Pandemic: Addressing the Supply Side

It must be obvious by now what the Coronavirus (formally COVID-19) is not!

• It is not a hoax!
• It is not slow moving! As of March 17th, according to the World Health Organization (WHO) the number of cases reported globally is now stand at 179,112 with 7426 deaths. This represents a 24 hour increase of 11,526 cases and 475 deaths, a jump of almost 7% in a single day.
• It is not anything like the NOW well-understood flu! While we are all now familiar with the flu with well mapped out mechanisms, well identified seasonality, a vaccine regimen, well known symptoms, common sense precautions, the typical outcomes (mild for most people, severe for high risk populations) and the related complications. The flu infection rate is high among the unvaccinated. There are typically 3-5 million severe cases globally (approximately 0.1%), leading to 300,000-650,000 deaths (approximately 0.01% mortality ~ that is, 10% of the severe cases result in deaths). Those who cannot remember the past are doomed to repeat it, said George Santayana. Because, when influenza first swept through the globe as a pandemic in 1918 (labeled as the Spanish Flu), it became the deadliest epidemic ever experienced by humans. With the world's population hovering around 2 billion then, the flu raged around the globe infecting 500 million (25% of the world's population) and causing anywhere from 17-50 million deaths (4-10% of the infected).

The spread of the disease is exponential. But the capacity to respond to the disease is arithmetically constrained (the insight of Malthus). It is clear that we have to take measures not only to arrest the spread of the virus (e.g. hygiene, social distancing) but also to beef up the capacity (caregivers and care equipment). This article provides a view on how we can use technology to help bridge this gap.

"The power of population is so superior to the power of the earth to produce subsistence for man that premature death must in some shape or other visit the human race.

The vices of mankind are active and able ministers of depopulation. They are the precursors in the great army of destruction, and often finish the dreadful work themselves.

But should they fail in this war of extermination, sickly seasons, epidemics, pestilence, and plague advance in terrific array, and sweep off their thousands and tens of thousands. 

Should success be still incomplete, gigantic inevitable famine stalks in the rear, and with one mighty blow levels the population with the food of the world."

From the 'Essay on the Principle of Population,' by Thomas Malthus, 1798

As a side note: it was this grim picture obtained from reading Malthus' work that inspired naturalists Charles Darwin and Alfred Russell Wallace to piece together the theory of natural selection. The geometric population growth and the arithmetic nature of food supply meant that the endless struggle to survive provided the balancing mechanism in nature. Nature "red in tooth and claw," as Tennyson framed it in his poem.

A Simple Yet Powerful Model to Understand the Pandemic

The pioneering work of Kermack and McKendrick in the 1920s, building on the earlier insights of William Hamer, Ronald Ross, Hilda Hudson and others, resulted in the basic compartmental models of epidemiology. Kermack and McKendrick created the SIR model for a fixed population, with S, I, R standing for populations compartmentalized into Susceptible-Infectious-Removed. Removed refers to either Recovered or Dead. R below primarily refers to Recovered shown along with the mortality rate. There are many variants of these compartmental models, essentially modifications of these basic stages. The rates of contact and progression of a population through these compartments over time are predicted using a set of differential equations. This is a deterministic yet intuitive and powerful approach of modeling the spread of a disease. There are also other methods and stochastic approaches.

The reason for dragging the SIR model into the discussion here is to provide a conceptually simple way to frame such discussions and comparisons. The compartmental models provide a way to think of the Coronavirus pandemic in terms of compartments: What is the susceptibility? What is the infection rate? What is the recovery/mortality rate?

For example, the 1918 influenza had the following outcome:

S (the entire global pop: 2 B??) > I (25%: 500M) > R (90-96% --> 17M-50M dead)

So what pattern can we make out so far for the coronavirus from the WHO COVID-19 data?

