COVID-19 Evidence Bites, Updated June 9, 2020

Filtered and Appraised by
Supervising Editor: Shahriar Zehtabchi, MD

This article was published in collaboration with MDCalc. Please see the MDCalc COVID-19 Resource Center for more information, including a critical review of recommended calcs for resource-limited situations and more.


What Is The Overall Case Fatality Rate of Covid-19? 

EVIDENCE BITE: This is not yet known definitively, but the answer most likely lies between 0.5 and 2%

SUMMARY: Most major variation is probably attributable to reporting anomalies. Extremely high fatality rates, as those seen early in Italy, are likely a reflection of both an elderly population and testing limited to the sickest cases. Lower rates, like those seen at first in Germany, were probably a reflection of a younger population initially exposed and widespread testing. As of this writing, for instance, Brazil’s fatality rate is roughly 5% according to the testing data, and this is almost certainly an artifact of inadequate and selective testing. In all cases the numbers mostly normalize as testing becomes widespread and more representative. Up-to-date case and case fatality rates are available here, here, and here.

What Is The True Accuracy of The Swab (RNA-PCR) Test? 

EVIDENCE BITE: When results are positive the test is usually correct, but it misses a lot of Covid-19.

SUMMARY: The Reverse Transcriptase-Polymerase Chain Reaction test is deeply flawed. When the test finds SARS-CoV-2 it seems to be overwhelmingly correct, but it often misses the virus. One official in China reported in February the PCR test was detecting only 30-50% of Covid-19 cases. There are few well reported, reliable studies, but one study examined swab results from body fluids in 205 hospitalized Covid patients. The virus was detected in just 32% of pharyngeal specimens, 63% of nasal specimens, and >90% was seen only in fluid from deep lung endoscopy. Another suggests the test misses only 20% of infections at 3 days after symptom onset, but more than that for all other days including missing all infections on the first day and 66% of infections at 3 weeks. A recent perspective piece further explored this literature, concluding 70% sensitivity is a best-case scenario. 

This problem seems to have worsened with the rapid PCR that is now among the most common. A Medscape report highlights one study that showed the rapid Abbot test missed about half of infections compared to a non-rapid competitor PCR. Since competitor PCR tests seem to peak at 80% sensitivity this is particularly disturbing. 

Confirming these findings, blood antibody response and CT scan reports have both suggested the PCR tests miss frequently (nearly half based on antibodies, and a third based on CT), and reports comparing PCR to CT find the same.

At least one author group has explored the potentially catastrophic ramifications of relying on the PCR swab test for pandemic management, as we are currently doing. New technologies or innovative use of current technologies will have to step up. One potential approach has been suggested combining imaging and PCR testing, we eagerly await further advances or innovations. In the meantime it is generally unsafe to rely on PCR testing to rule out Covid-19.

What is the Survival Rate for Covid-19 With Invasive Mechanical Ventilation?

EVIDENCE BITE: Anecdotes and case series suggest overall survival is low with endotracheal intubation and ventilation.

SUMMARY: This is a moving target. Based mostly on aggregated anecdotes it appeared, early on, most endotracheally intubated patients quickly exhibited major physiologic decompensation. Reports of hypercoagulable state, agitated delirium, abrupt renal failure, and worsening ARDS following intubation were, and remain, common.

In a prominent chart review of Chinese cases 86% in a small cohort of invasively ventilated patients died. Other pre-publication reports found varying, but often similarly low survival rates, substantiated by one early case series. A case series from two academic centers in New York City offered a more hopeful picture (15% of intubated patients died, though follow-up was limited), but this was followed by a much larger chart review of 5700 hospitalized patients from the same area in which 88% of ventilated patients died, including 97% of those over age 65. 

As a result there has been a push to innovate before invasive ventilation using proning, nasal 02, NIPPV, and other methods. According to virtually all involved, and limited case series data, this is yielding fruit. However current best estimates of mortality with mechanical ventilation remain high, at 30-40%, though this is not nearly as high as the earlier estimates, which is encouraging.

How Likely Is Someone To Contract Covid-19 After Close Contact With Symptomatic Cases?  

EVIDENCE BITE: Not likely. Roughly <1-5% of close contacts end up falling ill with Covid-19, and about 10-15% of within-household close contacts do.

SUMMARY: There is a broad experience with rigorous contact tracing, quarantining close contacts (usually defined as those within 1 meter for more than 15 minutes, or in some cases 1 hour), and watching for development of disease. According to the WHO and reports from the Chinese CDC and American CDC, the overall rate of new Covid-19 infection in these settings ranges from 0.5% to nearly 5%. Within households, close contact led to predictably higher transmission, but still only 10-15%. 

