I reviewed the often-quoted website “Institute for Health Metrics and Evaluation” today, and unless something dramatically changes in 24 hours, they will fall on the side of overly-optimistic professional publications which create false hope around epidemics. Their estimates are very low, and if you look above in the grey line, diverge from other estimates substantially next week — a large difference between the grey and the orange/blue lines. Many people use their projections. I think this may be a bad thing.
Their latest estimate is that COVID Daily Fatalities will start leveling off next week at around 2,600 to 4,500 (I can’t even imagine they’re publishing around 1,461 on the low end since we’ve already effectively surpassed that). If all goes as projected, we can expect to hear that we’ll have no more than 95,000 total fatalities (high end: 175,000), the peak is in two weeks, and White House announcements will be seen as prudent.
Institute for Health Metrics and Evaluation: Only 3x worse than the Flu, not 10x or 100x
That’s what you see in the chart above. Unfortunately, this is their first revision, published one day after their prior estimate of 80,000 fatalities.
I really don’t think that’s the case, and worry a lot about their reporting. Let me explain why.
There are many ways to model this epidemic in terms of cases and total fatalities. I don’t use case information to model what’s happening, I use fatality data. Why? When most of what you see in the news is about cases? Because frankly it’s very bad data. If you have a question why, happy to explain. I’ve said it many times, and many other people smarter than I am have too.
I’ve tested many models using fatality data, and assumptions and keep adjusting to find the lowest error system — models and parameters which get closer and closer to actual fatality data. I’m certain that, for example, we’ll see total fatalities move this week to over 30,000, not 2,600. It’s not a number to me, it’s living people in my mind, and should be scaring the hell out of everyone. To me the errors in models, when they are too low, are excuses for more people to make poor, risky decisions. I just had a conversation with a friend today who is struggling to convince a republican family member in Louisiana to take this seriously. Risky behavior is a huge problem that continues to be driven by overly-optimistic or frankly terrible messaging from political and non-political sources.
The IMHE model in the above charts, and which has been flagged by writers on KOS as reliable, will diverge from other models over the next few days, which is in the lead graphic. By Monday, there will be a difference between fatalities in reality and IMHE (which will lose a lot of credibility), other models continue to be more accurate (high-fatality, peak early May), or IMHE will adjust their models upwards substantially, more than 30%-50%. You can’t build models which don’t track reality.
Here’s my current projection.
This morning I assessed what’s called a additional forecasting model called “SIR” (Susceptible-Infected-Removed) (or Recovered) to see how good it could be at predicting total mortality. The good news is that using the same assumptions I have for a Logistic Function, it’s slightly more accurate than my prior model. If so, it would mean 30% lower total mortality, which is huge of course. I’m going to continue to look at it for another few days and probably switch.
The SIR model also implies that transmission will effectively halt when 70% of the population has had an infection, which also has dramatic consequences. I can’t project that with my original model.
So today, I’m left feeling slightly more hopeful — I see lower total mortality than I originally calculated. I’m alarmed at reputable organizations continuing to lowball this epidemic by a factor of 10x or 100x, which can drive risky behavior. I’m also alarmed that CEBM keeps raising the US Case Fatality Rate, which has risen an additional 0.2% to 2.55% from a day or so ago. I hope we don’t get back to 3.4% which I used in my original estimates, but if it rises 0.1% a day we’ll be there again next Friday.
Stay Calm. Critical decisions need a clear head.
Stay Sane. Incessant COVID noise from deranged people will derange you.
Stay Home. There’s no safety in numbers, for the time being.
Disclaimer: I’m not a statistician, nor am I an epidemiologist or physician. I do have substantial training in science and mathematics. My assumptions are based on publicly available data, and interpreted without domain-specific expertise to temper my judgements.