7,700

That’s the average number of people who die every single day in the U.S., which BTW is  twice the number we lost in the 9/11 attacks!  (Source: 2.8m/yr divided by 365 days)

The key question nobody is asking is how much has COVID-19 changed that number?

When I analyze a business project, I start with two pieces of data:

  1. The net impact it had on the business
  2. The impact it had on the specific part of the business it was meant to impact. We call this this the categorical impact (because it measures how much impact the project had on specific categories).

It is common to find that a project had a noticeable impact on #2, but not #1.

Let me illustrate with an example.

Let’s say that a potato chip company (I’ve never worked for one) wants to know if the new variety of sea salt potato chips it introduced was a success.

The sea salt potato chip product manager is apt to focus on #2, the sales of her new sea salt chips. Those are doing great! This initiative has been a huge success! Promote me!

I’m often the spoiler of this good news and I come in and show #1, the net impact it had on the business.

About 60% of the time #1 is close to zero. In other words, for every bag of sea salt chips a customer bought, they bought one less bag of the existing varieties of chips. All of her fellow product managers say, Boo! Your new product is just stealing sales from our products.

In the COVID-19, so far, numbers reported have been focused solely on #2, the categorical impact: How many people who die who also happen to have COVID-19?

The more telling number will be how much COVID-19 has impacted #1, the total average number of deaths in the U.S. I suspect that it may be a noticeable impact for the peak time periods, but may be less so as we expand the time window to a few months.

But, as always — I COULD BE WRONG!

That will tell us to what extent COVID-19 had an incremental impact on death rate and to what extent it shifted deaths from one category (e.g. pneumonia) to another (COVID-19) and maybe shifted some deaths by a few weeks or a few months.

Unfortunately, we won’t likely know this for some time. But, it’s good to keep in mind that we should look at it.

High-level estimate

According this site, the U.S. has conducted 1.3 million COVID-19 tests (as of 4/2) and found 240 thousand positive cases of COVID-19, for a positive test rate of 18.5%.

If 18.5% represents the average rate of COVID-19 in the general population, then I would guesstimate that we should expect 1,424 deaths per day (7,700 x 18.5%) of people who test positive for COVID-19 in the U.S.  And, that’s doesn’t necessarily say COVID caused the death. It’s just that they died and happened to test positive, just like the 18.5% of the other folks that didn’t die and also tested positive.

I think 18.5% is too high for the general population, because of selection biases in who has been given the test.

If I cut that estimate in half and say 9% of the population in the U.S. currently has COVID-19, then my guesstimate changes to 693 deaths per day (7,700 x 9%) as people who test positive with COVID-19.

 

Exploring my house

As I rediscover parts of my home that I’ve long neglected, I was going through some old books in my library.

I opened Notes on Love and Courage by Hugh Prather to a page that says:

The character of a church, business or a government can be seen in its attitude toward its detractors.

True.

I would extend that to people, too. Especially so for people on Twitter.

I try to follow The Golden Rule: treat others as you would wish to be treated.

How I read the numbers

I saw a FB friend ask why Missouri has 3x more C-19 cases than Kansas and insinuate the difference is in how seriously the Governors of each state is handling it (Kansas more authoritarian, Missouri leaving it up to locals).

Here are some things to think about:

According to Politico’s Covid Tracker, here are the stats as of 3/29, plus the population of each state:

  • KS: 4,513 tests given, 319 positive, 1.8 million state population
  • MO: 12,385 tests given, 838 positive, 6.1 million state population

This means:

  • Missouri has given about 3x the number of tests as Kansas (2.74 to be exact).
  • It has about 3x as many cases (2.63 to be exact).
  • It also has about 3x the population as KS (3.4 to be exact).

So, all the numbers are in close proportion to the states’ populations. It’s not clear that the way the two governor’s have approached the situation have had much of an impact.

Also, both states are running about 6-7% positive rate on tests given. More on that in a later.

As I’ve said in other posts, until there is no bottleneck or selection bias in getting tested, the number of positive cases will be more of a function of the number of tests given and not a good indicator of the actual number of C-19 cases.