Obviously, avoiding all contact with patients suffering any level of illness from Covid-19 is the safest approach. However, it is notable that tracing consistently shows low attack rates among even close contacts. The nature of contact does seem to matter, as those in close quarters and crowded situations do appear to be at higher risk. This strongly suggests social distancing and other physical measures can be effective means of lowering the R-naught.

Can Covid-19 Be Transmitted by Asymptomatic People?

EVIDENCE BITE: Asymptomatic transmission is probably less important (in public health terms) than symptomatic transmission

SUMMARY: There are a number of reports of possible asymptomatic transmission, and there continue to be reviews amplifying these reports (and mathematical modeling based on them). However, proving the phenomenon, or establishing its role, remains tricky. The WHO continues to assert that symptomatic patients, even those with mild symptoms not easily visible (e.g. mild cough or fever) are the main transmission source. They acknowledge reports of asymptomatic transmission, but state it is “not yet known how often it happens.”

After literature and contract tracing data review, we agree with this assessment.

One commonly cited report of putative asymptomatic transmission includes a supplementary appendix describing the index case as clearly symptomatic. The second, an asymptomatic woman whose family became ill, occurred in an area with public contact (no lockdown) but may have been asymptomatic transmission. The third case describes a gentleman who transmitted to his family and others on a hospital ward before reporting symptoms. In addition, one contact tracing report from Singapore found 6% of new infections could only have been from pre-symptomatic persons. 

Finally, a well investigated nursing home outbreak suggests some patients reporting no symptoms were likely transmitting. Over half, however, were cognitively impaired and others had baseline cough, making the data fraught. 

In contrast, contact tracings of many hundreds of thousands of cases around the world have identified index events of contact with a symptomatic patient in the overwhelming majority. Computer modeling studies have suggested patients with incubating infection may be infectious up to 2 or more days prior to symptoms, though these studies and contact tracings rely on patient recall. One South Korean report discussed patients with mild symptoms which the authors termed ‘subclinical’. These individuals were important spreaders and had symptoms like sore throat and cough mild enough not to be apparent without directed examination and interview. 

As the world becomes increasingly aware of Covid-19 and its dangers, and people more assiduously avoid symptomatic persons, pre-symptomatic or subclinical spread may be increasingly important. Indeed with significant social distancing, widespread awareness, and mask wearing, one would expect to see asymptomatic transmissions quickly overtake all other forms. At this point, however, truly asymptomatic transmission seems, based on contact tracings, to remain, at most, a minority of transmissions. 

Can One Person Be Infected By Covid-19 Twice?

EVIDENCE BITE: Case reports suggest this may happen; it appears to be rare

SUMMARY: First, the reason this issue is important is not individual risk (suffering a second bout). That is bound to be frustrating and difficult, but the public health concern is transmissibility. If those with prior infection can still infect others, re-population of public spaces may be unsafe until there is an effective vaccine. 

We have reviewed emerging information including news and scientific reports about those who retest positive. By far the most useful and extensive data are from the South Korean CDC which describes 447 cases of testing positive after recovering from Covid-19. Of these, 284 were symptomatic when tested, a median of 44 days after first infection onset. These patients underwent viral culture and 790 persons with prior contact to them were quarantined. No viral cultures grew (i.e. they weren’t infectious cases) and no new Covid-19 infections were found among the contacts. These results jibe with extensive worldwide contact tracing, and with monkey studies suggesting infection is protective. 

Conversely, one non-peer reviewed examination of common (non-SARS) coronavirus infections found a meaningful percentage (12 of 86 infected patients, 14%) tested positive for the same coronavirus strain within the year. Most (9 of 12) were children, and the three adults experienced minimal or no symptoms. There was no evaluation of transmissibility. One press report we are aware of and a case report of two elderly patients both appear to document second symptomatic bouts of Covid-19 disease, though again no transmissions from these are known or reported.

In addition, one concerning serum antibody study suggested 30% of patients mounted little or no detectable antibody response, feeding fears of non-immunity. A second recent study, as well as others now, however, have found 97-100% exhibiting strong responses. The reason for the discrepancy is unclear, but we believe (hope) the latter is correct.

For now reinfection with documented transmissibility remains unreported. Fingers crossed.

If You Have An Upper Respiratory Infection, How Likely Is It To Be Covid-19?

EVIDENCE BITE: Likely. We’re in the midst of a global pandemic of Covid-19, and influenza activity is extremely low.

