skip to Main Content

COVID19 By The Numbers- for people who like numbers!

by Buzz Hollander MD

***COVID19 disclaimer: I am not an epidemiologist nor an infectious disease specialist. I am a humble family doctor who likes to study numbers. This is for other people who like to study numbers. Take what follows with a grain of salt as many of the professionals would disagree with my conclusions.****

The number to ignore is… “confirmed cases”. 

Many people have raised our awareness that the number of known cases is grossly under-represented due to the lack of testing. Testing varies by country. It varies within a country by region. It varies by timing within a region, as most testing is limited in the early phases of an outbreak as a country “gears up” for the infection, and might be limited again later in an outbreak as a country “gives up” on aggressive testing. Here in Hawaii County we are still in the “gearing up” phase, with perhaps 200 test results reported as of March 28th. That’s 0.1% of the population with a known test result. While an accelerating number of deaths is always alarming, an accelerating number of “confirmed cases” might just be indicating an accelerating capacity to test.

The best number to watch is… deaths. 

This is not just because deaths are a fair surrogate marker for the suffering and loss a country experiences. It is because deaths are the hardest number to get wrong. Imperfect, yes – surely some people died of unrelated bacterial pneumonias or heart failure cases and were mis-ascribed to COVID19 deaths. However, this is our most solid number to track the progression of this disease. It is also a “lagging indicator”, meaning the by the time someone dies from COVID19, weeks have probably passed since they became sick. That means that the death toll better represents the severity of a growing infection 2-3 weeks ago rather than now. This is why it is such a hugely good sign when the death rate starts to drop, and so alarming when it is rising exponentially. That rapid growth is the signal that 2-3 weeks ago cases were exploding, and only the leading edge of the wave is hitting.

The numbers we want to know about our COVID19 test are… missing.

We have a fair sense that the many tests out there for COVID19 are fairly accurate. That’s about it. Every test needs a “gold standard” to evaluate it. If I invent a new stool test to screen for colon cancer, I compare my test results to what is found on colonoscopy, as colonoscopy can detect the vast majority of colon cancers and is considered our “gold standard” for the purpose. To the degree that the gold standard fails to be accurate, my new stool test becomes harder to evaluate. When it comes to COVID19, we have a test, but no gold standard with which to compare it. If someone took the time to follow a couple thousand people tested for COVID19, and observed them for 3 weeks, and made sure a CT was done on all the positives, and evidence of viral shedding was studied on all the negatives, we could have a good sense of how “sensitive” (able to avoid false negatives and miss people WITH the disease) and “specific” (able to avoid false positives and mis-label people WITHOUT the disease) the test is. No one has the time or resources to do this, yet. 

Why does this matter? Well, if we knew the test rarely had false positives, we could aggressively test low risk people without symptoms. If positive, we isolate them and reduce transmission. However, some Chinese researchers have suggested that the false positive rate might be rather high. That’s okay if you only test high-risk people, but if you start testing low-risk (asymptomatic) people your results get wonky. Think of using an accurate test, like tests for banned substances in athletes, and assume it is about 97% specific (ie few false positives). It works pretty well at finding a high rate of true positives in a collection of professional bicyclists who were not anticipating testing. But, if you tested 100 accountants who ride recreationally, you might expect to get 3 positives, all of them false positives. When we test asymptomatic people in households or cruise ships, we don’t know if we are getting mostly true positives or false positives, and it skews how we might view the contagiousness of this virus, and which measures of containment are needed. We also know there is a substantial false negative rate with these tests, often needing two or more tests to find a positive COVID19 case, and this makes triage of sick patients all the harder. My take-away point is this: when you hear that 600 out of 3000 tested people on a cruise ship of 3600  tested positive for COVID19, but only half had symptoms, think of professional bikers and accountants. Did 600 people on the cruise really have COVID19? Or was it closer to 300? We don’t really know until we understand our test accuracy better.

The mortality rate of COVID19 is going to end up being… low.

