticoagulants for other conditions was not assessed because of the frequent lack of adjustment for in-hospital initiation/changes of anticoagulant agents, and the related recommendation was thus based on panel opinion only; (7) finally, considering the frequent presence of data on in-hospital use of anticoagulants in the literature on inpatients with COVID-19, some studies meeting inclusion criteria may have not been caught by our search, especially if not focused on the use of anticoagulant agents.

Overall, despite the low certainty of evidence inherent in the observational nature of most retrieved studies, a general trend toward reduced mortality in patients receiving anticoagulant agents (mainly at prophylactic dosages) was appreciable in studies with mortality as a primary endpoint. This finding, in our opinion, may support a strong recommendation towards the administration of prophylactic anticoagulants despite low certainty of evidence. This recommendation is supported by the following consequent considerations: (1) in the large majority of studies the reduction in mortality associated with the use of anticoagulant agents was more marked or present only in multivariable analyses, after adjustment for potential confounding factors; (2) this finding indicates that the use of anticoagulant was more likely in patients at worse prognosis; (3) therefore, it may be speculated that any residual confounding unaccounted for in multivariable analyses would introduce a bias against the efficacy of anticoagulant agents.

Considering the available evidence and the need for evaluating full peer-reviewed data from recently completed RCTs (see above), no recommendations based on GRADE criteria were made in the present version of the guidelines about the possible prophylactic use of higher dosages of anticoagulant agents in patients with COVID-19 without thrombotic complications, either in general or according to d-dimer values (only best practice recommendations are provided, to be reassessed the future). The same applies to the choice among different anticoagulant agents.

* These recommendations are intended for inpatients with COVID-19 outside ICU.

The rationale for administering systemic steroids in patients with COVID-19 is their anti-inflammatory effect (that on the one hand could be beneficial by counteracting excessive inflammation, but on the other hand could be detrimental by hampering the natural host response to the virus). At the beginning of the COVID-19 pandemic, the use of systemic steroids for the treatment of severe coronavirus disease was a highly controversial topic owing to an unclear balance between possible benefits and arms in previous small experiences in patients with severe pulmonary infections due to SARS-CoV-1 or MERS-CoV [121, 122]. After some exploratory, observational experiences in patients with COVID-19 (some examples are [123‐128]), the results of the RECOVERY RCT showed a favorable effect of dexamethasone (6 mg daily for 10 days) in terms of survival in two subgroups of hospitalized patients with COVID-19 (i.e., patients requiring invasive mechanical ventilation and those not requiring invasive mechanical ventilation but needing oxygen supplementation, whereas no effect was observed in patients not requiring supplementary oxygen) [61]. Besides the RECOVERY RCT, our literature search initially led to the inclusion of four other RCTs addressing (as primary population or, more often, in subgroup analyses) the efficacy of systemic steroids in inpatients with COVID-19 not subjected to invasive mechanical ventilation. The choice of patients not subjected to invasive mechanical ventilation instead of ICU patients as the population of interest for the present question was based on the frequent stratification in subgroups according to the type of ventilatory support instead of ward of stay. Thus, it was taken into account by the panel that the present review may have in part implied an indirect extrapolation of results in non-invasively ventilated patients with severe COVID-19 in ICU to non-invasively ventilated patients outside ICU. Two RCTs assessed the efficacy of systemic methylprednisolone administration in patients with COVID-19 not undergoing invasive mechanical ventilation, showing conflicting results (one reported a favorable effect on mortality and the other one did not) [129, 130]. The reasons for this difference remain partly unclear and include, beside the small sample of at least one of the two studies, differences in study design, such as the different inclusion criteria, the different doses of the drug, and the timing of administration. However, these factors seem unable to completely explain the fact that, in the larger of the two studies, no beneficial effect of methylprednisolone was observed (in post hoc analyses) either among intubated or non-intubated patients. Conversely, in the same trial there was an apparent age-dependent efficacy, i.e., a favorable effect of methylprednisolone on survival in patients aged 60 years or older. A third, small, open-label RCT including 64 patients was recently published and retrieved during the latest update of our search; that RCT compared methylprednisolone vs. standard of care alone, showing a possible advantage of methylprednisolone with respect to the primary composite endpoint of death, admission to the ICU, or requirement for non-invasive ventilation [131]. In another small RCT (86 patients), need for ventilation was registered in 18% and 38% of patients in methylprednisolone and dexamethasone arms, respectively [132]. Two other RCTs were conducted in ICU that included also patients not subjected to invasive mechanical ventilation at baseline [133, 134]. Although they were conducted in the ICU and not in the population of interest for the present review, it is of interest that also for these two trials results were somewhat conflicting. Indeed, although both RCTs were interrupted early, enrollment was discontinued for futility in one case and after the release of the positive results of the RECOVERY trial in the other one (in this latter RCT, although a significant effect was not observed in the entire population, a possible favorable effect of a fixed-dose 7-day course of hydrocortisone was detected in pre-planned subgroup analyses in patients not receiving invasive mechanical ventilation at baseline). A more detailed discussion of the results of all RCTs included in the present systematic review is available in the Supplementary Material. Another RCT was excluded as conducted only in mechanically ventilated patients [135].

