SARS-2 March 24 Update

2020-03-25Posted by xaminmo

First, the update:

03-24 Tuesday
1-day Projection: WW 425953/18600; PRC 81595/3283; NON 346031/15382; IT 69104/6744; US 57310/731; TX 916/10
Actual Numbers: WW 417966/18615; PRC 81591/3281; NON 336375/15334; IT 69176/6820; US 53740/706; TX 955/12

03-25 Wednesday
1-day Projection: WW 461807/21005; PRC 81686/3288; NON 381238/17782; IT 74856/7654; US 66137/903; TX 1203/16

Infection spread appears to be reducing in US and TX, even though TX was above projections for yesterday.
US shows 134% instead of 135%, and the last 3 days were 130%, 131%, and 123%.
TX shows 135% instead of 147%, and the last 3 days were 108%, 121%, and 126%.

This implies that the declaration of pandemic may have had positive effect.

The numbers from tonight or tomorrow will start reflecting any changes caused by the declaration of national emergency.

Milestones would have pushed out a day, but my methodology was poor. I now show the number for the condition in the left of the note, not the estimated number for that day. Also, I dropped stage 2 here, and just show worst case reserve (smallest number) and best case reserve (largest number). I do not have proper numbers to separate stage 1 (over standard) vs stage 2 (over minimum reserve).

04-05 29k TEXAS ICU OVERRUN STAGE 1
04-06 36k TEXAS VENT OVERRUN STAGE 1
04-10 146k TEXAS ICU STG3
04-11 178k TEXAS VENT STG3 & Hosp STG1
04-14 400k may be the inflection point for Texas
04-15 532k TEXAS HOSP STG3

Milestones are based on guesstimates, because the exact number of ICU beds and placement of ventillators is proprietary data that is hard to find and confirm. My estimates are based on: TX 28.7m pop; 2.9 beds per 1000 in TX (83230); 32% unoccupied (26633) / 11% are ICU (2929); 6.7% ventilators (1787); 10% limited function ventilation (2663) / 5-15% conf need hosp(532k-177k @ overrun); 2-10% ICU (29k-146k); 1-5% need vent (36k-178k)

Testing infrastructure may fail before the inflection point, leading to a false decrease in numbers reported. If that happens, we may not numerically reach the inflection point, which is 60% of the population for R0 of 2.2. Remember, we only confirm 12-15% of the actual cases, as many are mild or even asymptomatic, but are still infectious to others.

I expect these to move further out each of the next few days, and then it will probably look like it levels off a bit for several days before reducing further.

We are not out of the woods. Hospital usage is 1-4 weeks. Average infection course is 20 days with no hospitalization, but once people are bad enough to need O2 or ventilation, it takes longer to recover well enough to not need it anymore. If we were to fall to 112% today (unrealistic extreme) and stay there, the milestone dates become 04-23, 04-25, 05-07, 05-09, 05-16, 05-18. That spreads out enough that we may gain an extra couple of days on each due to early cases resolving (people do not stay in the hospital forever).

The current milestones for the US, since TX is about 9% of the US:
03-31 29k -> 322k US ICU OVERRUN STAGE 1
03-31 36k -> 400k US VENT OVERRUN STAGE 1
04-05 146k -> 1622k US ICU STG3
04-06 178k -> 1977k US VENT STG3
04-08 400k -> 4444k US inflection point
04-09 532k -> 5800k US HOSP STG 2

These are MUCH more fuzzy, since it is not exactly 9%, and ICU, Vent and bed capacities vary. That adds 1-2 days uncertainty. This also will be affected by any changes in the numbers as discussed above.

Here are the milestones from 04-23 data for comparison to track our progress flattening the curve:
04-06 33k TEXAS ICU OVERRUN STAGE 1
04-07 45k TEXAS VENT OVERRUN STAGE 1
04-10 150k TEXAS VENT/ICU STG2 & Hosp STG 1
04-13 366k TEXAS VENT/ICU STG3 & Hosp STG 2
04-15 660k TEXAS Hosp STG 3

NOTE that the very first milestone was 04-02 based on 570 ICU beds, and was abandoned as a predictor.

When testing and behavior are nonchanging:
Model is +/- 25% per week
That is +/- 1.5 days for stage 1
That is +/- 2.5 days for stage 2
That is +/- 3.5 days for stage 3

Inflection point for R0 of 2.2 is 60%.
Current reporting rate is 12-15%.
Assuming we keep testing by same criteria, the Tx inflection point is 400k confirmed.
We may not reach this numerically due to testing/infrastructure failure.
eg, we may reach stage 3 while our confirmed rate is much lower, but still around the same days.