SUMMARY: There is no excellent evidence to answer this, but it is relevant to ‘pandemic thinking’. The prevalence of Covid-19 in quarantines and among fever clinic populations has been surprisingly low. Moreover, most state level testing in the U.S. shows low rates (10-30%) of SARS-Cov-2. This may be misleading since both asymptomatic people and hospitalized patients with known alternative diagnoses are included. If exclusively people with URI symptoms or influenza-like illness in active pandemic areas were tested, the positive rate would likely be higher. Based on current trends and disease activity in the United States it is reasonable to estimate that a URI is more likely to be due to Covid-19 than any other single cause in areas with active outbreaks. While it may not be more than half, it is likely no other agent would be causing more such illnesses, by frequency.

Therefore, based on sound reasoning and cautious (pandemic) thinking, everyone with even mild URI or influenza-like symptoms should be self quarantining immediately, watching closely for symptoms, and finding a place to be tested. As noted above based on the properties of the test, a negative test does not mean a person with symptoms does not have Covid. The PCR test is notoriously unreliable. However a positive test is correct, and helpful. All persons who quarantine should do so until they meet (at minimum) CDC clearance criteria described below, i.e. >7 days from onset, 3 days without fever, and with consistent symptom improvement.

How Long Am I Infectious Once I Have Covid-19?

EVIDENCE BITE: In mild cases infectiousness probably ends once fever is gone and other signs and symptoms have improved for at least 2-3 days. In severe and persistent cases infectiousness may continue for weeks.

SUMMARY: There is minimal data to inform this. The CDC’s approach has been to accept a best guess using known properties of other coronaviruses and other viral upper respiratory illnesses. With this background they suggest people are no longer infectious when a) symptoms are improving, b) fever has been absent for 3 days, and c) at least 7 days have passed from onset of symptoms. Case series’ and other recent reports confirm infectiousness (based on viral load, a surrogate marker) is probably greatest in the early days of symptoms.

None of this is well studied, and one chart review from China suggests ‘viral shedding’ is possible up to 37 days from initial infection. However, what the authors meant by ‘viral shedding’ is a positive viral RNA (PCR) test, which is quite different from being able to transmit Covid-19—the issue that matters most. Moreover, based on case tracing data from the WHO, China, and others, the overwhelming majority of transmission has occurred due to early symptomatic spreaders, and perhaps rarely asymptomatic persons. There are no reports we are aware of detailing transmission of the virus by post-recovery persons. 

Are There Known Predictors of Severe Covid-19 Disease?


SUMMARY: There have been many publications and pre-publications reporting associations between clinical characteristics and severity of Covid-19 infection or outcome. One peer reviewed dataset we are aware of is here, and one of the larger, more methodologically rigorous studies we are aware of is here. A recent, methodologically rigorous example is here and another is here, with a handy online risk calculator. We also summarized some relevant early data on indicators of mortality. However while, for instance, increasing age correlates with outcome, and integrating this with existing judgment is wise, predictors from case reviews are inherently unreliable. They are cherry-picked from datasets retrospectively and for Covid-19 often have limited follow-up. Prospective validation studies are always necessary, and until then physician judgment will typically be better. Even the most rigorous studies noted above have not been prospectively validated (despite language that might suggest otherwise). For clinicians seeking to predict outcomes we suggest carefully considering these data as part of, not as a replacement for, best judgment.

Should Affected Populations Be Universally Wearing Masks In Public Settings?

EVIDENCE BITE: We think so

SUMMARY: The question of whether to wear masks is important. For broad psychology (‘pandemic thinking’) alone we believe this may be useful, but we have no hard data to support that belief.

It is worth understanding how this might work, however. While individuals not yet infected are theoretically not protected by masks (hands, touching, and droplets spread the virus) actively infected patients may spread less with a mask. Therefore the utility of widespread masking is based on hoping those infected are wearing them as well.

Frustratingly, a small lab study suggested the virus can be transmitted through both cotton and surgical masks with coughing, though the authors retracted the data upon request by the editors, seemingly for under-reporting the possibility of cross-mask transmission. Moreover, the WHO does not see masking as independently adequate for protection. However, it still seems likely that the transmission is reduced by masks. The masks are guaranteed to reduce the concentration of virus emitted, and probably reduce virus escape during normal breathing. They would also cut down on droplet spread and may protect against airborne droplets landing on mucus membranes. 

There may be other benefits. For one, wearing a mask may remind people to more consistently socially distance and take other important physical precautions that reduce spread. For all of these reasons, and because we believe it is wise to simply emulate successful methods from other countries, we recommend wearing masks in public settings.

Can Covid-19 Be Transmitted Through Airborne Spread?

