This entry will focus on Arizona, Florida and Texas and compare them to New York, Massachusetts, and New Jersey. Previous entries: https://eeholmes.github.io/CoV19/Updates.html

Like the northeast states, Arizona, Florida and Texas are now entering an epidemic phase where they have lost control of the virus. One of the difficulties for the US is that we cannot close state borders effectively. States can require quarantines but that is hard to enforce. We have seen over and over, how hotspots develop in locations that are draws from people from other regions. Travel into and out of US hotspots continually seeds infections in other areas and at some point enough seedings occur and enough take hold. Most disease seedings into a new area die out (the person doesn’t spread it to enough people), but a few become ‘super-spreader’ events–i.e. the person attends a few big social gatherings in that key pre-symptom window when they feel “ok” but have high virus load in their lungs and exhale that all over an enclosed space with many people. The epidemics in AZ, TX, and FL are certain to spread. There is constant travel into and out of those states.

The pattern of the epidemics in the northeast were multiple “seedings” of the disease due to travel in from Europe (particularly Italy) followed by “silent” (not much testing) spread throughout the community, and then a “flash” of very fast increases in cases, followed approximately two weeks later by very fast increases in deaths. The “flash” is just how exponential growth appears to us. When numbers are doubling at low levels, the increases are not noticeable and then once the numbers get bigger, they get big really really fast.

Florida, Arizona, and Texas are now showing a similar pattern. The initial lockdown was not successful (cases were not sufficiently driven down), once lockdown was lifted community spread resumed. Now they are in the “flash” phase where cases increase rapidly but deaths have not yet entered the flash phase.

Here is a plot of the 7-day average new cases per day per million people. The x-axis is days since the average per capita new cases per day reached 100. The dashed lines are NY, NJ and MA while the solid lines are AZ, FL, and TX. The AZ-TX-FL lines are quite similar in the rate of increase to the NY-NJ-MA lines. I put Sweden on just because I hear Sweden used as an example of a country that let the virus “take its own course” but that is entirely untrue. They imposed many mitigations to prevent a large epidemic occurring and rather staying with steady, manageable, levels. I also included Denmark to show a European country that controlled the virus and didn’t let it spread at all (most of Europe looks like this, exceptions being Belgium, UK, Spain, Italy, and Sweden)

However the epidemics in the northeast (and eastern midwest) were most severe in the big cities (number of deaths and deaths per million), with death rates of 0.26% NYC, 0.25% Newark and ).12% Boston.

region death.per.100
Essex New Jersey US 0.2525736
NYC New York US 0.2691795
Suffolk Massachusetts US 0.1248901

Let’s look at all the counties over 750,000 people that experienced an epidemic and are now well past it with current new cases less than 10% of what they were at the peak. There are 16 such counties representing the NYC Metro region, Newark Metro region, Boston Metro area, Detroit Metro area, and New Haven, CT. Chicago doesn’t appear on this list as its epidemic has not reached 10% of its peak. Note, the 5 NYC boroughs appear just as “NYC” in the table.

region max.pos curr.pos
Nassau New York US 1005.7411 29.68900
Suffolk New York US 983.5329 33.37781
Westchester New York US 981.3154 38.68562
Bergen New Jersey US 708.6140 42.75591
Suffolk Massachusetts US 639.1997 41.93804
Essex New Jersey US 628.1262 29.50209
NYC New York US 625.4357 40.66300
Fairfield Connecticut US 515.3465 20.74713
Middlesex New Jersey US 508.1869 22.85543
Essex Massachusetts US 474.9026 38.20223
Wayne Michigan US 437.9603 40.34168
New Haven Connecticut US 357.1608 16.88032
Middlesex Massachusetts US 354.9050 27.92085
Worcester Massachusetts US 320.9299 31.12986
Macomb Michigan US 255.6473 20.43217
Oakland Michigan US 255.5921 24.87763