This is all bistromath, and both reporting rates, and trends are changing daily at this point.

Also note that the numbers lag by up to 24 hours based on reporting systems.

Also Also note that it still takes around 12 days for action changes to affect trends, since people infected today will not be detected for a while.

Also Also Also note that there are many people who have NO symptoms at all, but are still infecting others. This, along with poor testing, explains the 12-15% confirmation rate.

https://drive.google.com/file/d/1vocCN445AZyVBBLsv0kJR8ZDP9DM0UST/view

Overrun Planning

2020-03-24Posted by xaminmo

Stats for Sunday, Monday, and predictions for Tuesday, and Texas hospital overruns.

03-22 Sunday

7-day Projection: WW 262771/11285; PRC 81266/3324; NON 259371/14696; IT 78550/8986; US 25567/280; TX 521/21. plus or minus 25%.
1-day Projection: WW 340729/14895; PRC 81360/3265; NON 260988/11728; IT 61049/5774; US 34015/386; TX 857/5.
Actual Numbers: WW 335955/14632; PRC 81397/3265; NON 254558/11367; IT 59138/5476; US 33272/417; TX 627/8

03-23 Monday

1-day Projection: WW 370630/16503; PRC 81489/3271; NON 290297/13301; IT 65275/6215; US 43432/566; TX 627/13. TX Deaths is Still early and erratic
Actual Numbers: WW 378287/18600; PRC 81496/3274; NON 296791/13223; IT 63927/6077; US 43667/552; TX 758/9. IT is slowing, which is good. JHU new dataset.

03-24 Tuesday

1-day Projection: WW 425,953/18,600; PRC 81,595/3,283; NON 346,031/15,382; IT 69,104/6,744; US 57,310/731; TX 916/10
Actual Numbers: To be determined. Infection spread for US is fluctuating slightly.

The following major milestones assume no gross change in testing rates nor confirmed infection rates. Stage 1 is over standard capacity. Stage 2 is over worst case reserve capacity. Stage 3 is over best case reserve capacity. This does not cover additional production, but does cover identified lower-function and out-of-date equipment from federal, military, and major hospital stockpiles. At Stage 1, alternative locations are getting converted for use by patients, such as closed medical buildings. At stage 2, we’re relying on medical and nursing students as front-line caregivers, and MASH style pop-up tent hospital expansions start getting deployed where possible. At stage 3, we’re draping parking garages, and getting scouts with first-aid badges to help. Elderly or anyone with comorbidities will be comforted, but won’t get access to mechanical ventilation. Only those with the best chance of survival with, and a high risk of death without, would get advanced care.

04-05 33k TEXAS ICU OVERRUN STAGE 1 / 28.7m pop; 2.9 beds per 1000 (83230); 32% unoccupied (26633) / 11% are ICU (2929); 6.7% ventilators (1787) and 10% limited function ventilation (2663) / 5-15% conf need hosp(532k-177k @ overrun); 2-10% ICU (29k-146k); 1-5% need vent (36k-178k)
04-06 45k TEXAS VENT OVERRUN STAGE 1 / 28.7m pop; 2.9 beds per 1000 (83230); 32% unoccupied (26633) / 11% are ICU (2929); 6.7% ventilators (1787) and 10% limited function ventilation (2663) / 5-15% conf need hosp(532k-177k @ overrun); 2-10% ICU (29k-146k); 1-5% need vent (36k-178k)
04-10 150k TEXAS VENT/ICU STG2 & Hosp STG 1 / 28.7m pop; 2.9 beds per 1000 (83230); 32% unoccupied (26633) / 11% are ICU (2929); 6.7% ventilators (1787) and 10% limited function ventilation (2663) / 5-15% conf need hosp(532k-177k @ overrun); 2-10% ICU (29k-146k); 1-5% need vent (36k-178k)
04-13 366k TEXAS VENT/ICU STG3 & Hosp STG 2 / 28.7m pop; 2.9 beds per 1000 (83230); 32% unoccupied (26633) / 11% are ICU (2929); 6.7% ventilators (1787) and 10% limited function ventilation (2663) / 5-15% conf need hosp(532k-177k @ overrun); 2-10% ICU (29k-146k); 1-5% need vent (36k-178k)
04-15 660k TEXAS Hosp STG 3 / 28.7m pop; 2.9 beds per 1000 (83230); 32% unoccupied (26633) / 11% are ICU (2929); 6.7% ventilators (1787) and 10% limited function ventilation (2663) / 5-15% conf need hosp(532k-177k @ overrun); 2-10% ICU (29k-146k); 1-5% need vent (36k-178k)