Despite the limitations reported above (e.g., conflicting results with regard to methylprednisolone and hydrocortisone administration), a key aim of our literature search was that of trying to define a threshold of severity for differentiating non-invasively ventilated patients in which the intervention (steroid administration) was associated with better clinically relevant outcomes from non-invasively ventilated patients in which there was no favorable (and possibly an unfavorable) effect of steroid administration. However, it turned out that such a threshold could not be defined, based on currently available evidence from RCTs, for the following major reasons: (1) the lack of details regarding the type of non-invasive respiratory support in the RECOVERY trial; (2) the lack of information about viral clearance, which could have allowed an additional stratification for assessing efficacy in subgroups independent (or in combination) with the level of respiratory support; (3) the above reported conflicting results of the RCTs addressing the efficacy of steroids other than dexamethasone, which still deserve further investigation.

With the same aim (defining a threshold of disease severity to guide a targeted steroid administration in non-invasively ventilated patients) we also reviewed the observational literature on this subject, with a focus on the following relevant clinical outcomes: (1) survival; (2) need for invasive mechanical ventilation; (3) progression to severe disease. However, it rapidly became clear that no clear conclusion could be drawn. The major factors hampering a clear definition of a threshold of severity from observational studies were (1) the various definitions of severe COVID-19 pneumonia or disease used in the different studies (i.e., subgroups hardly comparable between studies); (2) the different steroid drugs and dosages used in the different studies; (3) the non-homogeneous definitions of primary outcome measures.

In available RCTs, a general trend towards improved survival was consistently observed with dexamethasone administration in critically ill patients undergoing invasive mechanical ventilation, whereas for inpatients with COVID-19 not undergoing invasive mechanical ventilation the evidence is less straightforward. Certainly, the positive effect of dexamethasone in inpatients with COVID-19 requiring supplemental oxygen (and of methylprednisolone in those aged 60 years or older), cannot be ignored. The evidence in support of these effects is less solid, based on post hoc subgroup analyses, but, in the opinion of the panel, sufficient to justify the use of these drugs in these populations. However, it must be underlined that these recommendations will likely need to be refined in the future according to a more precise stratification based on the type of non-invasive respiratory support (e.g., nasal cannula, Venturi mask, continuous positive airway pressure [CPAP], non-invasive ventilation [NIV]), and on other patients’ characteristics (e.g., viral load) that may be considered in the risk/benefit balance of steroid administration. Finally, as the related evidence stemmed from RCTs conducted in ICUs, albeit also in non-invasively ventilated patients, the panel decided not to provide recommendations about the possible use of hydrocortisone at the present time, pending further evidence that could be extrapolated to inpatients with COVID-19 outside ICU.

*These recommendations are intended for inpatients with COVID-19 outside ICU.

**Equivalent dosages of other steroids may be considered if dexamethasone is not available (although this should be considered as best practice recommendation, taking also into account the indirectness of evidence for steroids other than dexamethasone).

The possibility of exerting direct antiviral activity against the virus has prompted several studies to be conducted, including a non-negligible number of RCTs assessing the efficacy and safety of either already existing or novel drugs with direct antiviral activity for the treatment of patients with COVID-19. For the purpose of guidelines development, we focused the discussion of the available evidence on the results of RCTs assessing the efficacy of antiviral drugs compared to standard care alone (with/or without placebo) in hospitalized patients with COVID-19 not subjected to invasive mechanical ventilation (as the population assessed by the present question, for the same reasons detailed in question 4), with respect to the following clinically relevant endpoints: (1) mortality; (2) need for invasive mechanical ventilation; (3) clinical improvement or recovery. In the following paragraphs we summarize the major indications stemming from RCTs for guiding panel recommendations on the possible use of antivirals in the target population. A more detailed report of the results of RCTs included in the present review is available in the Supplementary Material.

