I started writing this article a few days ago on March 5th, but I didn’t get to finish it until March 16th. It’s still useful to help make sense of the problem.
Living in the Seattle metro area in Washington state, much of my attention has been focused on the coronavirus. With the media, stock markets, and social media going crazy over the topic, I spent a lot of time digging into the data to develop my own conclusions.
Out of personal bias and self-interest, my focus has been on the United States and, in particular, the Seattle area.
I have several conclusions that I’ll share with you and a number of second-order implications that only a handful of people here locally are paying attention to (which I think is a mistake).
- Sample-Size Problem
- Excessive Focus on Fatality Rate + Insufficient Focus on R0 (how many people an infected person infects before they’ve recovered)
- Psychological “Panic” about the Wrong Things (toilet paper)
- Insufficient Focus on Medical System Scalability (equipment, personnel, & system throughput)
1) Sample-Size Problem
As of March 5th, 4 pm ET, the United States had:
Total Cases: 164
Total Deaths: 11
Friends of mine have said, “Why is everyone freaking out over 11 deaths?”
“Only 164 cases, that's not bad compared to other countries."
The problem with this conclusion is the sample size.
Consider two extreme scenarios:
Scenario A: Sample the Entire Population
Total Cases: 164
Individuals Tested: 330 million (entire population of the United States)
Scenario B: Tiny Sample Size
Total Cases: 164
Individuals Tested: 164
In both scenarios, the # of cases is identical: 164 patients.
However, intuitively, the conclusion we can draw is wildly different.
In the first scenario, when we test everyone in the country (and assuming the test is accurate), we have perfect confidence in the exact # of confirmed cases as of the day of the test.
In the second scenario, it is mathematically impossible for the country to have more than 164 confirmed cases because only 164 people were tested.
As of a few days ago, the United States was closer to Scenario B, as only 480 or so people had been tested. The country is moving toward scaling up testing to a much broader segment of the population (which I agree with wholeheartedly).
The sample size (or lack thereof) has enormous implications for the confidence we can have in the numbers.
In trying to solve any problem in any domain, it helps enormously to “size” the problem or determine its severity. If we can’t numerically define the problem, it’s hard to respond appropriately.
It leaves us prone to over- and under-reacting. The best way to respond to a crisis is to respond accurately.
2) Excessive Focus on Fatality Rate + Insufficient Focus on R0
When fatality statistics first came out, it looked like the overall fatality rate was around 2%. If you have been following my work for any length of time, you know that anytime you get summary-level statistics, you need to segment.
As you break down the fatality rate by age and prior medical history, a more nuanced picture emerges. Globally, no child under age 10 had died. The fatality rate for those between 11-39 yrs old was around 0.2% (similar to regular flu which has a case fatality rate of 0.1%).
If you looked at people 70+ yrs old who had chronic cardiac, respiratory, or diabetes conditions, their fatality rates were between 12%-14%.
Many of my acquaintances argued, "What's the big deal for people our age? Our kids won’t die from this. The practical fatality risk is close enough to regular flu that the risk is basically the same."
If the fatality rate were the only relevant issue and one wasn’t in a high-risk group, I can understand that point of view. No big deal.
However, it’s useful to remember that the fatality rate is a percentage.
Fatality Rate % x # of Patients = Total # of Deaths
If 10% of people die x 100 patients = 10 deaths
If 1% of people die x 1 billion patients = 10 million deaths
The key metric that is overlooked by a lot of people is how many patients there will be as the virus progresses.
It turns out that there is an actual metric that measures this. It’s known as R0 (pronounced R-nought).
R0 is the number of people an infected person infects before getting better.
THE R0 IS MORE IMPORTANT THAN THE FATALITY RATE.
If R0 = 2, that means 1 patient infects 2 new patients.
If R0 = 0 that means 1 patient doesn’t infect any other patients.
The R0 is difficult to know until after the pandemic ends. The closest proxy metric that’s easy to measure is # of new cases today vs. yesterday.