The actual number of C-19 cases, I suspect, is higher. Maybe 3 – 10x higher than cases identified through testing.

I know people in both states that have had C-19 symptoms who were not tested because they were not in a high risk group. They were told to go home and self-quarantine, and their conditions improved. They won’t know until they get an antibody test if they had C-19 or not.

But, what is interesting to me is the % positive tests. At 6-7% it could mean a few things:

  • That has come down from 10-15% range in both states just a few days ago. I’ve found with this data, though, that active cases and/or tests sometimes lag each other by a day. Tests given may have been updated and active cases won’t be until tomorrow. So, the 6-7% may not be accurate.
  • If the numbers are accurate and percent positives have truly declined, that tells me that the testing bottlenecks and selection bias are likely decreasing and more folks are getting tested. If so, the 6-7% may be closer to the percent of the population that has had C-19 since the tests started.
    • I don’t know if that’s good or bad.
    • It could be bad if this actually represents the first wave. That means we still have more than 90% of the population to go through.
    • It might be good if these actually represent the 2nd wave (which I still think it might), which might mean that 15-20% of the population has had it and we have less to go.
    •  I still don’t have a good feel for what percent of the population could get it. So, my inherent assumption in the previous two bullets that the virus will work through 100% of the population may be high.

Shark Attack Journalism

Subtitle: The Dangers of Reporting Numbers Without Proper Benchmarks

A few years ago shark attacks seemed to be on the rise given all the media attention they were getting.

It caused people to ask, Have sharks suddenly developed a taste for humans? Has global warming or agricultural runoff diminished their food supply, so they are more willing to try new things?

But, at the end of the year, we learned that number of shark attacks that year was average or below average.

That doesn’t diminish from the shark attacks themselves. I don’t want to be attacked by a shark and I think it’s terrible when someone does get attacked by one.

But, even in the years where shark attacks got lots of media attention, I still got in the water. Why? Because I did my research and felt comfortable that while getting attacked by a shark was possible, it wasn’t any more probable than it had been before. The only probability that had changed is that of me making the national headlines if I did get attacked.

This caused me to be skeptical when I see the reporting of raw numbers without proper benchmarks. It tells me the author is either ignorant of the importance of proper benchmarks or is intentionally leaving those benchmarks out for a ‘good story.’

If ignorant, they should not be reporting on things with numbers involved and I’m not interested in reading their work.

If intentional, then they are likely trying ‘not to let the facts get in the way of a good story.’

Unfortunately, with COVID-19, a lot of reporting has been done without proper benchmarks.

Here’s a good example of an article that lacks proper benchmarks. Here’s a paragraph from the article:

On Thursday, EMS received more than 6,000 calls. EMS lieutenant Vincent Variale, who also heads a supervisors’ union, said, “We’ve broken every call volume record we’ve ever seen before.”

6,000 is a raw number. That’s the number the author wants me to remember.

But, what’s the benchmark? A benchmark tells me how to think of this number.

The author hints at a benchmark with the quote that it breaks every call volume record.

But, as a reader I’m left to guess by how much. When I’m left to guess, I get skeptical. Too many times before when I’ve been left to guess, like with shark attacks, I’ve learned that the benchmarked numbers were less than alarming.

If a typical Thursday gets 1,000 calls and the previous record was 3,500 calls, then 6,000 might be alarming.

But, if the typical Thursday is 4,800 calls and the previous record was 5,900, then we are talking about an outlier, for sure, but still not too far out of the realm of possibility.

Does the paragraph that follows the one quoted above provide a clue to what might be happening?

Fire Commissioner Daniel Nigro on Friday asked the public to avoid making 911 calls except in real emergencies.

It’s hard to tell.

The way the author includes it, it could just be a public service reminder, or it could be that a good percentage of the 6,000 calls were non-emergency calls driven by the attention C-19 has been getting. For example, calls from people with a mild cough asking where they can get tested and how to get there without possibly infecting 10 others on the way.

If it turns out that the number of non-emergency calls was 2,000 for Thursday compared to 200 on a typical day, then the real story is the number of non-emergency calls, not necessarily the record number of calls.