16 new metro counties have reached the threshold 200 new cases per day per million people: Phoenix, 8 counties in the Miami-Dade region, Houston, Dallas, San Antonio, Atlanta (one county of it), Salt Lake City, and Memphis. Based on following the news, 200 is approximately the threshold when many US metro county health systems are quite close to being full and the county needs to begin taking strong actions to try to stop the spread of coronavirus. Houston (Harris County) announced that its hospitals had reached capacity at that threshold. San Antonio (Bexar County) sent an alert today that its hospitals are reaching capacity. Phoenix activated its hospital surge plans two days ago. Phoenix must have very high hospital capacity (or was doing extensive testing and catching many low-symptom cases) because it activated its surge plan when new cases per day per million was up above 300. Miami-Dade county must have even higher hospital capacity as it is reporting 30% of hospital and ICU beds available even with new cases per day per million also above 300. However, I have reports of individual hospitals in South Florida having reach capacity.

region max.pos
Maricopa Arizona US 448.1510
Orange Florida US 437.5567
Hillsborough Florida US 369.8644
Miami-Dade Florida US 340.8246
Pinellas Florida US 333.3347
Duval Florida US 321.1379
Lee Florida US 259.5458
Harris Texas US 258.0829
Gwinnett Georgia US 238.1842
Broward Florida US 234.7571
Mecklenburg North Carolina US 232.8725
Palm Beach Florida US 231.1644
Hidalgo Texas US 219.7025
Salt Lake Utah US 214.0819
Shelby Tennessee US 213.7142
Bexar Texas US 206.9180

We can compare the 7-day average new cases per day per million people for these metropolitan counties to the counties with Phoenix, Houston, San Antonio and Miami. The Phoenix, Houston, San Antonio and Miami lines are the solid line. It would appear, at this early stage, that these cities are following the pattern of the other cities that had epidemics and we can use that to make a guess at what’s ahead in Florida, Arizona and Texas.

Where are Arizona, Texas and Florida headed?

Using the patterns seen in the northeast outbreaks, we can make a guess at the mortalities and total cases to expect in the current outbreaks in the south/southwest. These estimates are probably a lower bound however. At around day 0 (when new cases per million were close to 100), NY, NJ, and MA employed a strict lockdown to stop their epidemics. AZ has gone well past that threshold without any severe mitigations, like a “shelter in place” order. Only recently were cities in Texas allowed to impose their own stricter regulations to enforce social distancing. We have yet to see a cases where a city was not put into lockdown to stop a coronavirus epidemic once is got out of control. So we are entering new territory here. Though Sweden is often touted as a country that did not employ mitigations, that’s not true. They didn’t lockdown like the rest of Europe, but certainly employed a variety of mitigations to reduce spread and have never entered a “flash” phase where they lost control of the epidemic in the way that we are seeing in some US states (see the first plot).

Mortality patterns

In the northeast, mortality started rising 7 (NY) to 14 (MA) days after cases started rising and it was only after daily new cases per million reached 100 (this is 10 on the graph because I divided new cases by 10) that the death curves started increasing. In NJ, this is really apparent. Reported deaths seem fine, fine, fine and then they started zooming up when daily new cases per million hit 200. Why the delay? First, people are on ventilators for a long time before dying and there are delays in death reporting. There are a few layers reporting (hospital, county, then state) before it appears on the state-wide lists. Some states have longer reporting delays than others.

In Arizona, Florida and Texas, we are not yet seeing a rise in mortalities. We will eventually. The disease still has no cure. But for Arizona in particular, we are seeing an unusually long delay before the daily death curve begins to track the new cases curve. My guess is sometime in the next 7 days, it’ll start tracking. You can just begin to see that the Arizona mortality curve is start to go up (20 days after new cases per day hit 100 per million). In the northeast, we saw the same delay but much shorter. New cases were going up, but the deaths were not so there was a sense that “Maybe we got lucky. Maybe people won’t die here.” and then the deaths start piling up. We saw that in Germany too. For so long, people thought Germany was somehow special. Nope it just took longer.