• Estimates of infection and mortality data: 179K cases and 7.5K deaths are noted globally (as of March 17th). We can observe an average mortality rate M of 4.4% (which the medical profession terms the Crude Fatality Rate or CFR). The infection and death toll in Italy (population: 60 million) is very high, nearly 28,000 cases and 2500 dead, almost a 9% mortality rate and 13% single day jump in infections. The infection rate in Italy is over 450+ per million of population. China (pop: 1.4 billion), where the earliest expression of the coronavirus was seen, saw 81,000 cases (~60 per million of population) and 3200 deaths (4% mortality) before the situation was brought under reasonable control (40 new cases now reported daily). The Hubei province (Wuhan is the capital) has roughly the same population as Italy (~60M). As the epicenter of the infection in China, if we assume that a bulk of the infections were in the Hubei province, the Hubei numbers then parallel or exceed Italy's. We see that the infection rate can reach 450+ cases per million and the mortality rate from 4-9% roughly. Again, to really emphasize this point, these are estimates from what is known currently and true rates are unknown.
• Estimating susceptibility: To move away from the complete unknown, it becomes important to discern patterns from what already happened. Toward the end of February, the WHO-China Joint Mission on COVID-19 provided the world some indications of how the disease was affecting the population by looking at the following: the virus, the outbreak, transmission dynamics, disease progression and severity, the China response and knowledge gaps. They were careful to assess susceptibility as follows: "As COVID-19 is a newly identified pathogen, there is no known pre-existing immunity in humans. Based on the epidemiologic characteristics observed so far in China, everyone is assumed to be susceptible, although there may be risk factors increasing susceptibility to infection. This requires further study, as well as to know whether there is neutralizing immunity after infection." Whether the world really heeded the findings soon enough and took action will be analyzed for decades to come. Some factors stood out: the age, sex/gender and existing underlying conditions of patients. This is just build up for what I really want to emphasize, so I will just provide the highlights and the link to the original report is provided. The younger (< 50 years of age) seemed to fare really well, women did better than men (the crude fatality rate was 2.8% vs. 4.7%, almost double for men), and underlying risk conditions (comorbidity) really heightened the risk of dying (for example, a CFR of 13% for someone with cardiovascular disease vs. 1.4% for patients exhibiting no comorbid conditions). So a simple back-of-the-envelope model for susceptibility could model age distribution classes (a two-class model could be [less than 50, 50 and above], a more sophisticated model could be more granular in classification, with K classes defined using the data in the WHO report); using 2:1 for relative risk between men:women for CFR; and 10X for comorbidity indicators. Just for contrast, the mean age of the dead in Italy was 80, the relative risk of men: women was 4:1 and underlying health conditions again mattered a lot (Source: Lancet paper: "COVID-19 and Italy: what next?")―which shows that the factors understood from China were still pretty relevant. While it may be obvious, the "neutralizing immunity after infection" part referenced above is the potential for reinfection. This is a crucial statement regarding susceptibility from the report: "there is no known pre-existing immunity in humans," meaning entire populations are susceptible. The 1918 influenza swept through entire populations. A crucial bit of information tucked away in the report is something referred to as R0, given as between 2-2.5 (note: the latest estimates now give a range for R0 as 1.4-3.9). In infectious disease modeling, R0 is the basic reproduction number, which is the average number of new infections caused by an individual mixing with a susceptible population, thus acting as a measure of contagiousness. This is related to the concept of Herd Immunity, a measure of how much a population is protected from an epidemic either through immunization or through the process of infection. This is the rationale for vaccination: the greater the proportion that is immune, the lower is the probability of infection for those not immune.

R0 and the Herd Immunity Threshold (HIT):

If the population N has a proportion L susceptible and M immune to an infection, we have N = L + M.  

If p = M/N (immune) and q = L/N (susceptible), p + q = 1

Now, R0*L = R0*(N-M) = N --> R0*(1-p) = 1 --> p = 1 - (1/R0)

Roughly, if we take R0 = 2.5 as an average for COVID-19, p = 0.60 (60%)

p = 0.60 represents the critical proportion of population needed to become immune, the Herd Immunity Threshold (HIT), to damp out the epidemic

Note 1: Current estimates for R0 are in the range 1.4-3.9 --> HIT = 29%-74%

Note 2: There are no currently known vaccines and no known pre-existing immunity for this novel coronavirus, which means that the entire population is susceptible, as far as we know. That is: L/N = 1/R0 --> 40% for R0 ~ 2.5; 71% for R0 ~ 1.4 

Given the population of California is 40M, the Governor of California can reasonably expect 16M - 28M (40%-71%) to become infected, depending on whether R0 ~ 2.5 or 1.4. Health experts determine what that proportion can be.

Note 3: We can appreciate why "social distancing" works --> because it directly impacts R0.

