Second Surge After Re-opening? A Detailed State-by-State Analysis

Updated: Jun 12

In response to the COVID-19 pandemic, nearly every state in the United States instituted lockdowns of some sort -- most with stay-at-home orders, and others never instituting stay-at-home orders but closing non-essential businesses and schools.


This swift response was largely sparked by an Imperial College report published in March 16, which sounded the alarm and helped the USA take this pandemic seriously. Using a basic exponential growth model assuming a fixed exponent of 2.2, this report suggested that without intervention, that >80% of the USA population would be infected (>265 million), and would lead to 2.2 million deaths. Reasoning that "mitigation is unlikely to be a viable option without overwhelming healthcare systems" based on recent experiences in Wuhan, they concluded that extreme steps would be necessary to suppress the virus, including "combination of case isolation, social distancing of the entire population, and either household quarantine or school and university closures" required to avoid this catastrophe. They also suggested that these measures "will need to be maintained until a vaccine becomes available (potentially 18 months or more)". Nature just published an "accelerated review" study that posited based on modeling data through early April, the lockdown policies prevented 60 million USA infections (including asymptomatic and undetected infections) and 4.8 million additional confirmed cases (thus resulting in a 3:1 reduction assuming current 2.0 million confirmed cases will end up near 2.4 million by the end of the summer) compared with a "do nothing strategy" in which no government interventions are proposed (incidentally also assuming people would not change their behavior and take any viral mitigation steps without government edicts).


For now, I will not go in detail into any criticisms of what I believe to be simplistic assumptions the Imperial college (and Nature paper) models, assuming homogeneous viral spread throughout countries with a constant exponential growth model, or their assumption (held by many at the time but very debatable now) that milder mitigation strategies would be insufficient to keep healthcare systems would be overwhelmed, or the complete lack of cost-benefit analysis to assess the cost vs. benefit of lockdowns vs. more targeted mitigation strategies. I think these are important issues to discuss, but will save those for a later post. Later in this post, however, I will come back to what I believe to be a commonly believed erroneous premise based on this work.


Irregardless of speculation on what would have happened without lockdowns, the lockdowns (i.e. stay-at-home orders, business and school closures) have clearly "flattened the curve" and reduced the number of infections, confirmed cases, hospitalizations and deaths caused by the virus. Various guidelines were proposed for when to reopen, with some states eager to reopen quickly and others taking a more conservative approach. Ten states lifted their stay-at-home orders or other restrictions in April, and another 12 lifted them in the first week of May. There was an apoplectic response to these openings, in particular Georgia's opening nearly universally panned as reckless (even by President Trump!), and serious concerns this would lead to a new surge. Between May 9 and May 22, another 15 states lifted stay-at-home orders and other restrictions. Most of the remaining 13 states have just recently relaxed restrictions or are about to do so.


In this post, I would like to explore "What has happened after reopening?", looking across all states to see what the data say.


I will focus this analysis on cases, since they are the earliest measure of viral spread we can get. The lag from infection to death is typically 4-6 weeks or even more, which is too long to use as a leading edge measure of surge. Hospitalizations are less lagged, perhaps 3 weeks from infection, but good data on hospitalizations is lacking for many municipalities.


There are many caveats to measuring cases as well. One is that these data only measure "confirmed cases" based on viral RNA tests, which are typically not given until after symptoms are observed and do not give instantaneous results. Thus, there is also a lag between positive test results (when most states record a case), and the initial infection, that I estimate to be 2 weeks.


Second, cases for a given municipality directly depend upon their testing rate. Most states have ramped up their testing efforts, as they should, over time. Proper analysis of case increases must also take into account testing increases. For example, a doubling of cases/day in a given month, if accompanied by a doubling of tests/day in that month, may not indicate a surge of infections, but may correspond to a stable plateau of daily cases with the increased testing simply capturing a higher proportion of the infections as confirmed cases.


Third, daily testing results are extremely noisy, with uneven reporting of testing and cases across different days of the week. For example, some states may not report cases as much on the weekend leading to lulls on weekend days and a spike on Mondays. Smoothing the testing and case counts with a 7-day moving average would remove many of these artifacts.