At the beginning of the COVID-19 pandemic, there was great expectation that antiviral agents, based on promising results of in vitro and exploratory observational studies, could significantly improve the outcome of patients with COVID-19. However, evidence provided by RCTs has mainly indicated a lack of effect, ultimately leading the panel to discourage the use of LPV/r and HCQ, while retaining a recommendation for the use remdesivir, in which a limited (but possible) favorable effect on survival in the target population cannot be currently excluded. Nonetheless, although the initial expectations were not fully met, this certainly does not mean that efforts to find other, efficacious antiviral treatments should be discontinued, especially considering the need to enrich our knowledge about any possible, optimized combined use of antiviral plus anti-inflammatory and/or immunomodulatory agents.

*These recommendations are intended for inpatients with COVID-19 outside ICU.

Besides their mandatory use in proven bacterial infections, antibiotics are mostly administered to hospitalized patients with COVID-19 not requiring invasive mechanical ventilation in the following two situations: (1) at hospital admission, in patients with suspected pulmonary bacterial co-infection or superinfection; (2) for the empirical treatment, pending microbiological results, of severe hospital-acquired bacterial superinfections occurring later during hospital stay. At least in general, the administration of antibiotics in this latter situation may follow the common rules and clinical reasoning for the empirical treatment of hospital-acquired infections. Therefore, the main question of the present section refers mostly to the former situation (i.e., suspicion of mixed viral/bacterial pneumonia at hospital admission): in fact, it remains unclear whether or not to administer antibiotics to patients with COVID-19 presenting with consolidative infiltrates at chest X-ray/computerized tomography and with increased laboratory inflammatory markers (e.g., white blood cell count, C-reactive protein) that may resemble bacterial disease. Overall, this may lead to an indiscriminate and perilous use of antibiotics in many patients with COVID-19 without bacterial infections, likely increasing the selective pressure for antimicrobial resistance both in patients and in the environment. For this reason, we focused the first part of our evidence evaluation on the prevalence of bacterial pneumonia in hospitalized patients with COVID-19, preferably at hospital admission, and reviewed systematic reviews and meta-analyses of observational studies. In a systematic review of 30 observational studies for a total of 3834 hospitalized patients with COVID-19, the pooled prevalence of laboratory-confirmed bacterial pulmonary co-infection was 7% (95% CI 3–12, with high heterogeneity, I² = 92.2%) [169]. When excluding studies conducted exclusively in ICU, the pooled prevalence of bacterial co-infection was lower (4%, with 95% CI 1–9%, still with high heterogeneity, I² = 91.7%). Of note, the term co-infection in this systematic review apparently included both co-infection and superinfections, and both community-acquired and hospital-acquired infections. The most commonly detected organism was Mycoplasma pneumoniae, followed by Pseudomonas aeruginosa. In another systematic review, including nine observational studies for a total of 806 hospitalized patients with COVID-19 (including both ICU and non-ICU patients with no clear distinction), 64 developed a bacterial or fungal infection (no differentiation between fungal and bacterial infections was presented nor was a pooled estimate of the prevalence provided) [170].

Despite such a low prevalence of bacterial infections, the same meta-analyses registered a high frequency of antibiotics administration to hospitalized patients with COVID-19 (more than 70%), which seems not to be justified [169, 170]. Yet, the presence of a beneficial effect would theoretically support widespread antibiotic administration even in the presence of a low prevalence of bacterial diseases. Therefore, as the second part of our evidence evaluation, we searched the literature with the aim of assessing the effect of antibiotic administration (as the intervention) in terms of relevant clinical outcomes (mortality, need for invasive mechanical ventilation) in hospitalized patients with COVID-19 not requiring invasive mechanical ventilation (main population). In addition, we aimed to identify subgroups in which the intervention (antibiotic administration) was most effective. However, identification of such subgroups was ultimately not possible (see “Effect of Antibiotics Administration”). Of note, in our assessment we were able to focus only on patient-level effects of antibiotic administration. Indeed, an important limitation was (and still is) that any possible long-term consequence at population-level of widespread antibiotic therapy cannot be evaluated at the present time, owing to the only recent spread of SARS-CoV-2.