Not WHO 3.4% high, anyway. The numbers reported from Wuhan initially, and now Italy, are rather shocking, 4+% in the former and up to 8% in the latter . Since the percentage comes from the number of deaths in the numerator (which we trust) divided by number of cases in the denominator (which we don’t), we take the mortality rates with a grain of salt. One theory for the variance is that overwhelmed health care systems, like northern Italy and Hubei Province, lead to excess deaths. The very interesting articles from Tomas Pueyo (a fellow non-epidemiologist) that have widely circulated and are worth a read repeat this theory: when we do not “flatten the curve” and hospitals get overwhelmed by a large initial wave of CIOVDFI19 infections, the death rate goes radically higher due to reduced hospital capacity. I agree with most of his conclusions, but I doubt this theory. Data from both Italy and China suggest that about 50% of people admitted to the ICU with severe disease die. A small study from China showed that 80% of people who made it to a ventilator died anyway, and a colleague from the West Coast is suggesting that number might be closer to 90% from his experience. Increasing ICU beds and ventilators would not seem to be the intervention that led to a 4% mortality rate in Hubei and a 0.5% mortality rate in the rest of China, an 8-fold improvement. Fewer people getting sick would be the significant intervention that led to the 8-fold improvement! Why the high mortality rate in those settings early in the outbreak, then? Testing lagged behind deaths in the early going in both places, and the Italians are still not testing in anything like the numbers we see the South Koreans or Germans doing, for instance. The places that test broadly, report mortality rates around or under 1%. This is the number to expect when the dust settles, and more likely lower than higher.

So, when it comes to the mortality rate, the country to follow is… South Korea.

Of large countries with a major epidemic of COVID19, South Korea tested the best.  If you had a cold, you got tested. They might even be the one place where the mortality rate might be falsely low from over-testing and finding false positives! Still, their mortality rate has to be considered the most solid of anywhere, since we trust their denominator (cases) as well as their numerator (deaths), and they are seeing a mortality rate right under 1%. When commentators draw on the data from Iceland, which has almost no cases reported in those over 80 years of age, to make the case that the mortality rate might be lower than that of influenza, maybe even 10 times lower (!), you have to question whether Iceland is the useful place from which to be gathering data, given its stark contrast to other places that have tested thoroughly. Try telling an ICU physician in Milan or Madrid that the mortality rate of COVID19 is less than the flu!

The most useful contagious disease to compare COVID19 to is… the flu.

This comparison is getting widely panned in the media, for good reason. What COVID19 is doing to the hardest hit regions in the world is worse than the worst of flu epidemics in the past century. However, it is still a useful comparison given our lack of data on COVID19, and our abundance of data for influenza. While COVID19 is probably rather more contagious than the flu, and 10 times more lethal, we can project from flu data and make some rapid conclusions about COVID19. We can extrapolate that if we sit back and do nothing and let COVID19 spread like – well, the flu – we can guess that it will cause at least as many cases as the flu, if not many more; and kill 10X as many people that contract it. This is a 20 second, back-of-the-napkin calculation that was available to every world leader by mid-February.  Some world leaders would probably like those 20 seconds back.  

The other relevant aspect of the flu is that the capacity of our ERs and ICUs have been indirectly determined by it. Car accidents, heart attacks, cancer complications – these are fairly steady over the year in large populations. The flu drowns the ER and ICUs for a few weeks or months every year. If a hospital cannot handle the surge of flu cases in your community in a “bad” flu year, it is forced to revamp and increase capacity. It is a fairly safe estimate that if COVID19 starts turning in seriously ill patients at a rate 2-3X what the flu does in a given community, that hospital will be overwhelmed, a la Bergamo. 

A mortality rate of 1% applied globally if COVID19 transmits like the flu would be… catastrophic.

Thinking the mortality rate will be low does not mean I find COVID19 anything but alarming. The math is simple, and has been widely shared. If 10% of people get the flu each year, and the flu is really not as easily spread as COVID19, we might expect that more than 10% of people would get COVID19 in the world’s population. Let’s just use that 10% number because it’s easy. Couple it with the 1% mortality rate. That means if you want to see how many people will die in your country if we take a “business as usual” approach, it would be 0.01 X 0.1, or 0.001 of the population, or 0.1%. 350 million Americans X 0.1% = 350,000 deaths. That’s 7X a typical flu season. Double that number if the transmission is twice as high, a reasonable supposition. Go up to 30%, and now we have a million Americans dying. These are catastrophic numbers unless your biggest cause is controlling population growth.