Based on these principles, here are the details of my analysis:

  1. The analysis will be based on daily testing and case counts from covidtracking.com which I believe to be the most rigorous source of state-level data in the USA.

  2. The "reopening dates" I used were based on the New York Times site.

  3. I will compute the growth of confirmed cases by taking the ratio of the daily case counts of June 9 to the daily case counts on the date 14 days after opening/lifting restrictions.

  4. I will compute the growth in testing by taking the ratio of daily testing counts from June 9 to the daily testing counts from the date 14 days after opening/lifting restrictions.

  5. To adjust for the noisy day effects, I will compute these ratios based on the 7-day moving average of these case counts, computed as part of our web-based graphing application for covid tracking data.

  6. I will also compute a "ratio of ratios", taking the ratio of the case growth rate in 2 to the testing growth rate in 3. This is an attempt to adjust for changes in the testing rate. If this number is <1, this suggests that the infection rate has declined, if ~1 it suggests a stead state, and if >1 this suggests a potential surge in infections. This relates to the testing positivity rate.

There are some caveats given measuring practices in some states that I will comment on individually. I will now show and discuss results of the analysis. A few points:

  • I will split the analysis into "early openers", states lifting restrictions by the first week in May, "moderate openers", states that lifted restrictions sometime in may, and "late openers", including states that opened since the last few days of May or that have yet to open.

  • Within each set of states, I will list states that appear to have infection rates that are declining, stable, or increasing, and those with practically no infections.


NOTE: The tables of results are much better viewable on a computer screen, not on a phone -- sorry about that if like me you read a lot of stuff on your phone!


Early Opening States:


The 22 early opening states are mostly "red" states with Republican governors, but also include "blue" states with Democratic governors Colorado, Kansas, and Rhode Island.


Declining Infection Rates

First here are the states that appear to have declining infection rates since opening:


Note:

Open=opening date+14 days

Now=June 9 (or 8 if that is most recent data)

Adjusted Infection ratio = case ratio/testing ratio

Date Cases Testing Adjusted

State Opened Open Now Ratio Open Now Ratio Infection Ratio

Rhode Island 5/8 217 83 0.38 2588 3047 1.18 0.32

Kansas 5/3 175 91 0.52 1840 3467 1.88 0.28

Colorado 4/26 425 229 0.54 3379 4479 1.33 0.41

Nebraska 5/4* 293 201 0.69 3149 2342 0.74 0.93

Indiana 5/4 511 399 0.78 5546 6210 1.12 0.70

Oklahoma 4/24* 104 96 0.92 3798 4197 1.11 0.83

South Dakota 4/20* 58 65 1.12 344 1759 5.11 0.22


All of these states experienced a marked decline in daily case rate since reopening, most accompanied by an increase in testing.


Some caveats and comments:

  • Nebraska never had a stay at home order, but they lifted restrictions on restaurants and salons in some areas starting 5/4

  • Oklahoma never had a stay at home order, but reopened salons on 4/24

  • South Dakota never had a stay at home order, but opened non-essential businesses on 4/20.

  • South Dakota had a slight increase in counts, which were very low across the states, but I included in the declining states since its testing increased steeply during that time.

  • Nebraska had a slight reduction in daily testing rate of similar magnitude as its reduction in cases, so could be classified in the next group of "stable infections"

  • Note that there are evidence of county-specific surges in some states, including Indiana, Nebraska and South Dakota, many apparently driven by cluster infections at institutions such as meat packing plants or prisons, that do not show up in the state-level aggregations.