With the exception of azithromycin, there is currently no high-level evidence either against or in favor of a widespread administration of antibiotics to all hospitalized patients with COVID-19. However, the low prevalence of bacterial co-infections registered in systematic reviews of observational studies could be more in line with a parsimonious use of antibiotics, thereby discouraging their routine use in the target population. Notably, no information is currently available about the possible effect of antibiotics in subgroups stratified according to specific clinical or laboratory variables; thus further research is needed to develop dedicated prediction models of bacterial infection in hospitalized patients with COVID-19 not requiring invasive mechanical ventilation; such models could help in identifying subgroups of patients in whom to administer empirical antibiotics for suspected bacterial pneumonia.

*These recommendations are intended for inpatients with COVID-19 outside ICU.

Since the first published reports of clinical manifestations and laboratory findings in hospitalized patients with COVID-19, it was postulated that two different components contributed to the organ damage: (1) the virus; (2) a dysregulated inflammatory host response to the virus [186‐188].

From an immunological perspective, different agents may act on these two components: (1) by neutralizing key viral antigens; (2) by reducing an excessive inflammatory host response. Regarding point (2), several anti-inflammatory or immunomodulatory agents, besides steroids, were administered as off-label treatments in many hospitalized patients with COVID-19 during the first months of the pandemic, supported by observational studies of various sizes and quality, indicating a possible favorable effect [189‐192]. Subsequently, the results of RCTs assessing the effect of some non-steroid anti-inflammatory and/or immunomodulatory agents on relevant clinical outcomes (mortality, need for invasive mechanical ventilation/ICU admission, clinical improvement) in hospitalized patients with COVID-19 became available. These RCTs, together with those on neutralizing monoclonal antibodies (whereas convalescent plasma is specifically addressed in question 8), were the focus of the literature search for the present question and are discussed in the following paragraphs. As for other questions, the population addressed in the present question was inpatients not requiring invasive mechanical ventilation, owing to the stratification according to the type of ventilatory support adopted by many RCTs.

At the current time, available evidence dictates against recommendation of neutralizing monoclonal antibodies in hospitalized patients with COVID-19, at least pending the results of the REGN-COV-2 RCT in patients not requiring oxygen and patients on low-flow oxygen.

In view of the conflicting results pertaining to primary endpoints and the apparent lack of effect on mortality, the panel initially decided against recommending tocilizumab in the target population. However, based on the favorable results of the RECOVERY RCT results, the panel revised the decision, posing a conditional recommendation to consider the administration of tocilizumab in patients not responding to steroid treatment and with oxygen saturation < 92% on room air (including those already on supplementary oxygen) and with increased inflammatory markers in the absence of proven/suspected bacterial or fungal infection. The panel nonetheless recognize that this recommendation is not based on high certainty of evidence, considering the non-univocal results of the various RCTs (see above).

In view of reported favorable effects in an RCT and pending further results from other randomized studies, the panel did not recommend against the use of baricitinib in hospitalized patients with COVID-19, provided it is authorized for this indication by the pertinent regulatory agencies.

Considering the current lack of data from large RCTs for other immunomodulatory agents, they cannot be currently recommended for the treatment of hospitalized patients with COVID-19. Nonetheless, since several RCTs are ongoing, they may lead to future modifications of the current recommendations should their results be favorable.

*These recommendations are intended for inpatients with COVID-19 outside ICU.

**In the RECOVERY trial, serum C-reactive protein ≥ 75 mg/L.

***Clinicians should be aware of the following: (1) the 75 mg/L cutoff is based on results of the RECOVERY RCT; (2) other markers of inflammation may be considered on a case-by-case basis (best practice recommendation); (3) another best practice recommendation is to avoid tocilizumab administration in patients with severe immunosuppression or in those with other contraindications to tocilizumab administration (low platelet count; risk of gastrointestinal perforation; increase of transaminases greater than  five times the upper limit of normal).

