Daily Covid-Insights and Charts — March 27
Let’s start with the cases vs. deaths and the total case count charts.
1) The US, Italy, Germany, Spain, the UK, Iran and France all had more new cases yesterday than the day before. The US, Spain, Iran, France and the UK also had their all-time highs. Although it’s good to keep an eye on the absolute numbers of cases, the overall growth trend in new cases is more meaningful. So here’s the 7-day compounded daily growth rate (CDGR) of new cases. If you’re familiar with CAGR, CDGR is the daily version of that. See note at the bottom on how I calculate it.
2) The 7-day Compounded Daily Growth Rate of new cases declined for all countries, including the US. Please do note that the number of new cases is entirely dependent on how many tests are conducted. We know now that most countries have test shortages, and they tighten the criteria to be tested as the virus spreads. There is a good chance that this decline in case growth is misleading. A more tangible metric for how Covid-19 is trending in a country is the number of deaths. The one thing to remember about deaths is that it is a laggard indicator, meaning it’s showing us how the virus was trending a week to two weeks ago. So here is the total deaths and the 7-day death CDGR for each country.
3) Germany, France, Iran and the UK had higher fatalities than the day before, which were also their all-time highs. We saw a modest decline in the US (246 vs. 249).
4) The 7-day death CDGR is declining for all countries listed here, except for the US. It has been steadily increasing for the US since they hit 150 cases. It is different behaviour than how it’s trended for other countries once they hit 150 cases. A possible explanation could be that there were deaths that were not recorded as Covid-19 over the first 2–3 weeks. I’m not suggesting any foul play here, just that Covid-19 was not really on the most people’s radar in the US, and given that the Covid-19 fatality rate is much higher for the elderly, their deaths could’ve been attributed to old age, pneumonia etc. If you’ve seen any sources about this or have any suggestions, please let me know!
Yesterday, I had provided a very rough back of the envelope framework for estimating how long lockdowns could last and used the US as an example. A friend of mine shared this great 538 article that cited a very recent survey of 20 actual experts (unlike yours truly) and had some pretty insightful results. You can find the article here and download the actual survey results here. A question in the survey was when the experts expected hospitalizations to peak, which could also indicate the peak of the pandemic. According to the survey, there’s a 69% chance the peak will fall between May and August. Some of the experts’ responses to the survey had fairly broad ranges, which I found concerning. Another development yesterday was that Neil Ferguson, the UK scientist behind the Imperial London projections that were fairly catastrophic (estimating 2.2M deaths in the US), significantly revised down his estimates. You can read about that here. There is also increasing skepticism of the accuracy of the case and death data coming from China — especially the hardest-hit Wuhan region. The sentiment seems to be that the actual numbers are a lot higher than official figures. I’ve shared a few of those stories on my daily posts before (here and here). Here’s the latest, which is about the number of urns not matching the official fatality count.
So we are seeing a wide range of responses from the experts, significant changes in projections in a matter of days, and declining confidence in the Chinese data that a lot of the western countries are trying to model their responses after. My takeaway from all this is we still know very little, a lot more ambiguity awaits us, and it’s too early to let our guard down.
Note — How I calculate 7-day CDGR:
Taking this method of calculating CAGR but applying to daily.
power(Day 8 total/Day 1 total, 1/7) — 1
So for the US, total deaths yesterday was 1050. beginning total 7 days ago was 150.
power(1050/150, 1/7) — 1 is 32%