The inflection point for R0 of 2.2 is 60%. Current reporting rate is 12-15%. Assuming we keep testing by same criteria, the Tx inflection point is about 400k confirmed. We may not reach this numerically due to testing/infrastructure failure. eg, we may reach stage 3 while our confirmed rate is much lower, but still around the same day, and still around the same number of actual infected. Also, there are different groups between spreaders and isolators. I don’t know the balance of those two groups. They could be 20/80 or 50/50.

Barring major changes, the model is +/- 25% per week. That is +/- 1.5 days for stage 1. That is +/- 2.5 days for stage 2. That is +/- 3.5 days for stage 3.

If the Pandemic declaration helped, then 03-24 will be at least 2% low for US and TX. If the Emergency declaration helped, then 03-25 will be at least 2% low for US and TX. Either of those should show a continual downward trend. We have had a lot of people ignoring expert and government recommendations, so I do not expect an abrupt change.

If we did abruptly fall to, say, 112% on 03-24 and stay there, then Stage 1 starts April 23-26; then Stage 2 starts May 7-11; Stage 3 may never happen due to average disease cycle of 20 days, and ICU cycle of 30 days. This would be a dream scenario, and is unlikely. More likely to see a 5% drop several days in a row. It’s unlikely to see the clam-down go below 110% until it looks really bad (and then it’s too late).

Texas is about 9% of the US capacity and slightly more capacity than average. Look for 11x numbers in the US column for similar problems.

33k -> 363k on 03/30 Stage 1 ICU
45k -> 495k on 03/31 Stage 1 Vent
150k -> 1650k on 04/04 Stage 2 / Stage 1 Hosp
366k -> 4026k on 04/07 Stage 3 / Stage 2 Hosp
660k -> 7260k on 04/09 Stage 3 Hosp

112% projections:

112% ICU Stage 1 is 04/10
112% Vent Stage 1 is 04/12
112% V2 / Hosp Stage 1 is 04/23
112% V3/H2 is 05/01
112% H3 is 05/06
112% numbers are +/- 3, 5, and 7 days.

This is all bistromath, and really, anything more than a week out is just guesswork. A lot can change in a week, and I’m expecting substantial changes over the next 3 days based on the activities 10-14 days ago.

The Spreadsheet has been updated. JHU replaced some data sources, so it was a little annoying, and a little more manual entry.

https://drive.google.com/file/d/1vocCN445AZyVBBLsv0kJR8ZDP9DM0UST/view

Projection Spreadsheet

2020-03-21Posted by xaminmo

Here’s the spreadsheet (attached). It’s not super sophisticated.

On “Projection”, yellow is projection, blue is confirmed from JHU, green is confirmed from some other agency such as the WHO, or a medical report that maybe didn’t make it into the stats in time that day. Coloring is all set manually.

I manually update the last three sheets by cut/paste from their GitHub project. I created dynamic tables at the top of each of those three pages.

That is what I cut/paste into “Projection” (transpose/transform on paste). I do it manually so I can copy down projections, and because that’s a bit complicated for my Excel skills.

As a day on “Projection” is updated, the calculation cells between confirmed and dead are manually copied down one row to replace the yellow/grey lines. That way they reflect actual percentage changes rather than projections still.

The graphs just got updated to dynamic. As columns are added to the tables on the last two sheets, it should update the chart, but I’ve not tested it yet.

UPDATED: SARS-COV-2_COVID-19_Projections_c

US must be testing

2020-03-20Posted by xaminmo

Must have been a lot of testing. US count doubled yesterday. Everything except China showed up as higher than projected infection rates. US and TX death rates are still low. Italy death rates are still high. Here are the numebers for 2020-03-19. Today’s numbers release sometime in the next 2-3 hours.