Expected Mortality and Cases

Using the northeast states, we can ballpark estimate what the total cases and mortalities might be in Arizona, Texas and Florida. I’m using NY, NJ, and MA because these states lost control of the epidemic. Even after the lockdown, the cases went up 4-5x in those states and went up 30x in NYC. Nothing the public officials could do would stop the increase. Eventually the lockdown worked (or it burned itself out?) but it took awhile. Now AZ, TX and FL are the new northeast. Here are the current positive cases and deaths per 100 in the northeast. At the state level, the range is 1.5 to 2 for positive cases per 100 and 0.11 to 0.17 for deaths.

region positive.per.100 deaths.per.100
Massachusetts US 1.573347 0.1168226
New Jersey US 1.923771 0.1682918
New York US 2.014659 0.1612455

For the urban centers, I used the data from all metro areas minus New York county (NYC) that had an epidemic (7-day average new cases per day over 200 per million people). The range is 0.8 to 3.6 for positives and 0.1 to 0.3 for deaths. I’ll use these as my upper and lower ranges for what to expect in the AZ, TX and FL metro counties. But I am going to multiply the death number by 4/7 because the CFR in the northeast was around 7% whereas in AZ, TX and FL, it has been closer to 4%. This may be over-optimistic on my part to use 4/7ths. Given that states are reporting that the median age of infection is lower than was reported earlier, I think the CFR is likely to be lower.

Metro counties that have experienced a large epidemic
region positive.per.100 deaths.per.100
Bergen New Jersey US 2.0676849 0.2127221
Essex Massachusetts US 2.0313447 0.1389040
Essex New Jersey US 2.3399981 0.2525736
Fairfield Connecticut US 1.7603558 0.1457599
Macomb Michigan US 0.8319488 0.1035502
Middlesex Massachusetts US 1.4803633 0.1147237
Middlesex New Jersey US 2.0269992 0.1596244
New Haven Connecticut US 1.4360807 0.1250648
NYC New York US 2.5721327 0.2691795
Oakland Michigan US 0.9451456 0.0862765
Suffolk Massachusetts US 2.4584933 0.1248901
Suffolk New York US 2.7937811 0.1339563
Wayne Michigan US 1.2907131 0.1548010
Westchester New York US 3.5915023 0.1466658
Worcester Massachusetts US 1.4811792 0.1107604
The state estimates based on what NY, NJ, and MA experienced are 3, 4.2, 8.1 times higher for AZ, FL, and TX than current deaths on the low end and 4.4, 6.1, 11.7 times higher than current reported deaths on the high end.
Estimated mortalities based on epidemics in NY, NJ and MA
region current.cases min.cases max.cases current.deaths min.deaths median.deaths max.deaths
Arizona US 70124 114519 146641 1600 4859 6707 7000
Florida US 132545 337919 432703 3390 14338 19790 20654
Texas US 145894 456206 584168 2388 19356 26717 27884

The metro county estimates (only deaths shown) based on what other metro counties have experienced are

Estimated mortalities based on epidemics in other metro counties
region current.deaths min.deaths median.deaths max.deaths
Bexar Texas US 105 988 1590 3082
Broward Florida US 382 963 1550 3004
Duval Florida US 64 472 760 1473
Gwinnett Georgia US 169 462 743 1440
Harris Texas US 369 2324 3741 7250
Hidalgo Texas US 31 428 690 1336
Hillsborough Florida US 132 726 1168 2264
Lee Florida US 152 380 612 1185
Maricopa Arizona US 741 2211 3560 6899
Mecklenburg North Carolina US 145 547 881 1708
Miami-Dade Florida US 947 1339 2157 4179
Orange Florida US 56 687 1106 2143
Palm Beach Florida US 492 738 1188 2302
Pinellas Florida US 149 481 774 1500
Salt Lake Utah US 106 572 921 1785
Shelby Tennessee US 182 462 744 1442

Here are the predicted deaths in graphical form.