Scenario 1: Exponential Growth Phase
# of New Cases Today > # of New Cases Yesterday
This is where things are in the United States and Italy right now (March 15, 2020).
Scenario 2: Inflection Point Phase
# of New Cases Today = # of New Cases Yesterday
This is a transition point before the outbreak slows down. This is an encouraging sign and it typically means that if we keep doing what we’re doing, things should improve.
Scenario 3: Deceleration Phase
# of New Cases Today < # of New Cases Yesterday
This means that the outbreak is slowing down. This is the goal.
The above three scenarios are predicated on sufficiently accurate data. The easiest way to intentionally manipulate or unintentionally skew the numbers is to limit testing. In that case, you have no idea what is happening. Not Good.
3) Psychological “Panic” about the Wrong Things (toilet paper)
Locally, there have been a lot of people buying toilet paper. There are videos of people running into Costco (a warehouse-style retail store that sells large quantities of goods at low prices) circulating locally.
From a purely functional standpoint, the demand for toilet paper is not going to go up from this pandemic. So, logically, there’s no need for extra toilet paper.
However, there is a realistic supply chain risk. If borders close and quarantines are imposed, trucks, truck drivers, and the goods they carry don’t get transported.
In addition, if you need to self-quarantine, you can’t go to a store to replenish your inventory. If the virus is extremely widespread in your community, it does make sense to go to the store to buy larger quantities less frequently.
The problem with psychologically-motivated “panics” is that they create a distraction. Even if you think buying toilet paper is stupid, if everyone else is buying toilet paper, it somewhat forces you to do the same.
While you do have to monitor panic-driven out-of-stock situations, it becomes a distraction from the supply constraints we should absolutely be worried about… specifically, medical system capacity.
4) Insufficient Focus on Medical System Scalability (equipment, personnel, & system throughput)
One of the key leverage points on the fatality rate is access to intensive care medical treatment.
[When I’m not teaching about case interviews and working with clients, I volunteer and train as an emergency medical responder. I’m trained for urban search and rescue, wilderness search and rescue, emergency medical response, basic life support, and mass casualty events.]
During the early weeks of the coronavirus, the case fatality rate was much higher than it is now. The initial death rate was around 15%; it then dropped to around 3% in China and 1%-3% in the rest of the world. The higher initial death rate appears to be due to two factors.
First, encountering a new virus whose contagiousness traits were not understood.
Second, local hospitals in Wuhan were completely overwhelmed with patients.
In short, when the demand for hospitalization > hospital capacity, the death rate goes up quite dramatically.
When I started writing this article, very few people in the United States grasped this point.
I’m encouraged to see that the term “social distancing” has become common vocabulary here in the United States.
The risk isn’t that someone in a low-risk group will die.
The risk is that hospitals will be overwhelmed and have to turn away patients they could otherwise save.
If your hospital has exceeded capacity and you break your leg, guess what? That’s not life-threatening. You may not get treated.
If your hospital is overrun, you get in a car accident, and you need oxygen therapy to survive, there may not be any oxygen to give.
In looking at the medical literature on coronavirus, the current treatment protocol is essentially symptom management. There is no known cure for the virus.
As a result, doctors and nurses basically try to keep you alive long enough for your body to fight off the virus itself.
For coronavirus, the biggest life-threatening symptom is respiratory distress and failure, a.k.a. you can’t breathe.
Treatment has involved using various forms of oxygen therapy. Non-invasive oxygen therapy involves putting a mask on your face with a high level of oxygen. (This is similar in concept to the oxygen masks that fall from the ceiling in an airplane).
Invasive oxygen therapy involves intubation (putting a tube down your throat) and hooking you up to a machine that helps you breathe. This is known as a “ventilator.”
In my analysis, the three most constrained and “at-risk" resources in the medical system are ventilators, regular and intensive care hospital beds, and medical staff.
As of seven years ago, there were only 62,000 ventilators in the United States. The federal government has a reserve supply of an additional 9,000 ventilators. That puts our potential supply around 70,000 ventilators. The thing is, most of the ventilators are already in use.