It was Edward Jenner, the English physician, who taking the tales of dairymaids who contracted the mild cowpox that they were spared the ravages of the deadly smallpox seriously, launched a series of experiments in 1796 to test his hypothesis and create the first vaccine (for smallpox).  He published his methods and insights in 1797-1798. So cowpox ("vaccinia") provides natural immunity to smallpox, which gives us the word vaccination. So vaccination is simply a method of supplying the immunity that humans lack naturally, which otherwise is acquired only as a survivor. Prior to Jenner's work, a risky approach was to expose onself to the disease in a small dose by introducing pox matter taken from a smallpox victim under one's skin using lancets, a practice called variolation (smallpox was called "variola"). Smallpox was finally eradicated globally in 1977 through a worldwide campaign launched by the World Health Organization (WHO). 

(See for example: "Edward Jenner and the History of Vaccination," by Stefan Riedel, BUMC 2005)

• Estimates the Times of Progression of the Disease: The China report provides some indications of how quickly the disease progresses from onset of symptoms to hospitalization and recovery in the non-fatal cases. Here is the relevant paragraph: "Using available preliminary data, the median time from onset to clinical recovery for mild cases is approximately 2 weeks and is 3-6 weeks for patients with severe or critical disease. Preliminary data suggests that the time period from onset to the development of severe disease, including hypoxia, is 1 week. Among patients who have died, the time from symptom onset to outcome ranges from 2-8 weeks." Again the conditions in a particular country make a big difference but we are able to see that even to put together simple models of capacity we need to understand the cycle times. The S --> I progression depends on the transmission rate within a community. "At some point early in the outbreak, some cases generated human-to-human transmission chains that seeded the subsequent community outbreak prior to the implementation of the comprehensive control measures that were rolled out in Wuhan." It became clear from the report that "close proximity and contact among people in these settings [that is, closed settings] and the potential for environmental contamination are important factors, which could amplify transmission." Again, much of the world was still unaware of how dramatically their lives would change in just a matter of weeks as the importance of these factors hit home and social distancing and shelter-in-place became the norm. It is the I --> R flow that indicates a time of 2 weeks or so for the mild cases and 3-6 weeks for the moderate, severe and critical cases.

Little's law, a theorem formulated in 1954 and named after John Little, an MIT Professor of Operations Research, establishes a relationship between average quantity or number in a system Q (e.g. patients, tests), the average rate of throughput R (e.g. patients per day, tests per hour) and the average cycle time T (e.g. days, hours) as: 

Q = R*T

If we have R = 100 patients arriving per day, and the average cycle time required to process them T say is 10 days, then we have Q = 1000 patients being processed at any time, which gives you an idea of the kind of capacity to plan for. 

This is the simple version. This equation can be augmented to consider safety capacity based on the variability of the throughput and the cycle time.  

Little's Law is used in queueing theory to determine response times, wait times, etc.

With this information, we can look across the full cycle S --> I --> R to give us a rough and realistic picture of the pandemic as it evolves:

S (the entire global pop: 6 B??) > I (60-450+ cases per million) > R (91%-96%)

S --> I requires spread control measures (e.g. hygiene, social distancing/density control, etc). The entire population is susceptible (assuming R0 as 2.5, we can estimate 40% of the population to become infected). Relative susceptibility depends on age, sex/gender and underlying conditions. Early testing makes a critical difference in all parts of this cycle. (To be clear, the disease poses a risk to all ages, even if the eventual outcomes for 50+ are different in terms of mortality. Also, the unaffected are still carriers of the virus and can infect compromised populations.)

I --> R requires testing to start with and the cycle times depend on triage categories mild, moderate, severe and critical. Understanding the risk of underlying conditions is crucial in directing treatment. Even crude estimates of capacity and availability can help society put together a more informed response to the crisis. Assumptions of cycle times, say 2 weeks for a mild case and 3-8 weeks for moderate/severe/critical cases, can inform wait times and capacity planning (typical areas of operations research with a multitude of techniques already available).

It must be obvious that country/region specific numbers can be plugged into the above model to determine where they are placed in the lifecycle of impact.

Looking at the Supply Side

Social Choice

On December 8, 1998, Amartya Sen from Trinity College, Cambridge delivered his Nobel lecture titled "The Possibility of Social Choice." He was awarded the Nobel Prize in Economics "for his contributions to welfare economics." His contribution was described by the Nobel Committee as: "Research on fundamental problems in welfare economics. Studies of social choice, welfare measurement, and poverty." As Sen points out in this lecture, social choice theory (developed in its path-breaking modern form by the brilliant economist Kenneth Arrow) is a broad area tackling issues such the fairness and consistency of majority decisions; judging society as a whole on its ability to accommodate disparate interests; how to measure the situation of people who make up a society (for example: poverty); how society accommodates the rights, liberties and preferences of society; how to socially evaluate public goods like the environment and epidemiological security; and how group decisions contribute to famines, gender inequality, etc.