Stable Infection Rates

These states demonstrate relatively stable infection rates since opening


Date Cases Testing Adjusted

State Opened Open Now Ratio Open Now Ratio Infection Ratio

Mississippi 4/27 243 288 1.19 2789 4565 1.65 0.73

Georgia 4/24 691 752 1.09* 8457 11842 1.40*** 0.78

Texas 4/30 1208 1526 1.26 16834 23019 1.37 0.92

Alabama 4/30 296 397 1.34 3830 6277 1.64 0.82

Missouri 5/3 135 191 1.41 4387 5752 1.31 1.08

Florida 5/4 780 1222 1.57 16479 29973 1.82 0.86


These states contain many of the states most discussed in the media, including Georgia, Texas, and Florida. All of these have experienced increase in daily case numbers, but these increases have been accompanied by a commensurate increase in daily testing rate, so it is not clear that they should be considered surges. But they also have clearly not declined in their daily infection rates, either.


Some caveats and comments:

  • Mississippi could have been in the first group of decliners based on their adjusted infection ratio, but I kept it in this group.

  • Regarding Georgia: Unlike most states that count a case on day of positive test, Georgia attributes a count to the date of first symptoms, and if not available date sample was taken for test, or if not available date of positive test. Thus, the most recent 14 days of data may not be complete, which is transparently acknowledged on their web page. Thus, for Georgia I measure counts from opening until 2 weeks ago, 5/25. Also, I have questions about the full accuracy of Georgia’s testing numbers -- Georgia at some point was conflating serology/antibody tests with PCR viral tests. The large dip of 74,000 in daily tests on 5/27 appears a response to correct this error, but it is not clear to me if this fixed the error. However, looking at the plot of case counts over time in the many weeks since opening, it seems clear there is no surge in Georgia to date, and stable infection rates seems to be the accurate determination.

  • Texas is a state that is important to look at county-by-county given its size and diversity. There is evidence of an increase in infections in Dallas and Houston areas, which is supported by growing covid-19 related hospitalizations increasing, but the data in the rest of the state results in an increase in line with the increase in testing. Also noteworthy is that there appear to be sharp increases since Memorial Day in El Paso, Galveston, and Brownsville in the southeast that are noteworthy.

  • Some Alabama towns including Montgomery and Mobile showed evidence of increase in May, but have subsequently subsided.

  • For Florida, since opening the daily increases in cases have been matched or exceeded by the daily increases in testing, suggesting a stable infection rate. However, there is a sharp increase in cases starting June 1st, which is about the right timing for a "Memorial Day weekend" surge. If we split the data at June 1st, we see from 14-days post-opening to June 1st, the case ratio is 0.93, testing ratio 1.14, and adjusted infection ratio of 0.81, clearly showing stable infections. The case ratio from June 1st-9th has shot up to 1.57. The testing ratio during that time period also shot up to 1.57 yielding an adjusted infection ratio is 1.00, so maybe it is still stable and an artifact of increase testing, but this bears watching.

Growing Infection Rates

These states demonstrate show growth in daily case rates that exceed the growth in testing rate, and are cause for concern.

Date Cases Testing Adjusted

State Opened Open Now Ratio Open Now Ratio Infection Ratio

Tennessee 4/30 371 457 1.23 9427 8437 0.89 1.38

South Carolina 5/4 184 402 2.18 6729 4172 0.62 3.52

Utah 5/1* 142 337 2.37 3504 3581 1.02 2.32

Arkansas 5/6* 120 377 3.14 3292 4639 1.41 2.23

These are the states that bear watching. While their statewide case counts are relatively low, they show evidence of increasing, clear evidence in some cases.


Some caveats and comments:

  • Of these states, Tennessee has been opened the longest and has a relatively mild increase of daily case rates during that time, so is somewhat stable. However, given that the state has not increased the testing rate over the past month, I include it in this group.

  • South Carolina is perhaps the most alarming case in this analysis. I haven't heard much discussion of this state, but not only does it have a greater than doubling in daily case rate, it also has not increased its testing, with stable testing numbers since 5/15. The daily testing numbers are all over the place for South Carolina (look at numbers with no moving average smoothing), which produces outliers that may have a strong effect on the moving average and testing ratio, but this is one to follow for sure.

  • Arkansas has also showed a stead increase in daily case counts -- their testing has also ramped up, but the case counts have clearly grown faster than the testing rate indicating a likely increase in infection rate. By the way Arkansas never locked down, but closed salons, theaters, etc., which were reopened on 5/1.