Data suggests that between 10%-20% of people who get sick need to be hospitalized. About 10% of those hospitalized (1%-2% of all those who get sick) need a ventilator.
Let's do some “estimation” math. (Hey, this stuff is useful in real life not just case interviews.)
Let's assume that 80% of ventilators are in use for regular hospital care. That means we have around 12,000 ventilators in hospitals that aren’t actively being used. If we add in the 10,000 units in the government stockpile, that's 22,000 units available.
If we assume 1% of patients require a ventilator, we can do some algebra to figure out how many patients the US medical system can handle.
(It's also important to realize this huge insight: a patient who needs a ventilator typically requires one for about 3 - 4 weeks in order to fully recover. In my model, I assumed an optimistic 21 days.)
Cumulative # of Patients Last 21 days x % of Patients Who Need Ventilators = # of Ventilators Needed
If we assume the saturation point is where the # of ventilators needed = # of ventilators available, then the equation becomes:
Cumulative # of Patients Last 21 days x % of Patients who Need Ventilators = # of Ventilators Available
Let's plug in some numbers:
Cumulative # of Patients Last 21 days x 0.01 = 22,000
Divide both sides by 0.01.
Cumulative # of Patients Last 21 days = 2,200,000
When you look at the daily growth rate of new cases, we are on a trajectory to hit that number of cases in the middle of April 2020. That's the incredible compounding effective of exponential growth.
[Globally it took 90 days to get the first 100,000 patients. The second 100,000 patients took 11 days.]
** It's important to note that overall # of cases in the United States (and elsewhere around the world) and the ICU saturation point is HIGHLY dependent on choices we as a society and as individuals make. My forecast (and all Covid19 forecasts) is NOT a foregone completely unavoidable conclusion. The true outcome is know fully knowable in advance. The purpose of models isn't to forecast the outcome to the last significant digit. The purpose of a model is to answer the question, "Should we be worried and take action?" And to that question, my firm conclusion is Yes! **
There will be regional differences. Here in the Seattle metro area, we are a hotspot for the outbreak and our hospitals will likely be overwhelmed a few weeks before hospitals in other areas are. (At the same time, we and a few other states notably California and New York, have been aggressive earlier on social distancing. This would be an offsetting factor.)
This is why epidemiologists and infectious disease experts have been extremely concerned.
It’s also why Seattle and Washington state government leaders are extremely concerned.
To understand what the fuss was all about, I built my own model based on assumptions provided by epidemiologists. When I saw the exponential growth curve of my own model, I became absolutely terrified.
I’ve never been terrified in my life before.
Several prominent individuals in the venture capital, startup, and finance communities did the same thing. All have independently come to the exact same conclusion. If we do nothing, the preventable loss of human life will be unimaginable.
There’s a certain loss of life that will occur due to this virus no matter what we do. That is a human tragedy that is not in our control.
However, there’s another quite significant loss of life that’s preventable. These losses are within our control. This is what I’m personally focused on and concerned about at the societal level.
Given clinical trials on antiviral drugs and vaccines wouldn't yield results for another 12-18 months, the only means to combat the virus today is through social distancing.
Personally, I have cut all in-person contact with the outside world. I’ve canceled all in-person appointments. I only go out to the grocery store during off-peak hours and would only see a doctor if facing a life-threatening situation (if it's only moderate or severe, I would call first).
My lifestyle changes:
- No eating in restaurants (which have now been closed by the governor)
- No cafes (which have also since been closed)
- No going to movie theaters
- No going to the barber (yes, I’ve been cutting my own hair)
- No trade shows
- No airplane travel
- No in-person business meetings
- No seeing friends in person
The key insight into why these severe actions are needed is this:
NEWLY INFECTED PATIENTS ARE CONTAGIOUS FOR APPROXIMATELY 5 DAYS ** BEFORE ** THEY FEEL SICK.
Think about that for a moment.
Let's assume that I had contact with and got infected by the virus a week or two ago.
I was originally scheduled to fly to San Jose, CA to attend a trade show. I was supposed to go to the Seattle airport and walk by several hundred people. I was supposed to be on a plane with 150 other passengers.