But it was Sen's research on famine in India (the Bengal Famine of 1943) and in other countries (famines in Ethiopia in 1973-74, in Bangladesh in 1974, in the Sahel region of Africa in the 1970s), summarized in his brilliant work "Poverty and Famines: An Essay on Entitlement and Deprivation," which provided the startling picture that a famine can occur not only from total food shortage but also from societal structures of entitlement (rights of ownership and exchange) that deprive people of the legal means to command enough food for sustenance. In fact, what becomes clear is that societal structures may cause famines to happen even with a food supply deemed adequate during other times. As Sen describes, in an economy where private ownership and exchange are possible and trade and production are supported, a flawed societal structure can cause the endowment of a person (what they own) and the exchange potential (what alternatives a person can command) to fail catastrophically, plunging them into a crisis.

Why are social choice considerations important for the pandemic?

• For countries to evaluate if they have taken a public good like epidemiological security seriously. As the economy staggers with both the enormous direct and indirect costs of the impact of the coronavirus pandemic, it becomes abundantly clear that economic models have not considered such factors. (There is potential to apply the stochastic models of ruin theory to form a better picture of pandemic impact.) Both psychological (behavioral) factors and tangible factors (jobs, demand and supply, etc) play a role. As a rough proxy for global economic panic, by the end of February, the Dow Jones Index was within easy reach of 30,000, and today (March 19th) it hovers around 20,000. This translates to trillions in market loss. Even the intervention of the Feds in the US to stabilize the economy by reducing the exchange rate to zero and pumping trillions to reassure the jittery has as little effect. The mis-valuation of epidemiological security has enormous consequences for the nation, as we are finding out all over the world.
• For countries to determine (taking a leaf from Sen's book) if the existing societal structures (rights of ownership and exchange) are resilient enough during a pandemic when the consideration is healthcare. In other words, just as food is a consideration for famine, care becomes the commodity of concern during an epidemic or pandemic. The whole point of Sen's analysis of famine was looking at the failure of legal mechanisms of entitlement and its role in deprivation. Do ownership and exchange structures provide adequate care? In a highly privatized healthcare setup like the one in the US, this is a matter of primary concern! Does availability of insurance entitle one to testing and care? How is sick leave covered during a pandemic? Is a sick person forced to work to avoid job loss during such a serious pandemic? Even the definition of Paid Time Off (PTO) will get really tested during such times. The coronavirus cares little if one is insured or not―how will the uninsured or underinsured be treated? How is emergency medical funding accessed? Can privatized medicine handle the entire lifecycle of emergency coordination with government bodies (like the CDC in the US)? As the US scrambles to address such concerns, it is again abundantly clear that current structures cannot provide adequate support without drastic emergency amendments to various charters. As we have seen, countries with a better social safety net can fare better in terms of such deprivation. It highlights something that welfare economics has always pointed out ―that economic security depends on holistic assessment of a population and on real interpretation of what constitutes welfare. As the US evaluates its response to the pandemic and also how to craft a healthcare policy for all citizens, the factors discussed here should become part of the healthcare fabric.

So, in terms of social choice, we see how that affects the valuation of epidemiological security and the overall supply and functioning of care during a pandemic.

The Role of Technology

Social choice considerations examine the important societal constructs in terms of choices made, rights and liberties protected, steps taken, etc―that is the social machinery that operates to keep the flow going. The Malthusian principle exposed the direct gap in supply that occurs when the pandemic sweeps through the population in exponential fashion and the arithmetic supply of care (people and equipment) cannot keep up (as is the current situation in many countries like Italy, Spain, Iran, etc). This is the prospect that other countries like the US are facing as well.

We can objectively learn from the China response as the nation adopted a science and risk-based approach as the outbreak evolved and they learned more about the previously unknown virus. This is very important as CNN reports now (March 19th) that "China has reported no new locally transmitted coronavirus cases for the first time since the pandemic began, marking a major turning point in the global battle to contain Covid-19."

What they learned from their response:

• Creating the diagnostic framework, mapping the spread, assessing the incubation periods --> time is of the essence in responding to a pandemic and this requires the use of cutting edge tech to support the analysis and support containment strategies (e.g. move to online platforms for information and care, 5G for rural outreach)
• Collaboration and coordination --> health experts, public health officials, community representatives, frontline care providers and equipment suppliers
• Rapidly identifying gaps in the public health emergency response capacity and communicating/alerting those concerned --> information, tools, protection for workers, methods of care, equipment, etc.