The most useful to consider from a Public Health Policy perspective is… deaths per million (DPM).

This number does not have much traction in the media, but it is very useful. 3,000 deaths in China with its 1.4B population is remarkable compared to 10,000 deaths in Italy with its 60M population. China is sitting at 0.5 deaths per million inhabitants. Even in Hubei Province, with its population of 60M, the DPM is still only around 50. Flu season in the USA sees a DPM around 150. Reassuring, right?

The reason why COVID19 is causing mayhem in hospitals in some places, and hardly a ripple in others, is… uneven distribution.

The flu tends to spread through the country in a fairly orderly and predictable wave. COVID19 hits more like a cell of tornados, selectively smashing certain areas. Wuhan got smashed; the rest of China did not. Lombardy, Italy got smashed and still is; Rome is not. Seattle, New York and New Orleans are getting smashed; most other cities are not (yet). Italy still only had a DPM in the 80s with a death toll around 5,000- so what was the problem with Italy? The problem is that when the Italian DPM was 80 for the entire country, the DPM in Lombardy was around 300. “Only” double the death toll we see with the flu, and the system was overwhelmed and reports from hospitals in the North sounded straight out of dystopian novels. Their DPM has since risen to around 600. This is why Italy shut down the entire country, and not just the North, once the severity of its problem became apparent. When people talk about 60% of Americans getting COVID19 if we do not take preventative measures, it would lead to a DPM on the order of 6000.  Folks that think that all this “shelter at home” stuff is a huge over-reaction can mull over that number – while they are staying at home.

The rate of infection in the general population now is quite low… probably under 1%.

We hear about the case numbers in cruise ships and nursing homes that get publicized, and it’s easy to think that those people grabbing rolls of toilet paper out of our hands at the Costco are surely infected. Actually, they’re probably not. In this case, the denominator is the population with which you are concerned. Let’s take our country. The denominator is 350 million. The infection rate, which in epidemiological parlance is the “prevalence” – what percentage of people in a population have a given problem – is the number of cases divided by the total population. We know we cannot trust confirmed cases as our numerator. However, we can make educated guesses based on the number of deaths. Since we are trusting our mortality rate of 1%, it takes about 100 cases stewing out in the real world for 2-3 weeks to see one death. Here in Hawaii County, having not had our first death, or even publicly announced confirmed ICU case of COVID19, we can assume that the number of cases as of two or three weeks ago (the amount of time from infection to death) was probably under 100. If we really don’t even have a COVID19 case in the ICU, since people end up in the ICU earlier and more often than people who die of COVID19, we might argue that as of a week ago Hawaii County probably had fewer than 50 cases. 50/200,000 is a tiny prevalence. However, since our “confirmed cases” continue to rise, and those are only the tip of the iceberg, we can trust that number of actual cases to be far higher now – but still well under 1% of the population. That level would require 2,000 actual cases. When populations are fully tested, we get a better sense of the true prevalence. An Infectious Disease specialist friend in Seattle (who does not endorse my conclusions on prevalence!) shared that they tested an entire nursing home where a death was reported: 30 employees and 100 nursing home inhabitants, and 6 positive tests (counting the death). Prevalence:  5%. The infamous Princess Cruise line with 3000 passengers stranded at sea ended up with about 600 positive tests after quarantine was complete, a rate of around 20%. They tested every one of the 3000 citizens in the Lombardy town of Vo: 66 positives in the town that essentially started the Lombardy outbreak with its first death, or a prevalence of 2.2%. (This “experiment” in complete testing, by the way, resulted in a prevalence of zero after isolations were completed). The Kirkland, WA, nursing home: 300 residents and staff, 63 infections, for a prevalence of 21%. In a Dutch hospital system where health care workers were starting to fall ill, out of some 6000 workers, 1353 people with any cold symptoms were tested, and there were 86 cases, or 6.4%. These are some of the worst possible sub-populations – a Lombardy town, the worst of the cruise ships, two Seattle nursing homes, a hospital system in the Netherlands – and the prevalence ranges from 2-21%. Your neighbor probably does not have COVID19. Yet.

The “acceptable” prevalence rate for an entire population is… well under 1%.