I was supposed to be meeting 15,000 trade show attendees from around the world.
I usually eat out a few times a week. I usually go to my favorite Mexican restaurant that sees several hundred patrons per day. The Starbucks I go to gets a few hundred people per day. The barber sees over a hundred customers per day.
I could have easily exposed thousands of people to the virus all without knowing it.
To be clear, I have no known exposure to coronavirus. The key adjective here is “known.” That’s the problem with this virus, you don’t know if/when you’re contagious.
The only way to know is through testing, and the testing capacity in the United States is infinitesimally small relative to the need.
As a result, social distancing is the only tool available to combat the virus.
Even though I’m in a low-risk demographic, I’ve taken these extreme measures for several reasons:
- Eliminating Guilt: I have many neighbors and extended family members that are high-risk. I don’t want to inadvertently contribute to anyone dying because I wanted a decaf oat milk vanilla latte.
- Reduce Personal Risk: Although my chance of death is low, there are two specifics risks I’m concerned about. Death isn’t one of them.
- Inability to Access Oxygen Therapy — If I were to get the virus, I would likely have a presentation of COVID-19 that is quite easily survivable with proper medical care. As I mentioned earlier, I think there’s a 95% probability that oxygen therapy, in the Seattle area in particular, will have to be rationed and triaged.
If 10 people need oxygen, and you can only treat 1 patient, how do you decide?
I prefer to consider that scenario as an intellectual exercise and not as a participant.
[As part of my training, I’ve been trained in the use of emergency oxygen (think drownings, heart attacks, asthma/anaphylaxis). I’ve been reading the treatment protocols with keen interest.]
For those admitted to hospitals who subsequently survive, they seem to require 2-4 weeks of oxygen therapy. That is a long time to need oxygen therapy when there’s a surge of new patients coming in the door.
In a shortage situation, doctors are talking about time limits on oxygen therapy. You get oxygen for a certain number of days, if you don’t improve in that time, they remove oxygen support and give it to someone else. In other words, they will let you die if you don’t improve.
I’ve also been reading the individual case histories of patients. I’m surprised at the severity of oxygen therapy needed for low-risk patients. In some cases, it appears high-flow oxygen at 10 liters per minute is needed.
For context, if your heart stops beating and you stop breathing completely, medical professionals would give you 15 liters per minute. It’s shocking to me that one might need 10 liters per minute of 100% oxygen just to be able to lie in a bed.
At that level, one would likely die without oxygen support. Keep in mind, this is for someone in a low-risk category. (I don’t have data on comorbidities; so the data is certainly incomplete on this).
I figure that I can’t control if I get a life-threatening illness that ultimately kills me.
However, I’m going to be super pissed if I have a survivable illness, but end up dying in the hospital parking lot waiting in line — especially if that situation was preventable (which, to large extent, it is).
- Long-Term Lung Damage — Little is known about the long-term ramifications of COVID-19. Some patients in Hong Kong who have fully recovered have had lung tissue scarring with a 20%-30% reduction in oxygen transport capability. There isn’t sufficient longitudinal data on this disease. In short, we don’t know what we don’t know.
Realistically, it will not be possible to avoid the virus forever. Most epidemiological estimates indicate 20%-40% of the world will eventually get COVID-19.
That said, if I have to get it, I’d prefer to get it much later in the process… after the hospital demand vs. capacity crisis has abated… after hospital procedures have been refined… after testing turnaround times have come down… after the 200 clinical trials in Wuhan have reached some useful conclusions… after some anti-viral medications have been proven to be safe and effective… after a vaccine has been developed.
In short, my conclusion after all of this research and analysis is simple:
Update March 20, 2020
Thanks to a few readers who pointed out a math error I made in a previous version of this article. I actually made two errors. The first several people caught (a computation error). The second nobody noticed. The second error (which has since been revised) is that a patient that requires a ventilator typically needs 3 - 4 weeks of it before they recover. This significantly increases the demand for ICU beds and ventilators.