On January 19th, a patient who came to the Providence St. Joseph Health's urgent care center in the Washington State with a cough after returning from a trip to Wuhan, China became the first positive case of COVID-19 in the US. Today, the number of coronavirus cases in the US has reached 10,000 already.

Source for the Providence example discussed below is from the excellent article Practical takeaways from America's COVID-19 Ground Zero (Laura Lovett, March 12, 2020)

Providence is leveraging its expertise and infrastructure built up to handle epidemics like SARS and Ebola. ""We had an infrastructure we built we never threw out," said the chief clinical officer Dr. Amy Compton-Phillips. Providence is now in the frontline of the coronavirus battle. The approach they have taken is important and should be a model for health centers across the country. Providence decided to leverage technology to bridge the demand-supply gap and with a strategy for triage, testing and treatment.

For triage, they implemented a chatbot to handle queries and online platform for consultation and a nurse line.

"Getting tested for the virus in the U.S. has been notoriously difficult thus far, however Providence is looking for ways to test a broader swath of patients in a speedier way." Adopting stringent criteria to screen patients for testing has been the difficult way out, highlighting the difficulties that await the US on this front.

For treatment again, Providence is adopting a triage model to treat the bulk of the patients at home. According to Dr. Compton-Phillips: "We worked very closely with our telehealth group, and they were able to create at capacity to give patients seen in the ED a thermometer and a pulse oximeter, and have them monitor at home using our telehealth capacity to be able to say how are you doing, and are you safe to stay at home, and are you going the wrong way? It's one of the things we've seen with this particular germ that patients can be OK for a while then decompensated rapidly."

How Technology Can Help Right Now

We can learn from WHO's strategic objectives as outlined in their situation reports:

STRATEGIC OBJECTIVES

WHO’s strategic objectives for this response are to:

* Interrupt human-to-human transmission including reducing secondary infections among close contacts and health care workers, preventing transmission amplification events, and preventing further international spread (**);

* Identify, isolate and care for patients early, including providing optimized care for infected patients;

* Identify and reduce transmission from the animal source;

* Address crucial unknowns regarding clinical severity, extent of transmission and infection, treatment options, and accelerate the development of diagnostics, therapeutics and vaccines;

* Communicate critical risk and event information to all communities and counter misinformation;

* Minimize social and economic impact through multisectoral partnerships.

(**) This can be achieved through a combination of public health measures, such as rapid identification, diagnosis and management of the cases, identification and follow up of the contacts, infection prevention and control in health care settings, implementation of health measures for travelers, awareness-raising in the population and risk communication.

To paraphrase these needs:

• Flatten the curve (lower the demand and augment the supply)
• Identify the risks
• Address the unknowns
• Communicate critical information (and eliminate misinformation)
• Identify the optimum path for triage, testing and treatment
• Minimize the economic and social impact

The efforts mounted against the pandemic outline a path to augment capacity with technology such as:

• General information management about COVID-19 with access to trusted data sources (e.g. OpenWHO)
• Real time models and analytics to map the spread of the disease (clusters, transmission vectors, compartmental information, etc)
• Real time communications (equipment, networks, tools)
• Real time collaboration technology (for public health officials, community organizers, frontline caregivers, hospital staff, etc.)
• Tele-consultation and tele-health technology (supporting home care, remote care, etc)
• Technology to support remote testing (as creative as we can get)
• Real time technology for the care supply chain (capacity planning, logistics, etc)
• Technology to rapidly assess care gaps (e.g. information, training, people, equipment, protocols, IT, devices, etc.)
• Advanced technology to support precision treatment on-site
• Wearables to augment monitoring capacity
• Robotics to help reduce staff exposure and with isolation care
• And more

In a nutshell, information exchange, real time communications/interactions, and operational support and analytics for the caregiver community. Of course it goes without saying that we can leverage big data, AI, IoT, 5G, blockchain, qubits, robots, drones and whatever else, if that helps this cause. The important thing is that we are able to rapidly multiply the virtual caregiving capabilities to help the systems that are going to get swamped, if they are already not.

"The future rewards those who press on. I don't have time to feel sorry for myself. I don't have time to complain. I'm going to press on." 

-- President Barack Obama

Look forward to your questions, comments and feedback.

―Suresh Babu, March 20, 2020