Updated charts and estimates, 9 April 2020

Yesterday the respected Institute for Health Metrics and Evaluation (IHME) at the University of Washington released an updated model of COVID and various projections.  Their current estimate for total US deaths is down to 60415, down from over 93,000 last week. The reason given for the drop is a combination of better data and the impacts of mitigation measures like travel restrictions and “social distancing.”  This is a long post, the latest data and models are worth taking a close look as there is both good news and bad.

A couple of caveats about this estimate.  First, it is only through August, so additional deaths after that are not included.  Second, it may be overestimating the impact of the mitigation measures.  Finally, it assumes the mitigation measures will remain in place for longer than they are presently scheduled to do so.  My guess is that the next revision will be upwards, but we’ll see.  For what it’s worth, my “in house” statistical model is a bit higher, at 72,820. For perspective, note these total numbers are very similar to the toll from the H3N2 outbreak in 2017 (61000), and less than the Sydney Flu of the late 1990’s (114,526).  Of course, as previously noted, while this virus may end up with a similar death toll to influenza, the rate of deaths are five or six times faster.

Here are the latest charts normalized to population so we are comparing like numbers.  First the country level data.  Italy and Spain are slowly starting to converge, but continue to trend higher than the forecast curves.  Other countries like Germany and the Netherlands are doing better.  The UK and France are on a similar trajectory – the France data is “jumpy” due to reporting issues … as usual, feel free to ignore Iran and China, those numbers are likely problematic (scientific way of saying “bogus”).  South Korea shows what aggressive early action can do for an outbreak like this.  The US is back at Day 30.  As usual, click any chart to embiggen.

And here is a closer look at some US states, with Chatham County Georgia also noted:

Chatham County GA is home to Enki, and also worth a look as a snapshot of what is going on outside the big cities in the US.  There are two markers for Chatham – the actual reported numbers (circle with an X through it) and an estimated number (asterisk).   Why two?  It is pretty clear that Chatham County is lagging in testing and reporting.  I strongly suspect this is a statewide problem.  It is taking up to two weeks to get test results, and it is pretty clear from conversations with practitioners at the local hospitals that many cases – and deaths – are not being recorded in a timely fashion.  LabCorp in particular has come in for heavy criticism for accepting specimens for testing but ultimately being unable to process them.

Lets take a closer look at some of these numbers.   Using the IHME data, the final death toll in New York will be on the order of 6.65 per 10,000.  That is astonishing, and far worse than many other areas of the world, even Italy and Spain, who are converging towards a rate of around 4 per 10,000.  Why is this?  Several reasons, including the population density, late action, recording issues, and sub-par health care system for an allegedly developed country.   There are those who will not like that characterization, but given the advance warning, why should New York be so high? Note that is not a criticism of the doctors and nurses in New York – they are working themselves literally to death trying to deal with this crisis.  It is an indictment of the politicians and leaders who created a system where practitioners don’t have the resources or flexibility do to their jobs.  Nationally, the IHME model is showing  1.79 per 10,000, while the Enki statistical model is a bit higher at 2.2 per 10,000. IHME is showing the peak death rate in the US in just a few days, on the 12th (driven largely by New York), whereas the Enki stat model is later, on the 16th. This could be a problem, given the New York centered views of the major news networks.  The peak for the rest of the country will be one to two weeks later.

On recording issues, it is worth noting that COVID deaths may be overestimated i New York.  If someone is positive for the SARS-COV-2 virus, it is often being listed as the primary cause of death, even if there are other significant factors like end stage congestive heart failure.  On the other hand, in other areas, testing limitations means some deaths where COVID may be an important factor are likely being missed.  So for now I’m treating the numbers as being what they are for the most part, since these two biases may be cancelling each other out.  This is something that will have to be sorted out later.

For Georgia, IHME is showing 2.63 per 10,000, while Enki is at 2.8.  Both are showing the peak death rate (and resource utilization) in Chatham County being projected to be the week of the 21st.  If those numbers and rates hold up, Chatham County can expect between 65 and 70 deaths from COVID-19 over the next month.  That may not sound like a lot, but that translates into potentially stretching the local hospital resources to the limit with hundreds of admissions. COVID patients are resource intensive to treat, and highly contagious. One slip can expose a practitioner to the virus.  California, which has done a relatively good job responding to the crisis, reports that 10% of their cases are health care workers.  About half of these exposures are from an unknown source.  Other countries are reporting much higher rates of cases among practitioners, which reinforces the vital need for proper infection control protocols.

I’ve been informally collection impressions from local health care workers on their situation.  From that anecdotal evidence, this is a huge area of concern.  The local hospitals are already rationing protective equipment, and there are reports they are critically low on some supplies. There is also concerns about staffing at one local hospital, with the nursing staff being reduced (the hospital was doing something smart, trying to clear out patients and get ready, but given the smaller population reducing staffing, which is the wrong move given COVID patients require more time per patient!). But in other areas they seem prepared, and have a plan. The other hospital network is more problematic.  I won’t go in to details, as the evidence is anecdotal, but I’m worried about them.  Hopefully they will work it out; we’ll see how everyone performs over the next two weeks.

For those who are impatient with the mitigation measures, and the profound economic impacts, please be aware that the data shows they are making a difference. Locally, we are just entering the steep part of the curve, and our local health care practitioners – especially nurses, who are the front lines of health care – are in for a rough ride.  Give them whatever support you can – and the best thing you can do is not get sick.

6 thoughts on “Updated charts and estimates, 9 April 2020

  1. Thanks for the information in this blog.

    New York 6.65 per 10,000.

    What about Lombardy a region of Italy with a population of 10 million. As of 8th April 2020 deaths at 9,722. So Lomabardy just about at 10 per 10,000 and although past peak, death toll will continue to tick up.

    Some regions appear very badly affected with hopefully other regions less so, so bringing down average country numbers.

    Does this happen in the same way for regular flu? Or are flu causualties more evenly distributed?

    • Yes, outbreaks tend to be concentrated in local hot spots like that, especially when mitigation measures are taken. New York may end up between a third to half of the US total, if the IHME models estimates are correct.

  2. Do you have any info as to why Germany is seemingly having more success treating patients and reducing deaths than other countries? I was trying to read one article referencing another but that article was in German.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.