Ohio's rapid open started 5/1, the creep back up has started.

[Ms. Toad]

Last edited Fri May 22, 2020, 03:33 PM - Edit history (1)

Ohio started its reopening process on May 1 (without waiting for a 2-week decline in cases). It has opened at a pretty rapid pace - as of Monday (5/18) hair salons were open, and on June 1, for example, caterers will be permitted serve banquests for up to 300 people, so long as they can meet social distancing guidelines). I figured we'd see an uptick in cases starting around May 15. Looks about right.

The relatively small numbers are susceptible to daily noise - so I'm plotting rolling 7-day averages.

The last 7 days of rolling 7-day average ICU admissions (5/16 - 5/22) have stayed the same or incrased increased (In the 7-day period before, the rolling average increased on only 2 of the 7 days.). Each number below represents the 7-day average number of ICU admissions starting May 9, continuing through today, May 22. I.e. the first number is the average number of ICU admissions from May 3 - May 9. The second number is the average number of ICU admissions from May 4 - May 10, and so on. Bold indicates an increase from the previous 7-day average.

5/9 - 19.14285714
5/10 - 18.14285714
5/11 - 18.14285714
5/12 - 15.57142857
5/13 - 13.85714286
5/14 - 14.42857143
5/15- 12.71428571
5/16 - 14.28571429
5/17 - 14.28571429
5/18 - 15.85714286
5/19 - 17.85714286
5/20 - 17.28571429
5/21 - 18.42857143
5/22 - 19.85714286

If you plot this curve, the uptick is striking.

Out of the last 7 days of rolling 7-day average new cases (5/16 - 5/22), 5 of the last 7 have increased. In the prior 7 days (5/9 - 5/15) there was only an uptick on two days.

Each number below represents the 7-day average number of new cases starting May 9, continuing through today, May 22. I.e. the first number is the average number of new cases from May 3 - May 9. The second number is the average number of new cases from May 4 - May 10, and so on. Bold indicates an increase from the previous 7-day average.


5/9 - 623.1428571
5/10 -595.2857143
5/11 - 614.7142857
5/12 - 611.5714286
5/13 - 592.1428571
5/14 - 603.7142857
5/15 - 562.5714286
5/16 - 539.5714286
5/17 - 548.8571429
5/18 - 525.2857143
5/19 - 528.8571429
5/20 - 530.7142857
5/21 - 544.2857143
5/22 - 548.5714286

I plotted this second curve (using the daily numbers, rather than averages) last night and noticed a 3-day uptick in actual cases

These numbers are not as solid as my national numbers. Things jump around more on a day-to-day basis when a single case is such a large portion of the overall cases But it still looks to me like we are trending up on the leading indicators.

[FarPoint]

1. I have no idea what the long numbers/ list represents...

Maybe a link could help...

[Ms. Toad]

2. I described each in the sentence that preceded the numbers.

I'll edit to clarify. But, as I indicated, theses are runing averages - so there is no direct link.

[doc03]

3. The county I live in Belmont only had a dozen or

so cases and 2 deaths May 1st now we have 374 cases an increase of 26.3 % in 7 days and 11 deaths. We have a prison with 2750 inmates and over 400 employees that account for many of them. Two residents and one employee at the nursing home near me died this week.

[Ms. Toad]

4. Definitely hot spots -

One of the reasons I find it challenging to work the data at this stage in the game. The data is so small that whatever I am seeing can be caused by an isolated outbreak (rather than the general increase in contacts associated wtih rapid opening)

[doc03]

6. Some of the other prisons like in Mahoning county were hit real bad earlier

this one started late it will get much worse.

[Tennessee Hillbilly]

5. The number of tests per day should be accounted for.

If they've increased the amount of testing, then more new cases will be discovered. But you didn't mention this possibility, so there's no way to know for sure from the data you gave.

[Ms. Toad]

7. Tests are only remotely related to ICU admissions,

the first set of data I included. Testing of people admitted to the ICU has not changed much in this time period.

[localroger]

8. It's exponential, so this is much more dramatic than it looks

The rolling average lags by the length of the average, so this is only now showing the full effect of the first few days of reopening. Note how quickly it stopped the decline cold. This will accelerate quickly. It won't look too bad through next week, but by the middle of June it's gonna be rolling.

[Ms. Toad]

9. It's not a firm enough trend (and the numbers are too "wobbly")

for me to make predictions I would consider reliable yet. (You may or may not recall that I've been making predictions since sometime in late July when I predicted when the death toll would exceed that of SARS - so I'm not shy about predictions when I think they are warranted, which is frequently long before others believe they are warranted).

