A Risk-Management Approach to Defeating SARS-CoV2 and COVID-19
In 1921, Professor Frank Knight (an economist at the University of Chicago) published his most famous work, Risk, Uncertainty and Profit, where he differentiated ‘risk’ (comprising the realm of future unknowns that depend “on the future being like the past”) from ‘true uncertainty’ (those situations where the future is not just unknown, but truly unknowable, because of an extreme lack of similarity with any relevant prior cases). As such, he quipped that
[true uncertainty occupies that space where] opinions (and not scientific knowledge) actually guide most of our conduct(p. 233).
Unfortunately, we are in the midst of a global pandemic that resides much closer to the realm of true uncertainty than risk, giving rise to myriads of ‘opinions’ but scant ‘scientific knowledge’ that is truly actionable. Six weeks ago, the World Health Organization, the Centers for Disease Control, the U.S. Surgeon General, and other ‘experts’ were all telling us that masks and face coverings were “not effective against respiratory illnesses like the flu and COVID-19” with the Surgeon General specifically stating that people “who don’t know how to wear them properly tend to touch their faces a lot and actually can increase the spread of coronavirus.” However, now all three are actively encouraging citizens to wear face coverings to help “slow the spread of COVID-19”.
It seems that even the world’s ‘experts’ are adhering to Frank Knight’s prophetic predictions about how
the human reaction to situations of [true uncertainty are] apt to be erratic and extremely various from one individual to another, but the ‘normal’ reaction is subject to well-recognized deviations from the conduct which sound logic would dictate(emphasis added, p. 235).
With that said, my aim with this blog post is to follow ‘sound logic’ in applying a risk management perspective to the COVID-19 pandemic. “Risk management” is a term commonly used in the fields of engineering, manufacturing, and business. The term is equally applicable, but less frequently invoked, with respect to healthcare. Risk management involves managing exposure to risk by identifying potentially-adverse future events, assessing the likelihood and severity of those events, then proactively mitigating the likelihood and/or severity of the events.
I am not a physician, nor am I an expert in epidemiology or epidemiologic models. However, I am intimately familiar both with the development of quantitative models and with issues related to the management of risk and uncertainty. I have an undergraduate engineering degree from MIT, a PhD in engineering from Oklahoma State University, and a PhD in entrepreneurship from Oklahoma State. My engineering dissertation involved the development of a quantitative risk-management model for controlling cost overruns associated with capital-intensive construction projects. My entrepreneurship dissertation examined the ways in which entrepreneurs perceive and manage risk and uncertainty, with an emphasis on decision-making in the absence of relevant prior knowledge.
With that said, my goal here is to lay out a path forward that intentionally skews toward the ‘risk’ side of the spectrum and away from decisions that are wrought with ‘true uncertainty’ (i.e. an inherent lack of similitude with prior cases).
Cutting to the chase, here are my overarching conclusions:
- Applying social distancing guidelines uniformly across all risk categories will result in 10x more COVID-19 fatalities compared to a simple two-pronged approach, namely loosening the social distancing restrictions on those least likely to develop serious complications while tightening the guidelines for protecting those who are most vulnerable (coupled with providing robust community support to ease the burden of isolation).
- The path to population immunity should not proceed in such a way that would risk overwhelming local critical care resources. To that end, I still recommend a trigger-based approach for implementing appropriate changes in social distancing whenever usage of critical care resources exceeds predetermined thresholds.
As I conducted my risk management analysis, I relied upon the following core pieces of information:
- Population immunity is the key to defeating the SARS-CoV2 virus, which will require either
- widespread vaccination, and/or
- widespread exposure and recovery.
- (It bears noting that development of an extremely effective ‘treatment’ that dramatically reduces the fatality rate would allow even safer ‘widespread exposure’ and, as such, should be aggressively pursued. However, we cannot simply ‘wait for a miracle treatment’, as it took over ten years before an extremely effective treatment of HIV/AIDS was developed.)
- Waiting for a vaccine places immense societal risk upon a single discrete outcome (achieving a safe and effective vaccine) that requires two separate NEVER-BEEN-DONE-BEFORE achievements:  developing a safe and effective vaccine for a class of viruses (human coronaviruses) that we have never before conquered (despite previous attempts, e.g. SARS, MERS) and  doing that FASTER than we have ever developed ANY vaccine.
- At best, a safe and effective SARS-CoV2 vaccine is MANY months (or years) away. A recent analysis in the New York Times (based on input from vaccine-development experts) showed that the ‘typical’ timeline for (routine) vaccine development leaves us waiting until May 2036 (16 years!) before widespread vaccinations for SARS-CoV2 could likely occur. NYT then showed a host of extremely optimistic changes (to the normal development and approval processes) that could conceivably (if everything falls perfectly in line) result in a ‘best case scenario’ of vaccinations beginning August 2021 (15 months from now, if their proposed never-been-done-this-way-before process goes perfectly as ‘planned’).