  • Utah is an interesting case as well. They never instituted a stay-at-home order, but closed restaurants, salons, gyms, etc. which were then reopened starting 5/1. Like Florida, they were clearly stable for a time before experiencing a potential surge. There is a clear change point on 5/27 for this state. From 5/15-5/27, the case ratio was 1.0 and test ratio was 0.85, for an adjusted infection ratio of 1.17. Clearly stable. However, from 5/27-6/9, the case ratio is 2.37, testing ratio 1.20, and an adjusted infection ratio of 1.98. I'm not sure what is going on in Utah, but as I will discuss below, this coincides with a potential surge in Arizona and New Mexico and even El Paso, TX, indicating some potential "southwestern desert" surge starting at the end of May. Also, there is word of a large meat packing plant outbreak in Utah that might account for a part of this mini-surge.

Low Case Counts

These states have never really had many cases at all, with case counts never above 50.


Date Cases Testing Adjusted

State Opened Open Now Ratio Open Now Ratio Infection Ratio

North Dakota 5/1* 48 36 0.75 1316 1003 0.76 0.98

Wyoming 5/1* 10.3 8.6 0.83 522 656 1.26 0.66

West Virginia 5/3 18.6 18.3 0.98 1470 1970 1.34 0.73

Montana 4/26 0.7 4.4 6.29 1109 1567 1.41 4.45

Alaska 4/24 1.9 12.3 6.47 721 1530 2.12 3.07


These states have such low counts there is not much to say about them. North Dakota and Wyoming never issued stay at home orders, but lifted restrictions on businesses on 5/1.

CONCLUSIONS:

Overall, it appears the opening has been a relative success. They have not produced and immediate, broad surge in cases as some people feared. There has not been a resumption of exponential growth that some might expect based on their reading of the Imperial college report. Most states have had relatively stable infection rates, with some continuing to decline after lockdowns lifted. There are a few states showing evidence of a possible surge beginning, including South Carolina and Arkansas, as well as some states to watch including Tennessee, Florida and Texas. Utah also bears watching although their surge started 4 weeks after lifting restrictions, so seems like it should have some other hitherto unknown explanation. The low case counts in many of these states may have contributed to their lack of increase.



Moderate Opening States:

The 15 states that opened between May 9 and May 22 are mostly "blue" states with Democratic governors, with only 3 exceptions (Maryland, Iowa, and Arizona).

Declining Infection Rates

First here are the states that appear to have declining infection rates since opening:


Date Cases Testing Adjusted

State Opened Open Now Ratio Open Now Ratio Infection Ratio

Massachusetts . 5/18 1067 389 0.36 9131 7514 0.82 0.43

Connecticut 5/20 240 171 0.71 6571 6006 0.92 0.77

Maryland 5/15* 938 675 0.72 8088 8895 1.10 0.65

New Hampshire 5/11* 78 56 0.72 1937 1825 0.94 0.76

Virginia 5/15* 1083 786 0.73 10457 8872 0.85 0.86

Wisconsin 5/13 436 342 0.78 7396 12265 1.66 0.47

Minnesota 5/17 507 431 0.85 6978 14606 2.09 0.40

All of these states experienced a decline in daily case rate since reopening, many accompanied by a decrease in testing.


Some caveats and comments:

  • Massachusetts has experienced a dramatic reduction in daily case counts even since reopening. It is clearly under control in that state.

  • Maryland opened the state of 5/15, but Baltimore was not opened until 5/29

  • New Hampshire opened businesses, retail and salons on 5/11, and opened restaurants for outdoor seating 5/18.

  • Virginia started to open salons, restaurants, and outdoor activities on 5/15, but the stay at home order doesn't expire until 6/10.

  • Some states, including Virginia and Wisconsin, have experienced potential surges in certain counties, but these were not enough to suggest a surge state-wide.

  • Wisconsin has been a political lightening rod with the intra-state political battle leading to the state supreme court effectively opening the state, but this does not appear to have led to any surge to date.