South Korea had about 100 deaths after the first rush of their outbreak. Given that they are now on the back end of the bell curve, we would expect that we can use the death rate as a fair proxy for 1% of the total cases that have happened in that country. We would estimate 10,000 cases based on their death rate;  unsurprisingly, given the excellent testing in South Korea, at they point where they had 100 deaths, they had confirmed nearly… 9000 cases. About 52 million people live in South Korea. They kept their whole population prevalence rate closer to 0.01% than 1%! That is how you have only 100 deaths in a country nearly as populous as Hubei province and Italy! Our current “worst case scenario” which has completed its first wave of deaths – Wuhan – might be useful in determining an “acceptable” prevalence, based on a  flu-like impact. With 3000 deaths, we can assume there were about 100X as many cases, or 300,000. Within a population of 60M, that would be a prevalence of 0.5%. If Hubei had its nightmare with a prevalence of 0.5% – followed by a shut-down of all activity that could probably not be imitated in the western world – you can see why estimates of 60% prevalence or more lead to disturbing scenarios.

What we are seeing in the world right now suggests COVID19 deaths from the first wave of the pandemic will be… more like the flu.

Mercifully, death rates have not been anything like the above calculations. Projecting them forward is of great interest. Hubei province and Italy both have around 60 million inhabitants. A typical flu season might see 5,000-20,000 deaths from flu in a population that size (Italy seems to over-report flu-associated deaths, while China vastly under-reports them, but again – we choose to trust our mortality rate of around 0.1% with the flu).  Italy has passed that lower limit of 5000, but we hope will not close in on the upper boundary of flu deaths of 20,000. China has squelched community spread of COVID19 and ended up with perhaps half the deaths in Hubei one might expect from the flu. Even if the US ends up more like Italy than Hubei, and it is beginning to appear that Italy and Spain might be statistical outliers, we can probably multiply the death toll in Italy by six (given we have about 6 times as many citizens as Italy) to estimate the upper end of our death toll. 25,000-100,000 would seem a fair, broad guess. This is our flu season in a bad year. Why “only” as bad as a flu season? Only because we are locking our country down!

The country whose strategy we should not emulate is… the UK.

The British do some things their own way; they speak American with a funny accent, and they add a syllable to “aluminum.” They have also had a rather unique response to COVID19 that makes even ours seem “pro-active” by comparison. The reports from the UK are that “the plan” was to let COVID19 cases build up right to the point that the health care system can handle- and then lock things down! The problem is, relative to the start of their outbreak, they are waiting to lock things down until well after Wuhan got locked down, in terms of reported cases or DPM. This means they can expect a Wuhan-like experience in the weeks to come. And worse. Why did they deliberately wait before starting to shut things down? Their experts informed their policy makers that if they shut things down early, they would only delay the wave of deaths, but not the number of deaths, nor the chaos. I think this is a flawed theory. Acting early and aggressively – i.e., in the style of non-Hubei China, South Korea, or Hong Kong – has been very promising at limiting deaths per million to a fraction of what the late-responders experience. China has virtually eliminated new cases. South Korea has stabilized their death rate at a low level. Are they destined to have a second wave as bad as the first would have been without such restrictive measures? I would argue that they have some distinct advantages from having already faced the enemy and bought some time: more supplies to stock their hospitals with; more test kits; the ability to jump on top of regional outbreaks with testing and isolation techniques; and a seasoned populace who has had a chance to adjust to “work at home” where appropriate. The next flare will not be as bad as the first – megachurches will employ spacing; restaurants will open back up but with a mind towards lower seating densities; people will be smarter about how they greet each other; employees won’t dare come to work with a cough. While things might not be “business as usual” in Seoul come May, it will be a close enough approximation for its inhabitants to probably think: “this hardship was worth avoiding 20,000 deaths – and our economy probably shutting down, anyway.” Can they keep cases suppressed until a vaccine finally arrives in a year or two, maybe even sooner? Or an effective treatment surfaces? It seems plausible enough to say it was worth the risk of shutting down their economy. 