The plasma of patients recovering from COVID-19 (convalescent plasma) may contain polyclonal antibodies that could neutralize viral antigens and have a beneficial effect if administered to hospitalized patients with COVID-19 [218]. The term hyperimmune immunoglobulin preparation refers to purified antibodies from convalescent plasma [219]. Several observational studies or single-arm interventional trials have been published regarding the use of convalescent plasma in patients with COVID-19 during the first months of the pandemic, with some promising results [220‐236]. However, results from RCTs are now available, on which we focused our literature search, since they are able to provide higher certainty of evidence compared to observational and single-arm studies. Therefore, in the following paragraph we summarize the results of RCTs assessing efficacy in terms of relevant clinical outcomes (mortality, need for mechanical ventilation, and/or clinical improvement) of convalescent plasma or hyperimmune immunoglobulin for the treatment of hospitalized patients with COVID-19 not requiring invasive mechanical ventilation (population assessed in the present question instead of non-ICU patients according to the stratification according to ventilatory support in major RCT). Of note, results from pre-print (i.e., yet to be peer-reviewed) RCTs are also discussed and were taken into account for drafting recommendations. As for other sections of the guidelines, pre-print results were not considered high-certainty evidence because of the lack of peer-review. Nonetheless, they remain useful for supporting (or casting doubts about) the direction of the effect registered in peer-reviewed studies.

Available results from RCTs have failed to show a substantial advantage in terms of efficacy of convalescent plasma administration in the population of interest, although the favorable dose-dependent IgG effect after convalescent plasma infusions in older adults observed by Libster and colleagues delineates a possible place in therapy in specific situations, to be confirmed by further RCTs [240]. In our opinion, the current evidence on the possible efficacy of convalescent plasma remains unconvincing and controversial (as also confirmed by the different position of various organizations and societies regarding its emergency use [249]), regardless of any favorable safety result (large data from the expanded access program have shown a low incidence of adverse events related to transfusion and no evidence of disease enhancement [218, 220, 225]). On this basis, the panel would recommend further research on this topic, with the use of convalescent plasma to be currently reserved for patients enrolled in RCTs. In fact, several reasons, including the current lack of standards about the optimal dose (in terms of antibody titers) to be transfused, the aforementioned inconsistent results of clinical trials, and the lack of efficacy information about hyperimmune immunoglobulin preparations (RCTs are ongoing), do not allow one to rule out the possibility of an advantage under specific conditions of use and/or in certain patient subgroups, which still deserves further investigation. Of note, should a favorable effect be demonstrated in ongoing or future RCTs, any possible widespread use of convalescent plasma may be hampered by the difficulties in its collection and preparation, particularly during other possible, future exponential increases in the virus diffusion.

* These recommendations are intended for inpatients with COVID-19 outside ICU.

Non-invasive ventilation (NIV, which delivers a differential air pressure during inspiration and expiration) or continuous positive airway pressure (CPAP, which delivers a constant air pressure during inspiration and expiration) are widely used to treat patients with COVID-19 pneumonia with hypoxic respiratory failure and/or respiratory distress that do not respond to standard or high-flow oxygen supplementation, for the following reasons: (1) when intubation is still not indicated (thereby possibly avoiding it in the case of favorable response); (2) in patients with “do not intubate” (DNI) orders; (3) in unfortunate situations where the number of patients requiring invasive mechanical ventilation exceeds the availability of devices to provide it.

In all these situations, NIV and CPAP are used with the intention to improve patient outcomes, but no evidence on the efficacy of CPAP/NIV for the treatment of patients with COVID-19 with acute hypoxemic respiratory failure could be extrapolated from the studies retrieved in our literature review because of the lack of RCTs or high-quality comparative observational studies. In studies employing CPAP or NIV, all patients with acute respiratory failure not responding to standard oxygen supplementation generally receive NIV/CPAP, and a controlled study in which a proportion of patients is denied either procedure would be considered ethically unacceptable.

For this reason, we decided to assess the available descriptive evidence in the population of patients receiving NIV/CPAP. More specifically, we retrieved information about the proportions of NIV/CPAP patients who (1) eventually required intubation; (2) died before or after intubation. Certainly, these descriptive, indirect findings are of questionable validity and do not provide evidence in the absence of a control group. However, either as the treatment of choice or as ceiling treatment, in our opinion, an acceptable frequency of successful administration of CPAP/NIV (survival and cure without intubation) may still provide useful insights on any potential advantage of the use of these techniques.