- However, scientists have NEVER (not once) developed a successful vaccine against a human coronavirus. As such, a safe and effective SARS-CoV2 vaccine might never actually materialize (similar to the unfortunate reality that a vaccine for HIV has eluded scientists for over 40 years). It bears noting that SARS-CoV2 and HIV are both ‘enveloped viruses’ (i.e. the viral material is enclosed in a lipid bilayer envelope), which makes them more difficult to vaccinate against, but (fortunately) easier to disinfect and kill outside their host.
- Early on, there was disagreement regarding whether recovery from a SARS-CoV2 infection would confer immunity; however, the early reports of ‘reinfection’ were found to be caused by testing errors, with most experts now believing that “an initial infection from the [SARS-CoV2] coronavirus … will grant people immunity to the virus for some amount of time … [which] is generally the case with acute infections from other viruses, including other coronaviruses.” (As such, although unknowns still exist with respect to the duration of immunity after exposure, relying upon exposure and recovery to confer immunity is clearly on the ‘risk’ side of the uncertainty spectrum.)
- Johns Hopkins epidemiologists suggested that 70% population immunity will likely be required before the SARS-CoV2 virus will cease being a major threat. Harvard epidemiologists used a 50% threshold, but also acknowledged the potential for cross-immunity against SARS-CoV2 from existing common-cold coronaviruses (such as HCoV-OC43 and HCoV-HKU1), which could conceivably reduce the required level of population immunity below 50% (although this is completely out of our control, let us hope and pray that this is true!).
- Strong social distancing can dramatically reduce a current peak of SARS-CoV2 infections and COVID-19 fatalities; however, strong social distancing results in scant development of population immunity, which means a future peak of infections and fatalities remains highly likely (and, with potential seasonal effects, the future peak could actually result in more infections and more fatalities, as demonstrated by the Harvard model). Numerous past pandemics have come in ‘waves’, with the Spanish Flu exhibiting three distinct waves: a mild first wave starting spring 1918, a much more severe second wave beginning fall 1918, and a modest third wave in spring 1919.
- Protection against overwhelming local critical care resources has been the primary focus of every major epidemiological model (Imperial College, Harvard, IHME) and should remain a focus of local communities.
- The severity of symptoms caused by SARS-CoV2 varies widely across different age groups, with the fatality risks (relative to those of a seasonal flu) on healthy young people being considerably less (e.g. by a factor of 0.02 to 0.3) and the effects on older individuals being considerably worse (e.g. by a factor of 20 to 100). See Table 1 (below) from the Imperial College London - Report #9 (the report where they predicted up to 2.2. million COVID-19 fatalities in the U.S.). It should be noted that there is an extremely high degree of comorbidity between COVID-19 fatalities and certain underlying health conditions (e.g. hypertension, obesity, and diabetes); similarly, those underlying health conditions become more prevalent as we grow older. As such, although the Imperial College infection fatality ratio (IFR) inputs (based on data from Wuhan, China) are differentiated by age, that differentiation seems to be primarily a proxy for overall health frailty and the existence of underlying health conditions. In other words, an extremely healthy 80-year-old would likely have a better expected outcome after exposure to SARS-CoV2 than a 60-year-old overweight diabetic with extreme hypertension.
TABLE 1 (from Imperial College Report #9)
Given all the above, and with specific focus on the final bullet point, I assert that the most reliable path to population immunity, i.e. the path that results in 90% fewer overall COVID-19 fatalities, requires a two-pronged approach, namely loosening the social distancing restrictions on those least likely to develop serious complications from exposure to the virus while tightening the guidelines for protecting those who are most vulnerable (coupled with providing robust community support to ease the burden of isolation). Here is an overview of this proposed path:
- Encourage and allow normal levels of social interaction among the least vulnerable members of the population (which I refer to herein as Very Low Risk individuals, or VLRs).
- Encourage trigger-based social distancing among those who are at a low or moderate risk of severe complications from COVID-19 (Low to Moderate Risk individuals, or LMRs) and VLRs whenever they are interacting with non-VLR individuals.
- Explicitly shield the most vulnerable members of the population (High Risk individuals, or HRs, e.g. anyone who is in frail health and/or has an underlying health condition that is known to be comorbid with COVID-19 fatalities). This is especially true for individuals residing in nursing homes and other adult-care facilities. Extreme efforts (and strong community support) are needed to protect the residents of such facilities and those who work there.