EVIDENCE BITE: True ‘airborne transmission’ of SARS-CoV-2, if it occurs, is probably much less important (in public health terms) than contact spread

SUMMARY: Clearly it is possible to cough, sneeze, or even breathe, speak, sing, or sputter droplets into the air, and thus onto other people or surfaces in close proximity. It is widely known that influenza and other common respiratory viruses can be transmitted this way. However it seems likely from contract tracing data and SARS-CoV-2’s known properties that it is primarily spread through means of direct contact (touching a person or object that has active virus particles, then touching one’s own face). 

The broad confusion about whether airborne spread occurs with SARS-CoV-2 may partly be jargon. When experts describe a virus as ‘airborne’ they mean it is aerosolized, the particles are small (typically estimated at <5 micrometers, thus able to travel to deep lung spaces), and high concentrations in the aerosolized matter can transmit infection.

Thus ‘airborne spread’ does not mean aerosolized droplets, which are common to all viruses. Measles, chicken pox, and smallpox exhibit airborne spread, and measles has an R0 of 15. This means 15 (non-immune) people are, on average, infected by each sick person. Covid-19 R0 has been estimated most recently by the CDC at 5.7. 

Thus while occasional travel of viral particles through the air may lead to infection in some Covid-19 cases, SARS-CoV-2 droplets are mostly heavy and fall quickly to the floor. They generally do not remain airborne and are not small enough to travel deep in the lung. Moreover, most cases around the world have reliably identified an instance of physical interaction or extended close contact (conversing within 6 feet for >15 minutes) with a Covid-19-infected patient.

What Antiviral Therapies Show Promise For Managing The Covid-19 Pandemic?

EVIDENCE BITE: Possibly dexamethasone


Dexamethasone: A single large randomized trial, only available in pre-publication as of this writing, found a 3% (NNT 33) absolute mortality reduction among >6000 subjects randomized to dexamethasone or no dexamethasone. This benefit seemed most robust among those receiving mechanical ventilation (nearly 12% reduction, or NNT of 8), while the drug may have harmed those who were least ill and who received the drug early in their course. The trial was not placebo controlled or blinded, important points of potential bias, and will need to be replicated under more rigorous conditions for confidence. However use of the drug after hospitalization in patients requiring oxygen seems like a wise approach given these data. We eagerly await further trials, as well as peer reviewed publication.

Remdesivir: this nucleotide analog was developed to treat hepatitis, for which it failed, and later ebola, for which it also failed. The first published trial of 237 subjects in Wuhan, stopped early as the outbreak abated, nonetheless failed to show benefit in any measure (even viral load was unchanged, theoretically the drug’s mechanism of action). A second, larger trial of over 1000 subjects, however, found a 4-day benefit (15 vs. 11) in ‘time to recovery’, though no effect on mortality. Problematically, the primary outcome of this trial was changed after data collection began. The original primary outcome was “percentage of subjects reporting each severity rating on an 8-point ordinal scale” including death, mechanical ventilation, and other important endpoints. The new primary outcome, however, was not listed in the researchers’ original 30 outcomes of interest. Based on the drug’s history of failure our prior probability estimate for the effectiveness of remdesivir was low. This combined with outcome switching and irregularities leave us skeptical. We await attempts to replicate this outlier result and hope to see future trials focus on mortality.

Hydroxychloroquine: Excitement emerged in the wake of a case series suggesting HCQ reduced viral load; this was then refuted by a similar second series. At this point there are multiple systematic reviews of the effects of HCQ and CQ, some published, others pre-publication. Results from randomized trials are uniform (most recent example we know of here), though data are weak: no clinical benefit. One large, recent trial using HCQ for post-exposure (median 3 days post) also showed no benefit. At this point there is ample data documenting adverse effects as well. The drug should therefore only be used in research settings.

Convalescent plasma: This therapy was first formally reported in 5 people with severe, ventilated Covid-19 disease. We are aware of a single randomized trial of 103 subjects, terminated early. While no statistical benefits were found for any outcome, the results suggest clinically meaningful differences are possible and more rigorous, better trials should be conducted. 

Tocilizumab (brand name: actemra): Tocilizumab is a recombinant monoclonal antibody used to treat moderate to severe rheumatoid arthritis and idiopathic juvenile arthritis. Since the drug blocks the interleukin-6 (IL-6) receptor, it has also been used to treat the cytokine release syndrome in a specific form of immunotherapy to treat certain cancers. Therefore, it has been proposed that tocilizumab could potentially treat the cytokine storm in COVID-19 infection. Unfortunately, so far only one case series of 20 patients has reported any benefit (clinical improvement such as decreased oxygen requirement and resolution of fever) in COVID-19 patients. The FDA recently approved a clinical trial to assess the efficacy and safety of tocilizumab for treatment of COVID-19 pneumonia in hospitalized patients.

