Categories
COVID Archives

Chasing down irregularities in data…

It seems that as the weeks march on, the quality and consistency of the data are deteriorating. In general, the growth in COVID tends to be exaggerated by several factors, including:

  • Some states report cases based on # of tests, rather than number of people (many positive cases are tested multiple times)
  • Some states report positive Serum tests (anti-body tests) as new cases, even though these people had COVID in the past, and are recovered
  • Some states have reviewed their medical histories as far back as March 1st, and report suspected or probable cases, which are picked up as new cases
  • Likewise, some states have gone back in history to find probable deaths, and newly report them

I have no argument with the classification that states are doing, as it largely (but not always) makes sense. The issue is where the cases are placed in the data. Since most of the reporting sites (Johns Hopkins, COVID Tracking, Infection2020) pick up total cases and deaths each day from the various state sites, many historical cases show up as new cases, which exaggerate the growth of COVID. I try to back this data out when it’s obvious and where I can find the correct data, but it’s a chore. Today my problems are with Michigan data – I’ll describe more fully below.

I’ve been seeing more and more reports that the virus is attenuating, or declining in strength. I first heard about this from Professor Martin Zelder at UNC. He calculated that the percentage of COVID cases that end up with hospital stays in NC has declined from 12% to 3% in one month’s time. He points out that this could be because we are testing more and more people that are less sick as testing expands. Then we saw the report from Italy last week that the viral load in tested subjects was a fraction of what it was (it sparked a lot of controversy).

And here is the latest from Pennsylvania: https://www.foxnews.com/health/pittsburgh-hospital-officials-coronavirus-appears-less-potent-than-before

If this is true, then this is great news.

Testing reports reached a new record today, at over 550,000. I need to spend some time thinking about how to adjust for increases in test volume and whether to do it by state or nationally. All you scientists and actuaries on this list, please send me your ideas…

I’m working tomorrow night (jazz), so will report again on Sunday.

As always, feel free to send me your questions about my assumptions, methodology, or modeling in general.

  • Likely date of active case peak (Chalke modeling): April 10
  • Likely date of peak deaths (IHME): April 16 (last revision on May 29)
  • Total Test Results reported today: 550,915 (a new record)
  • Total Pending tests reported today: 1,749 (extremely low)
  • National reported case Growth Rate today: 1.27% (very low)

Shane Chalke Interviews

https://www.fredericksburg.com/opinion/editorial-unlock-demographically-not-geographically/article_a62e6e70-dccd-51cf-b7b2-16d77a90fd9c.html

Website

Groom Ventures has agreed to host a website that will archive my daily reports, and supplement with other commentary. John Groom worked at one of my companies back in the day, and is an excellent writer. The website is: www.howmuchrisk.com For those of you that post my daily report on Facebook, let me suggest you link to this site, as the direct Facebook posts don’t seem to copy the graphs.

Daily Analysis

Here is the national picture of active cases – continued, slow decline. I suspect it’s declining faster than this graph suggests, for the reasons I’ve outlined above.

Here are the new reported cases nationally. Look at the trend line. Still tracking in the right direction.

Here are the daily reported tests. Going through the roof now. The U.S. is now approaching 20 million tests.

Here are the daily death reports. Daily deaths continue to fall steadily. This is the week that the “experts” said we’d be seeing 3,000 deaths per day. Happily, it’s not happening.

On to the states. Oh, my look at this! Virginia has crossed NC in absolute active cases. VA still has more per capita, but this is really interesting. Next week I’ll update my side by side growth analysis by date to see if I can find any evidence that the differences in restrictions might explain this.

No surprises with Washington – drifting upward lately, but a small downturn today. The numbers are small, and Washington has had several previous episodes of data anomalies.

Florida has risen for a week now. I’m going to take a look to see if it’s all in Miami as soon as I get a chance.

Both NY and NJ are in the end game, with a small percentage of COVID compared to their peaks. NY is now down 88% from their peak and NJ is down 83%. Unless something flairs up in these two states, there isn’t much more to tell…

California continues to be problematic. CA now exceeds NY on a per capita basis. IHME now projects that the California daily death count doesn’t peak until June 27th, so California could be rising for another week or two.

Massachusetts continues to do well, with a remarkably rapid recovery. MA is 83% down from the peak. Again, unless something flairs up, this story is almost over.

Georgia’s decline continues to be painfully slow – enough said.

…And here is Michigan. Does something look wrong with this picture? I’ve been studying COVID data for 3 months now, and I’ve noticed that the data everywhere is relatively continuous. This sort of spike has never happened. After much investigation, I’ve concluded (but don’t know with certainty), that Michigan has reported an extra 5,014 “probable” cases today, and 239 “probable” deaths. Michigan is also a state that reports tests as cases, rather than people. They also confound things by reporting positive antibody tests as cases. I’ve found the data necessary to adjust for this, and will by my next report. This graph DOES NOT represent what’s going on in Michigan.

Continued, almost straight line, decline in Pennsylvania. PA is now down 68% from the peak.

Texas has been rising for the past 9 days. IHME projects peak deaths in Texas on May 17, so I need to do some work to correlate what I’m seeing with the IHME data.

And finally, here is Colorado. Colorado is one of the states that has had aberrations in their data. It’s showing about flat for 2 weeks, but I’ve shied away from any conclusions in this state due to the data irregularities.

So that’s it for today. The numbers are very small as a percentage of the population. Unless you’re in a high density area, your chances of contracting COVID are very small. However, even though the probability is very small, that doesn’t help if you’re the one catching it. Everyone please continue to be as cautious as circumstances dictate.

–Shane Chalke, FSA

Leave a Reply

Your email address will not be published. Required fields are marked *