- Each local community should establish their own set of ‘trigger-based social distancing’ guidelines. Those guidelines should start out relatively lax (e.g. avoid close contact with non-family members and wash hands frequently), and increase proportionately whenever local confirmed + presumptive SARS-CoV2 cases exceed certain predetermined thresholds (which should be directly tied to local critical care capacity), such as I previously discussed here.
- Individuals should be provided clear information regarding the risk factors, categorized both by age and underlying health conditions, then allowed to self-select into the VLR, LMR, and HR groups (with some obvious restrictions on self-selection, such as healthcare workers and nursing home staff should be excluded from the VLR group). In addition, very young children (IMHO) should NOT be categorized as Very Low Risk, even though very few children have died from exposure to the virus; because very young children are still developing in so many ways, parents should seriously consider shielding them from exposure to viruses that we know very little about, such as SARS-CoV2. It is entirely possible that exposure to SARS-CoV2 could be developmentally disastrous to a very young child’s biological systems (e.g. neurologic, immune, endocrine, respiratory, circulatory, etc.).
The above ‘path to population immunity’ can be summarized by the following visualization:
As mentioned above, population immunity can be expected when at least 70% of the population have developed antibodies to the virus. However, whereas the actual immunity threshold cannot be determined a priori and could be as low as 50%, I have conducted my analyses using both values. Nonetheless, from a risk-management perspective, the worst-case scenario (in this case, the 70% threshold) should garner particular attention. Based on recent U.S. Census data, we could achieve over 70% total population exposure via 100% exposure of everyone under 55 years of age (and zero exposure above 55); similarly, we could achieve 50% total population exposure via 70% exposure of everyone under 55. Although a myriad of combinations exists for achieving any given level of total population exposure, the primary take-away from the analyses presented herein remains the same: skewing exposure towards those who are at Very Low Risk of serious complications and death will result in FAR fewer overall fatalities (see analyses below).
Using the above exposure-by-age criteria, I ran a set of spreadsheet models using infection fatality ratio (IFR) data from three sources:  the Imperial College model’s IFR inputs (derived from fatality rates in Wuhan, China),  Lombardy, Italy, and  New York City. Each spreadsheet model assumes that either 70% or 50% total population exposure would be required before COVID-19 fatalities become negligible, and that Lombardy and New York only achieved 20% population exposure during their respective first waves of the pandemic (this assumption implies that fatalities in Lombardy and New York City will likely reach 3.5 times their current numbers before 70% population immunity would be achieved in each location, or 2.5 times for 50% population immunity). The tables below summarize the results of various iterations of the spreadsheet models.
The charts below show the projected U.S. COVID-19 fatalities and cumulative fatalities by age using the New York City IFR data, comparing 70% exposure evenly-distributed across the U.S. population versus 50% exposure evenly-distributed versus 100% exposure of only those under 55 years of age. To emphasize the importance of explicitly shielding the vulnerable from exposure, the charts also show the impact of a mere 10% exposure of those 70 years and older, which effectively doubles the number of cumulative fatalities relative to the ‘Under 55 Only’ model.
As can be seen both in the above tables and the charts below, targeting population exposure toward healthier members of the population (which I am proxying by age, due to a lack of more granular health-condition data) can reduce the total fatalities associated with COVID-19 by up to 96%.
The overall duration required to reach population immunity represents another extremely important factor with respect to shielding the vulnerable and overall risk management of the disease. Regardless of the measures taken to shield the most vulnerable members of society, there will be ‘leakage’ and the LONGER the duration before we achieve population immunity, the MORE leakage we will undoubtedly experience. In other words, the longer we treat social distancing as a ONE SIZE FITS ALL solution (rather than applying a targeted approach, both geographically and by risk category), the MORE overall fatalities we will experience
Whereas the models I have presented herein are admittedly unsophisticated, I suggest that an extensive sensitivity analysis be performed using the Imperial College London original model (which is now publicly available on GitHub) or something similar, augmented to evaluate the targeted-exposure approach to population immunity (as presented herein), in tandem with a localized trigger-based approach to protecting local critical care resources (as presented conceptually in both the Imperial College and Harvard models and also detailed here).
Steve Trost is Associate Director of the Institute for the Study of Free Enterprise and can be contacted at email@example.com. He has a bachelor’s degree in engineering from MIT, a master’s degree and PhD in engineering from Oklahoma State University and a PhD in entrepreneurship (also from OSU).
Follow Dr. Trost on twitter: @TrostParadox
Follow Dr. Trost on twitter: @TrostParadox
Disclaimer: All comments, observations, and statements presented herein represent the opinions of the author and in no way reflect the views of Oklahoma State University or the Institute for the Study of Free Enterprise.