Finally, owing to the wide heterogeneity in indications/uses and populations of patients with COVID-19 undergoing high-flow nasal oxygen (HNFO) in observational studies, the panel deemed it currently unfeasible to develop guidance for the use of HFNO in inpatients with COVID-19. Thus, this point was not addressed in the present review (but with the commitment to consider it in future guideline updates). Very importantly, any future assessment of the role of HFNO should consider the recently published results of the HENICOV RCT, comparing NIV vs. HFNO with respect to the primary outcome of days free of respiratory support at 28 days. Of note, no difference was observed in days free of respiratory support (20 vs.18 days for NIV and HFNO, respectively), although a lower number of intubation events and a higher number of invasive mechanical ventilation-free days were observed in the NIV arm [250, 251].

Overall, 25 studies were included, all of them observational (see Table S3 in the Supplementary Material for a schematic summary and the extended evidence summary in the Supplementary Material for further details on included studies) [252‐276]. Mortality in patients treated with CPAP ranged from 14–24% [252, 256, 264, 265, 268, 271] to 43–55% [254, 260], although possibly reaching 84% when CPAP was the ceiling treatment [276]. Mortality in patients treated with NIV was reported in five studies and ranged from 5% to 52% (but reaching 86% in DNI patients) [261, 263, 269, 272, 274]. Pooling outcomes for the CPAP and NIV groups, Avdeev and colleagues estimated a mortality of 23% [255], and Bellani and colleagues estimated a mortality of 25% [270], while pooled results from Burns et al. reported a 50% mortality among DNI patients [257]. With regard to intubation (and thus need for invasive mechanical ventilation), it was observed in 41–63% of patients treated with CPAP [253, 254, 265, 273], but in some cases the proportion was as low as 11% [256]. In the study by Avdeev and colleagues, in which outcomes for patients treated with CPAP and those treated with NIV were pooled, intubation was eventually considered necessary in 28% of patients requiring CPAP/NIV, whereas intubation occurred in 15% of patients requiring CPAP/NIV in the study by Bellani and colleagues [255, 270]. This heterogeneity of results can be attributed to different reasons: (1) the small sample size of several studies, which implies large statistical uncertainty in the estimates of mortality/intubation rates; (2) the various criteria for starting CPAP/NIV adopted in the different studies; (3) the possible inclusion of DNI patients in the denominator; (4) difference in the instrumentation used and in pressure settings; (5) the different settings (general wards, high-dependency respiratory unit [HDRU], or ICU) in which the patients received NIV/CPAP. With regard to the last point, it is worth noting that for the present question we did not exclude studies conducted exclusively in ICU, as patients were inherently not receiving invasive mechanical ventilation at baseline (although the questionable generalizability of these results to non-ICU settings was taken into account when developing recommendations). A more detailed description of all the aforementioned sources of heterogeneity in the different studies is provided in the Supplementary Material, which also reports the available, limited results concerning direct comparisons of CPAP vs. NIV, which seem to indicate similar mortality, although the analyses are frequently unadjusted and possibly biased. With regard to safety of CPAP/NIV (which was rarely assessed in included studies), Aliberti and coworkers reported a low prevalence of pneumothorax/pneumomediastinum in patients exposed to CPAP treatment (1.9%) [252], while Franco and colleagues reported none [259]. Tolerance to CPAP was generally under-reported, with available data on proportion of CPAP interruption ranging from 0% to 44% [264, 268]. No information could be extrapolated from included studies regarding safety of NIV.

All the studies that assessed clinical outcomes in hospitalized patients with COVID-19 with acute respiratory failure treated with CPAP and/or NIV have an observational design, and the great majority are based on single-center data. The studies included in the analysis demonstrated a significant heterogeneity in patients’ characteristics and clinical presentation, especially with regard to the severity of respiratory failure, DNI reporting, and criteria for CPAP/NIV initiation. Moreover, the instrumentation used and the pressure settings are reported only in a minority of cases. The lack of these data contributes to the difficult interpretation of the results and limits the validity and generalizability of the findings. Finally, characteristics and outcomes of patients exposed to NIV and CPAP are sometimes pooled, implying similarities between these non-invasive approaches. However, CPAP and NIV represent very different ventilatory support modalities, needing different instrumentation and settings. Differences in mortality and intubation rates vary consistently from study to study and can be only partially explained by different respiratory failure severity at baseline or any possible different proportion of elderly or DNI patients in the sample enrolled.