What about the UK? Barring some deus ex machina that British customs or lungs or genetics are somehow adequately different from the Spanish and Italians, the UK could well experience a Lombardy-like situation on a national level. What did they gain for it? An extra week or two of pub crawls and concerts. Their theory – that they would allow “herd immunity” to take effect in the younger, healthier population, blunting future problems with COVID19, was laughably naive if you consider the numbers, albeit crude numbers, we have been discussing. If “herd immunity” implies, as they stated, that 60% or more of UK citizens should get COVID19, this would be a public health catastrophe – and one that would destroy their economy more severely than any proposed austerity measures. Even if they somehow quarantined everyone over 60 (a real challenge given the need for most of the elderly to interact with those younger than themselves for their basic needs) – even if all of their COVID19 cases were younger people – even if their mortality rate was well under 0.5% – any manipulation of the numbers still yields DPM figures that would dwarf Lombardy. Is Lombardy, on the basis of going into a lock-down, due for a second wave that will then be worse than the UK’s experience, with residents burned out on quarantines and unwilling to comply for a second round? That seems to be the bet the UK was originally intent on making, before beginning to back down from this game of high-stakes chicken, and implanting its own, late, “shelter in place” measures. 

That gets us back to the South Korea analogy – can Lombardy, once this passes in a month or two, use the sort of containment they employed in Vo to prevent new clusters to develop, and while away their time in relative normalcy waiting for a vaccine to be developed? We shall see. It is the challenge awaiting the US later this spring. If there is to be no normalcy without incessant severe COVID19 spikes even after the first wave is done, the degree of psychic and economic suffering in the year to come before an effective vaccine arrives is going to make today’s angst in the US seem positively upbeat by comparison.

The Island of Hawaii should go into severe containment measures… yesterday.

If the best time to plant a tree is 10 years ago, our the Chinese proverb, then the best time to start a lockdown is before you have extensive community spread of COVID19. In a situation with precious little data, there is a plethora of data that the sooner you lock down and test extensively, the lower your DPMs end up in the first wave. Every day you wait, is an exponential increase in cases and deaths. Flying people in from Seattle, San Francisco, and Oahu is the best way to seed community spread. Shut it down and test, Hawaii! 

One last number… 500,000.

It’s not what you think. Most people have settled in to this “shelter at home” philosophy, figuring that the economic and societal suffering is worth it if it means a less chaotic first wave of deaths. I think the data speaks pretty clearly, despite all the uncertainty and need to project forward. What our eyes saw happen in Hubei, Lombardy, and Milan, can happen anywhere, if the prevalence is allowed to rise above a very low point for a highly infectious virus. The only way to suppress the prevalence to socially acceptable levels is with severe social distancing measures. The most viable strategy while we wait for a vaccine seems to be: Control the first wave though whatever it takes – and hope that mutations that calm the virus itself, breakthroughs in treatment, or reductions in transmission in “normal life” through lessons learned in the first wave, will lead to a “COVID19 Success Story.”

I have had many conversations in the past weeks with my college friends, most of whom exist not in medicine but in the world of business. Some are questioning whether the downside risk of COVID19 has been overstated, and if it has justified the certain dismantling of the economy we have undertaken for an uncertain reduction in deaths. They raise some fair points, although I disagree in general – the destruction wrought by COVID19 if left to progress mostly unabated is so severe on the level of both human and economic suffering, that I think standing on the sidelines would be a bad gamble. 

But back to that 500,000. That’s the number of deaths in America every year directly attributed to smoking. Every single year – not just one awful pandemic year! Sarcasm alert: If only we had the means to stop these 500,000 deaths, somehow, some way. But, no – outlawing tobacco products would be government straying way too far outside its lane in limiting personal liberties!

The hypocrisy is hard to deny. We are essentially put into house arrest on a massive scale, all in the name of not overburdening our health care system and to prevent a few hundred thousand unnecessary deaths in one year. God forbid we ban tobacco, though!

This tension, between government forcing us to follow their dictates for the common good, versus having the liberty to make our own poor decisions that lead to chaos and mass fatalities, is one with a rich history, especially in our country. When the dust clears from COVID19, yes, we will all use hand sanitizer more; yes, we might be more cautious in our style of greeting each other; no, we might not show up at work with a ripping cough; but will we be more, or less, open to the exchange of liberty for the public good? That will be one of many interesting story lines to follow in the year ahead. But for now – please have that conversation in the privacy of your own home.

Back To Top