 

 

 

 

 

 

What I currently believe about C-19

Updated (as noted) 3/30

I thought I would write out the things I believe of about C-19 at the moment and update as I learn more or change my mind. I plan to document what I change my mind on and why.

Here are my set of beliefs about C-19 that I feel reasonably confident (i.e. I have not seen any evidence to dissuade me):

  • For most people, it’s an illness similar to cold or flu with high recovery rates.
    • We now have good enough data to determine who is most at risk — those over 60 with other health issues.
    • >98% of people under age 50-60
    • 90-95% of over age 60
    • Some groups with specific underlying health issues maybe in the 70-80% range.
  • It’s probably more widespread than we think.
    • My guess is that 10-20% of the U.S. has already had it and recovered from it.
    • Another 10-20% currently has it.
      • Why? It seems when testing is based on symptoms, like in areas where test kits are scarce, the positive rate is in the 20-40% range.
      • When testing becomes more widespread, more like random testing, the positive rate drops to 10-20% range.
    • My guess is that the total number of cases are under estimated by a factor of 3 to 10. This is based on testing bottlenecks and the number of people who recovered before testing was available.
      • This would mean that the death rates are likely 1/3rd to 1/10th what we currently believe. Though, there may also be also missing some deaths in the count.
  • The growth curves of number of cases and number of deaths is more of a function of growth in # of tests administered per day than with spread of the virus.
    • The number of new cases per day will grow as long as number of tests given per day grows.
    • The number of new cases per day will level off when the number of tests given per day levels off.
  • A high percentage of folks who end up needing to be put on a ventilator don’t make it. I’ve heard 90-95%.
    • What I don’t know is how this compares to other illnesses like the flu or pneumonia for people with the same risk factors.
  • Masks, social distancing, targeted (to the folks with symptoms or tested positive) quarantines and hand washing seemed  be better ways to slow the spread than broad stay-at-home orders.
  • Government and government agencies have not demonstrated that they understand the concepts of wide testing, contact tracing and targeted isolation.
  • There were C-19 deaths before we started testing for it, but were thought to be caused by flu or pneumonia or natural causes.
  • I have little faith in the reporting of numbers from government or media sources.
    • It’s not that the numbers are wrong, per se.
    • It’s that they report numbers without benchmarks or context.
    • They report numbers in the most alarming, attention-getting way.
    • They appear unaware or ignorant to caveats in the numbers, especially caveats that might remove some of the alarm.
  • There may be good things to learn in Italy about how not to run hospitals. My suspicion is that their practices brought C-19 into the hospitals from low risk groups and spread it through to the high risk groups that may have already been hospitalized.
  • Update 3/30: I saw today that in some parts of Italy, deaths from C-19 may be 2-3x higher than recorded C-19. That’s not surprising. I don’t expect that all folks with C-19 got tested before they died. So, that’s early evidence that the numerator in the death rate calculation is underestimated. We also need antibody testing to see to what extent the denominator — total number of cases — is underestimated before we will have a truer sense of C-19’s death rate.

What I don’t know or I am less certain about.

  • What is the true typical incubation period? Early it was 14 days. Then that seemed to be revised to 5-7 days, but up to 14 days.
  • There seems to be a promising treatment, but it’s been quiet for a few days.
  • Are there longer term health consequences of having the virus?
  • Some people may not get it due to blood type, genetics, hygiene, contact practices, etc.

I will continue to edit this post and keep of log of my beliefs as new evidence comes it.

I’m no expert. But, I’ve been very unimpressed with the experts on this one, so far. I think they have demonstrated some elementary lapses of logic.

I do believe it is better to be safe than sorry (i.e. precautionary principle).

But, I also think some ways of being safe are superior to others and, in the U.S., we’ve quickly chosen some less good ways and seem to have become married to them.

For example…given what we have learned about who is at risk, it would make more sense to me to target stay-at-home orders to those folks and we all pitch in to helping them get through this in isolation.

We can do this by delivering groceries to them, talking to them on Facetime, getting masks to them, making sure they have enough hand sanitizer and soap and making sure they can get treated at the earliest signs of the illness.

