While the 2002-2004 SARS coronavirus typically led to more severe clinical presentation than seen with COVID-19 (14-20% of patients required ventilatory support, and the case fatality rate was 9.2% compared to 4.1% and 5.6% respectively for COVID-19),1 this actually contributed to effective containment of the outbreak. High rates of nosocomial transmission and severity of presentation made it easier to identify infected individuals. Nosocomial outbreaks identified in mainland China, Hong Kong, and Canada seemed to offer lessons on the potential for hospital settings to act as nexus points for the dissemination of the virus; a relatively high proportion of all infected individuals were health care workers, accounting for an estimated 37%-63% of confirmed cases.2 Specific patient procedures such as endotracheal intubation (invasive ventilatory support) were associated with nosocomial outbreaks of SARS, even when healthcare workers wore appropriate personal protective equipment.3
Given that SARS-CoV-2 is closely related to SARS, transmission dynamics among healthcare workers may be similar. Furthermore, since many individuals infected with SARS-CoV-2 are asymptomatic, it’s possible that healthcare workers are under-appreciated as potential transmission sources in the current pandemic. If infected individuals who would otherwise have a mild disease course are being over-admitted to hospitals, this may be leading to unnecessarily high exposure and infection among healthcare workers. This could give rise to a positive feedback effect leading to increased hospitalization and increased strain on capacity.
This possibility is underscored by a comprehensive analysis of SARS-CoV-2 particles in aerosols sampled in various hospital settings in Wuhan.4 Disturbingly, the authors found that concentrations were especially high in changing rooms where medical staff removed personal protective equipment, suggesting that virus particles were aerosolized just as staff were becoming most vulnerable.
The possibility that healthcare providers are acting as vectors was also highlighted by physicians on the frontlines in northern Italy, the region hardest-hit by COVID-19.5
“…we are learning that hospitals might be the main Covid-19 carriers, as they are rapidly populated by infected patients, facilitating transmission to uninfected patients. Patients are transported by our regional system, which also contributes to spreading the disease as its ambulances and personnel rapidly become vectors. Health workers are asymptomatic carriers or sick without surveillance…”
In the early phase of the pandemic in northern Italy, it occurred to me that over-hospitalization among those identified as infected with COVID-19 could be a major factor leading to the overwhelming of capacity. Indeed, it turns that out that until February 25th, Italy provided hospital admission to individuals with mild illness who tested positive for COVID-19. A sort of observer-expectancy effect born of assumptions about the course and severity of COVID-19 may have become a self-fulfilling prophecy, leading public health officials and healthcare workers to unwittingly contribute to the strain on their healthcare system, potentially increasing COVID-19 mortality. As it soon became clear that this was a mistake, Italy decided that patients with mild illness would no longer be tested.6
I believe it’s possible that differing rates of treatment and hospitalization of infected individuals, as well as healthcare workers acting as vectors of transmission, may account for many of the rather extreme disparities in apparent infection and death rates seen around the world. For example, community transmission was first recognized in both Canada and the US in late January, but Canada’s per capita COVID-19 death rate is 1/4th that of the US. It’s worth noting that Canada was especially hard-hit by SARS in 2003.7 Population density is not a reliable indicator of death rate either; some of the most densely populated areas on earth, Hong Kong and Singapore, also have lower per capita COVID-19 death rates than the US, less than 1/5th and 1/3rd that of the US respectively.
Iceland provides a potentially informative example, with 1,586 confirmed cases, 22 cumulative ICU admissions, and 6 deaths as of this post.8,9 Iceland’s testing offers what may be the single most representative sample with which to estimate COVID-19 prevalence in a population; large-scale, comprehensive screening has been carried out by deCODE genetics, and testing includes individuals who show no signs or symptoms of infection. So far, approximately 5% of the country has been tested, the highest proportion in the world. The data imply that approximately 1% of Iceland’s population is infected, which would give an infection fatality rate of 0.16%.
Crucially, Iceland’s ICU admission rate among those with confirmed infection is 1.4%. The corresponding rate for the US is 18.5% (though I can only find data on cumulative hospitalizations in the US rather than ICU admission specifically). This suggests that the number of infections in the US could be at least 13 times that of confirmed cases; if true, this implies that approximately 5 million people in the US are infected, giving an infection fatality rate of 0.21%. This IFR and the one for Iceland mentioned above are both within the range estimated by Oxford’s CEBM (0.1%-0.26%), approximately 1-3 times that of seasonal influenza.10 If some degree of over-hospitalization also contributes to the extreme discrepancy in admission rates, then this may be an under-appreciated factor driving a positive feedback effect, further increasing ICU admission and mortality.
Physicians at the epicenter of the pandemic in northern Italy have called for a shift of focus away from hospitalization:
“This disaster could be averted only by massive deployment of outreach services. Pandemic solutions are required for the entire population, not only for hospitals. Home care and mobile clinics avoid unnecessary movements and release pressure from hospitals. Early oxygen therapy, pulse oximeters, and nutrition can be delivered to the homes of mildly ill and convalescent patients, setting up a broad surveillance system with adequate isolation and leveraging innovative telemedicine instruments. This approach would limit hospitalization to a focused target of disease severity, thereby decreasing contagion, protecting patients and health care workers, and minimizing consumption of protective equipment.”5
 Ñamendys-Silva SA. Respiratory support for patients with COVID-19 infection. The Lancet Respiratory Medicine. 2020;8(4):e18. doi:10.1016/s2213-2600(20)30110-7
 Park BJ, Peck AJ, Kuehnert MJ, et al. Lack of SARS Transmission among Healthcare Workers, United States. Emerging Infectious Diseases. 2004;10(2):217-224. doi:10.3201/eid1002.030793
 Severe acute respiratory syndrome–Singapore, 2003.
Centers for Disease Control and Prevention (CDC).
MMWR Morb Mortal Wkly Rep. 2003 May 9; 52(18):405-11.
 Liu Y, Ning Z, Yu C, et al. Aerodynamic Characteristics and RNA Concentration of SARS-CoV-2 Aerosol in Wuhan Hospitals during COVID-19 Outbreak. bioRxiv. doi:10.1101/2020.03.08.982637
 Nacoti M, Ciocca A, Giupponi A, et al. At the Epicenter of the Covid-19 Pandemic and Humanitarian Crises in Italy: Changing Perspectives on Preparation and Mitigation. NEJM Catalyst. 2020 21 March. doi:10.1056/CAT.20.0080
 Van Beusekom M. “Italian doctors note high COVID-19 death rate, urge action”. CIDRAP, University of Minnesota. March 24, 2020. Retrieved April 7, 2020.
 Wong T, Wallington T, McDonald LC, et al. Late Recognition of SARS in Nosocomial Outbreak, Toronto. Emerging Infectious Diseases. 2005;11(2):322-325. doi:10.3201/eid1102.040607
 “COVID-19 á Landspítalanum”. (Icelandic) Retrieved April 7, 2020.
 “COVID-19 Update: 2 More Deaths, Half Of Cases Asymptomatic”. The Reykjavík Grapevine. April 6, 2020. Retrieved April 7, 2020.
 Heneghan C, Oke J. Global Covid-19 Case Fatality Rates. CEBM; updated 6th April 2020. https://www.cebm.net/covid-19/global-covid-19-case-fatality-rates/. Retrieved April 7, 2020.