Given these limitations which preclude a recommendation based on GRADE criteria, it is nonetheless the opinion of the panel, based on the proportions of survival in NIV/CPAP patients (also when used as ceiling treatment) reported in the different studies, despite their heterogeneity, that CPAP and NIV are feasible approaches in patients showing respiratory distress, persistent respiratory failure, or inadequate oxygenation despite standard supplementation. Complications, such as pneumothorax and pneumomediastinum, seems to be limited, and in the majority of cases likely caused by excessive positive end-expiratory pressure (PEEP). In the opinion of the panel, it would be reasonable not to exceed a PEEP of 10 cmH₂O, unless differently indicated by the pulmonary/critical care physician.

Considering the widespread utilization of non-invasive ventilatory approaches in patients with persistent hypoxemia and COVID-19 pneumonia, there is an urgent need for well-designed RCTs that would guide patients’ selection and suggest the best management in clinical practice.

Hopefully, results of several RCTs that are currently enrolling patients or activating participating centers (e.g., NCT04381923, NCT04326075, NCT04390191, NCT04715243) could eventually help to precisely define the place in therapy of CPAP and NIV in standardized therapeutic algorithms for patients with COVID-19.

*These recommendations are intended for inpatients with COVID-19 outside ICU.

Decisions about discharge of a patient with COVID-19 from an acute care hospital, once their condition has improved, imply two different but related considerations: (1) From a clinical standpoint, is the patient sufficiently improved to be discharged? (2) From a public health standpoint is the patient still contagious?

To properly answer these two questions, in the opinion of the panel the following specifications are required: (1) population of interest is represented by hospitalized patients with COVID-19; (2) the intervention under study is discharge from acute care hospitals, and (3) the optimal outcome is a balanced composite of clinically relevant outcomes, quality of life, and healthcare costs (the last point also implies effects of possible transmission of the virus if the patient is still contagious, thereby creating an overlap between the two questions). Through our systematic literature review, we did not find articles assessing the effect of hospital discharge in these terms that would have allowed us to confidently identify specific subgroups or characteristics of patients to optimize the decisions (and the timing) of discharge.

Considering this lack of evidence, which precluded the possibility of defining recommendations through the GRADE system, we ultimately based our recommendations on experts’ opinions. In this regard, in the official documents from national and international organizations there is substantial consistency about the opportunity to combine evidence of viral RNA clearance from the upper respiratory tract with the clinical improvement/resolution of symptoms for guiding discharge decisions [277‐279]. However, the panel underlines that this combination must be evaluated on a case-by-case basis, since contagiousness may become irrelevant for discharge decisions if the patient can adopt isolation at home or is discharged to a long-term care facility able to guarantee isolation. In our opinion, this moves the central question from “When can a patient with COVID-19 be discharged?” to “How can we reduce the excess in-hospital stay in patients with COVID-19 who still need to be isolated according to Italian regulations (see Table 2) but who no longer need hospitalization because of improved clinical conditions?” In our opinion, a proactive, organized collaboration between hospitals, long-term care facilities, and dedicated community facilities (for those isolated patients that could be discharged home but cannot adopt isolation) remains crucial for guaranteeing an adequate turnover in overcrowded hospitals and reducing any possible delay in the care of other patients with COVID-19 needing hospitalization.

The literature search for this question did not provide any evidence satisfying the agreed rigorous selection criteria (and thus providing evidence of sufficient quality to guide definition of recommendations based on GRADE criteria). In the opinion of the panel, from a clinical standpoint, a patient with COVID-19 may be discharged from acute care hospitals when oxygen supplementation is no longer required or with a maximum requirement of low-flow oxygen at 2 L/min through nasal cannula (with the exception of patients already under oxygen supplementation at home at baseline or patients requiring initiation of long-term oxygen therapy after discharge), in line with common practice with other types of non-contagious lower respiratory tract infections, and provided there are no complications or other medical reasons that require continuation of hospitalization. For patients with COVID-19 still requiring isolation but who could be discharged from a clinical standpoint, isolation outside the hospital (at home, in community facilities, or in long-term facilities, according to the specific need for non-acute care of any given patient) should be supported and made feasible for as many patients as possible, especially in the case of overcrowded hospitals.