Where     Cases   Deaths  Recovered  Mort/A   Mort/R
World   242,713    9,867     84,962   4.07%   10.41%
China    81,156    3,249     70,535   4.00%    4.40%
NonPRC  161,557    6,618     14,427   4.10%   31.45%
Italy    41,035    3,405      4,440   8.30%   43.40%
US       13,680      200        108   1.46%   64.94%
Texas       260        1          0   0.38%  100.00%

Today’s numbers release sometime in the next 2-3 hours. The estimates for today’s numbers (03-20) are 274102/11148 WW; 81210/3257 PRC; 195048/7975 NON; 47150/3893 IT; 24034/339 US; 391/8 TX

Data source is Johns Hopkins University repository here:
https://github.com/CSSEGISandData/COVID-19/tree/master/csse_covid_19_data/csse_covid_19_time_series

Projections are just Excel, divide today by yesterday to get the change day to day, then use a 7-day average to guess what tomorrow might be, and do that again for each row.

In the US, first and second hand reports show that people are tested for seasonal influenza first, and if they test positive, they are not tested for SARS-2. This is conserve the the limited number of tests. It’s less likely for people to have both COVID19 and influenza at the same time. This may change as testing capability increases. According to Trump, Rosche, Quest, Labcorp, and Wal-Mart may be rolling out rapid testing capability starting in April.

In the US, it looks like the confirmed cases are about 13% of the actual infected. The actual death rate for non-overwhelmed places tends to be about 4.5% of the confirmed cases, when there’s ~14% testing (PRC and WW average).

The real mortality rate will therefore be that much less, which puts it in the same order of magnitude as seasonal influenza, but without the ability for people to get vaccinated, and a higher infection rate due to longer infectious incubation period.

Projections for 03/17 and 03/22

2020-03-17Posted by xaminmo

I run an infection rate projection based on the last 7 days change rates using the JHU case data stats which are updated around 18:30 Central/US each day (vs 04:00 Central/US for the WHO data).

For Monday, 03-16, it was generally within ~1%:

World projection was 179621 confirmed and 7127 dead.
World actual was 181546 confirmed and 7126 dead.

China projection 81029 / 3213.
China actual was 81033 / 3217.

Non-China projection 99472 / did not project dead.
Non-China actual was 100513 / 3909

US Projection 4490 / 74.
US Actual was 4632 / 85.

TX Projection was 91 / 1.
TX Actual was 85 / 1.

For 03/17-18:30ish projections, I have:
Est: WW 196830/7885 ;
PRC 81063/3231;
NonPRC 116869/4721;
IT 31635/2574;
US 6132/115;
TX 100/1 though at low numbers, dead could be 1-6.
For 1 day, I’d expect to be within 2% on any of these numbers.

Note that the real-time trackers are not exactly aligned with this data, and the covid19info.info page I like so much is using a different data source for China, or has not refreshed from the formal stats. They do not match JHU nor WHO, but the other data matches. I’m thinking maybe they run a projection instead of real stats, and need to add newer datapoints.

Also note that Italy and China’s death rates are higher than the US, and higher than the world minus China. If it were just the US, then I would say maybe it’s affluence, or testing anomalies. Instead, I’m thinking that Italy had China’s first strain, and that perhaps the rest of the world has a secondary strain of the virus? I don’t really have the tools, skills, nor time to analyze the gene sequences to build a CV19 family tree to confirm this.

I don’t expect to see a change in the overall growth rates until sometime next week. Pandemic declaration was on the 11th, so Monday the 23rd will probably start to show a dip, plus or minus a day, depending on how much social distancing is actually happening, and when people started. Plenty are still ignoring, or selectively ignoring. I know I ordered food, and the family went to grab things from the university dormitory. Plenty of public facing workplaces still are interacting with unscreened, unlimited people.

For Sunday the 22nd, I’m expecting 262771/11285 worldwide; 81266/3324 China; 259371/14696 non-PRC; 78550/8986 Italy; 25567/280 US; 521/21 Texas. The US numbers could be higher if the two hot zones overrun their hospital capacity, or if a new hot zone pops up. 7 days is a long time, so let’s say it could be plus or minus 25%.

If you want to be safe, then remember that most incubation is under 14 days. People who have self-quarantined for 14 days can co-mingle with other people who have self-quarantined for 14 days and be at very low risk. The longest incubation on record was 27 days, so groups or mutual 27-day self-quarantine would be ultra-low risk. There’s never a zero risk, since there will invariably be one or two asymptomatic carriers. Also, anyone who lies, or anyone who breaks these rules will potentially be a vector for infection into that entire cluster of people. Self-quarantine means no contact outside of a closed group. I’m not sure the increased risk from food deliveries, well-spaced grocery store visits, etc. (Did the cook dip their glove in your food accidentally after scratching their face? Did the stock crew at the store have someone not safe? Did you disinfect the outside of the packages you got in the mail?).