Here’s some resources I have found useful:

Politico U.S. Covid Tracker — I like this because it has a benchmark: the number of tests given. I explain in my third bullet point above why I think this is a good benchmark.

 

 

 

U.S. is #1

It has been widely reported these past two days that the U.S. has taken over the top spot in number of confirmed coronavirus cases.

If a news organizations wants to gain my trust, they would also report where we stand on number of tests given. If a good portion of the population has already been infected, then I would expect the number of confirmed cases to be more of a function of how many tests you conduct than how much it has spread.

If we are behind other countries on number of tests given, this news might be more alarming, because that could indicate that it has spread further in our population than in other countries. That could mean a number of things.

For example, it could mean we did less effective things to contain it at the start.

For example, wearing masks and targeting quarantines to those with the virus, and who they came in contact with, may be more effective than a broad brush city-wide shutdown where the unknowing infected still roam around at their essential jobs and grocery stores, along with everyone else.

According the Politico US tracker, as of today the U.S. has conducted 686,000 tests, which has ramped up from under 200,000 on 3/20 and at that time was behind other countries like S. Korea and Italy.

I Googled the number of tests given by country and the most recent data for other countries is 7-10 days old.

The latest data I could find for S. Korea is from 3/20 with 316,000 tests given and Italy, also 3/20, had completed 207,000 tests.

My guess is that neither has ramped up testing as much as the U.S., but let’s say they each have been conducting 30,000 tests per day. That would put S. Korea at 526,000 tests and Italy at 417,000 and that would mean the U.S. has surpassed them in both confirmed cases and number of tests given by a fair margin.

If anyone can point me to more recent data on number of tests given, I’d like to take a look at it.

 

I’m interested in the results of COVID-19 antibody testing

This piece by someone who seems to more about this stuff than I in the Spectator US raises a question I have had about COVID-19 from the start.

How deadly is it, truly?

When we are biased to testing those most likely to have the virus, we may be under estimating the denominator (total number of cases) in the death rate calc.

The author cites Britain’s chief scientific officer as estimating that we might under estimating the total number of cases by a factor of 10.

I also had a hunch that may be a possibility. My hunch was based on how spread out and isolated the cases were in the U.S. in the early days. I felt those cases represented the tip of the iceberg, with a large part of the iceberg submerged as carriers who would never be tested and counted in the total number of cases because their symptoms had come and gone without ever being severe enough to warrant being tested.

I even knew several people in the last 6 weeks who had a flu-like illness and tested negative for flu and strep and was never tested for C-19.

We will get some insight to this when we start doing antibody testing. So far, myself, the author and Gov. Cuomo seem to be the only ones talking about that.

The author also points out that we may be over counting the number of deaths:

Now look at what has happened since the emergence of COVID-19. The list of notifiable diseases has been updated. This list — as well as containing smallpox (which has been extinct for many years) and conditions such as anthrax, brucellosis, plague and rabies (which most UK doctors will never see in their entire careers) — has now been amended to include COVID-19. But not flu. That means every positive test for COVID-19 must be notified, in a way that it just would not be for flu or most other infections.

In the current climate, anyone with a positive test for COVID-19 will certainly be known to clinical staff looking after them: if any of these patients dies, staff will have to record the COVID-19 designation on the death certificate — contrary to usual practice for most infections of this kind. There is a big difference between COVID-19 causing death, and COVID-19 being found in someone who died of other causes. Making COVID-19 notifiable might give the appearance of it causing increasing numbers of deaths, whether this is true or not. It might appear far more of a killer than flu, simply because of the way deaths are recorded.

I’m not trying to make light. Early in on something like this, it’s better to be safe than sorry.

Since this is a new virus that we haven’t built a previous immunity to, then it could infect a large part of the population. Even a tiny death rate on a large part of the population is a sizable raw number.

Though, I am concerned with how few people seem interested in even mentioning these caveat, even experts who should be.

In my opinion, when someone who should know better leaves out caveats like these, they are being intentionally misleading and I lose trust.