Protease inhibitors: Despite theoretical promise, the drugs failed to show a benefit in ‘time to improvement’ among 199 patients with severe Covid-19 in China. In a second trial of ‘triple therapy’ with a protease inhibitor, interferon, and ribavirin, 127 patients with mild Covid-19 experienced a 5-day benefit in the time it took for swab tests to become negative. There were also mild symptom benefits. 

In summary, preliminary data support dexamethasone as a Covid-19- therapy that may reduce mortality or need for mechanical ventilation. No other agents appear to have demonstrated an important effect as of this writing. 

Is N95 Mask Reuse Safe After Contact With Covid-19 Patients?

EVIDENCE BITE: N95 mask reuse is probably less safe for the wearer than single use, but certain measures reduce risk 

SUMMARY: The Covid-19 pandemic has led to severe shortages of N95 masks, forcing hospital workers to reuse them. There is no data from Covid-19 patient settings, but studies performed with similar viruses probably provide the best available approximation of risk. On a practical note, wearing masks constantly for long periods is exhausting, makes communication difficult, and many practitioners will break safety routines. The procedures described here are difficult to follow.

Covid-19 seems to be transmitted primarily via hand contact with respiratory droplets. With mask reuse self-inoculation is possible, just as it is during doffing procedures. Modeling studies with influenza suggest a moderate risk of self-inoculation with reuse. Studies also suggest the risk conferred by mask contamination during aerosol-generating procedures can probably be reduced by washing hands before donning and doffing, placing a surgical mask over one’s N95, and by having patients wear a mask whenever possible.

Mask fit problems, common after donning and doffing, can probably be mitigated by fit testing periodically, or after donning, with a diluted saccharine spray (something that might be easily improvised with Sweet N Low and an atomizer). The risk of nosocomial spread from a contaminated N95 via a sneeze or cough is probably negligible.

Based on our review we would suggest extended mask use rather than removing a mask between patients. Multiple masks per provider is best (the CDC recommends 5, one to be used each day, then rotated). Ideally these would be worn all day, with 3 being used for roughly 4 hours each during a 12-hour shift to allow for removal when eating, and during a presumptive second break. Leaving masks to self-decontaminate over 2-3 days is probably effective, and other methods such as hydrogen peroxide vapor, 70% ethanol spray, and heat or UV light, show mixed results.

For aerosol generating procedures, particularly intubation, if possible it is safest to use a new mask (even over one’s N95) and discard or decontaminate immediately afterward.

What is The Likely Success Rate of CPR Among Patients With Covid?

EVIDENCE BITE: For most hospitalized Covid patients with cardiac arrest, functional survival is probably <1%, and risk for healthcare workers during CPR is high; CPR should be rare when Covid-19 patients arrest.

SUMMARY: Covid-19 is primarily characterized by pneumonia. As such, there are two precipitating scenarios representing the majority of cardiac arrest events during the illness. The first is hypoxia, a potential precipitant in out-of-hospital arrests, or in the peri-intubation period. The second, by far more common, occurs in critical care settings following mechanical ventilation, sepsis, ARDS, and multi-organ failure. The validated GO-FAR score suggests functional survival after CPR in such a scenario is near zero (<1%), regardless of age.

The only chart review series we are aware of supports this notion, documenting 136 patients with known Covid-19 who underwent CPR after cardiac arrest. There was one neurologically intact survival. 

In comparison, limited retrospective data from the SARS outbreak suggests baseline risk of infection among healthcare workers may be roughly 10%, while involvement in CPR by, for instance, performing chest compressions is associated with a 3-5x higher risk. If causative, this represents an absolute risk increase of 40%, an enormous potential for harm. Provider infections also, by extension, harm patients via nosocomial spread and, in pandemic settings, via loss of a critical human resource that will now, at minimum, be quarantined for weeks. Recently released AHA guidelines implicitly recognize these issues, suggesting the two first steps after cardiac arrest in Covid patients should be 1) “limit personnel”, and 2) “consider resuscitation appropriateness.”

In summary, most in-hospital Covid-19 arrests represent near zero potential for functional survival. Exceptions will be younger patients without multi-organ failure or without comorbidities, or perhaps those arrested due to isolated hypoxia. In the absence of these variables, based on survival potential and provider risk, we feel CPR is best avoided.

Supervising editor: Shahriar Zehtabchi, MD