I trust what I am seeing enough to start calling it out as happening, but not enough to make any prediction beyond it that the downward trend is reversing on roughly the predicted time frame after opening.

I agree with you wholeheartedly from a thought experiment perspective. We started opening up with upwards of 20,000 cases. That is a pretty weighty place to start growing exponentially - and from what I've seen when I go out there are way too many yahoos insisting on their right to spread the plague (some of them being out right nasty about it).

[localroger]

10. It's very clear that something dramatic happened

From 5/9 to 5/16 the ICU admissions show an overall drop from ~20 to ~13, averaging about 1/day. The trendline actually drops a bit faster than that at first, wobbling down a bit more than -2 from the trendline on 5/13, which probably means the actual reversal occurred around the 13th and the moving average smoothed out the transition from a faster decline to a more modest increase. Had that trendline continued, the admissions on 5/22 should have been between 3 and 8. Instead they have ticked hard upward, from ~13 back to ~20, now never varying more than +/- 1 from that new trendline which suggests the new trend is fully within the average window.

The situation for new cases is similar; despite the noise, the trendline of the final four or five days is completely incompatible with what was happening the first four or five days. It's not too early to say something has changed, and dramatically.

What it is too early to do is estimate the new doubling rate. However, we will probably have enough data to do that within another week to 10 days.

[uponit7771]

11. There was a jail outbreak in Ohio if I'm not mistaken but the trend line is still up

[Ms. Toad]

12. Here is the underlying daily data for ICU admissions

Last edited Sat May 23, 2020, 12:44 AM - Edit history (1)

Weekends bolded:

5/2 - 10
5/3 - 12
5/4 - 12
5/5 - 33
5/6 - 28
5/7 - 16
5/8 - 21
5/9 - 12
5/10 - 5
5/11 - 12
5/12 - 15
5/13 - 16
5/14 - 20
5/15 - 9
5/16 - 23
5/17 - 5
5/18 - 23
5/19 - 29
5/20 - 12
5/21 - 28
5.22 - 19

[localroger]

13. You need the 7 days before 5/9 to generate the moving average

I suspect that would make it clear even from the raw data that something dramatic happened around 5/13.

[Ms. Toad]

14. I've added it.

It's not really a single day when something happened (the numbers are too small to be reliable, as single day data points, but the mddle week of the three (5/9 - 5/16) was the trough before things started trending up again. That dip, then upward trend is much clearer from the averages than it is from looking at the individual day's admissions (which are all over the place. (https://coronavirus.ohio.gov/wps/portal/gov/covid-19/dashboards/current-trends - bottom chart)

[localroger]

15. Well that's why we use moving averages

I work with industrial data collection systems and yeah, that's the difference between raw and filtered data. The filter adds some distortion, particularly time lag and it can hide legitimate step and impulse events, but it can reveal a solid change in trend that is otherwise hidden by noise. I am comfortable enough with this data to declare that a lot of Ohioans were probably jumping the gun during the week of the 11th before the official reopening on the 15th.

[Ms. Toad]

16. Ohio started reopening things on the 1st.

for non-urgent medical matters, offices and manufacturing on the 4th. So not really jumping the gun - but significantly more people were out and about starting the week of the 4th.

[localroger]

17. That actually explains a lot

The downward curve is a bit irregular even with the moving average, but this explains it. There were different things coming online at different times. But once the reopening was complete, the trend became much more regular.

[Ms. Toad]

18. it was a rolling start with something new opening up about every 3 days or so.

Dine-in restaurants started back on Friday (21st). Caterers can serve groups of up to 300, as of June 1. Childcare comes back online May 31 - it will be different (and fewer in childcare) but that will free more parents to return to work

The other thing going on was informal reopening. Even before things were formally allowed to open, there were a lot of idiots who started informally opening things up (like having parties, going out more to grocery, etc. stores, rebelling against wearing masks). I went out shopping on the 9th (I believe - it may have been the 8th) and the 16th. On the 9th about 60% of the people I encountered were wearing masks (except in Target, where it was only about 10%). When I went out to largely the same stores on the 6th, mask wearing had dropped across the board to about 10% - AND - 67% of the employees in one store and 10% in another were not wearing masks (mandated for employees). As it has everywhere - the anti-mask crowd is becoming more vocal and visible about not wearing masks.

Today is the highest number of deaths since May 6. (84 - the average recently has ranged between 37 and 44). But ICU admissions are down today.