Stable Infection Rates

These states demonstrate relatively stable infection rates since opening.


Date Cases Testing Adjusted

State Opened Open Now Ratio Open Now Ratio Infection Ratio

Iowa 5/15* 346 317 0.92 3867 4862 1.26 0.73

Kentucky 5/20* 190 217 1.14 6492 7081 1.09* 1.05

Nevada 5/9 148 171 1.16 3559 4500 1.26 0.92

North Carolina 5/22 821 1038 1.26 11799 14399 1.22 1.03

Louisiana 5/15 268 409 1.53 6174 8648 1.40 1.09

All of these states have experienced increase in daily case numbers, but these increases have been accompanied by a commensurate increase in daily testing rate, so it is not clear that they should be considered surges. But they also have clearly not declined in their daily infection rates, either.


Some caveats and comments:

  • Iowa never instituted stay-at-home orders, but started reopening businesses, restaurants, houses of worship, and salons in 77/93 counties starting 5/1 and all counties by 5/15.

  • Kentucky instituted "healthy at home" suggestions accompanied by closures of nonessential businesses, houses of worship, and salons, opening these businesses starting 5/20.

  • North Carolina has experienced steady linear growth in daily case counts since March, but these have increased in concordance with a commensurate increase in daily testing rates, indicating a plateau of stable disease. For North Carolina, since the 14-day mark post-opening is 6/6, leaving only 2-3 days to assess growth, we computed ratios for the past 7 days, which means for them we used a baseline number is 10-days post-closing instead of 14 days post closing.

  • Louisiana has shown evidence of some counties with problematic increases in cases, but by and large the increase in cases has tracked with increases in daily testing rate.


Growing Infection Rates

The only state in this cohort showing a clearly increasing infection rate is Arizona.


Date Cases Testing Adjusted

State Opened Open Now Ratio Open Now Ratio Infection Ratio

Arizona 5/15 408 1007 2.47 4830 7911 1.64 1.51


Comments:

  • Arizona has shown a stark change point and increase in cases starting about 5/27, coinciding with the lifting of lockdowns. This steady growth in daily case counts is accompanied by an increase in daily testing, but the increase in cases is clearly at a faster rate than the increase in testing, indicating a surge in infection rate. This may be the #1 hot spot in the USA right now.

  • Hot spots include, but are not limited to, counties in the Navajo nation in the northeastern part of the state.

  • It is unclear why Arizona would experience this at lifting lockdowns while other states have not, but there seems to be a clear phenomenon starting on 5/27 affecting case counts in the "desert southwest", most notably Arizona, but also including Utah, New Mexico, and El Paso, TX which is in that region. It is unclear what is causing this surge. There are wildfires in the area -- could they be irritating the respiratory tract of people, making them more susceptible to deep infection?


Low Case Counts

These states have never really had many cases at all, with case counts never above 50.

Date Cases Testing Adjusted

State Opened Open Now Ratio Open Now Ratio Infection Ratio

Maine 5/1 33 33 1.00 1957 1350 0.69 1.45

Vermont 5/15 3.2 13.7 4.28* 879 1087 1.24 3.46


These states have such low counts there is not much to say about them. Note that for Maine, personal care facilities opened 5/1, rural areas opened 5/11, and statewide on 6/1.


CONCLUSIONS:

Not much to see from these states, except the disturbing surge currently ongoing in Arizona.



Recently Opened/Not Yet Opened States:


The remaining 13 states (and one territory, Puerto Rico) have either just opened since 5/29, or have not opened yet. These include primarily "blue" states with Democratic governors, with the only exceptions Ohio and Idaho.


For these states, I compute the change in cases and testing ratios using the past 14 days, from May 27 through June 9 to give a sense of recent trends.


Declining Infection Rates

First here are the states that appear to have declining infection rates in the past two weeks:

Date Cases Testing Adjusted

State Opened Open Now Ratio Open Now Ratio Infection Ratio

Delaware 5/31** 147 48 0.33 1540 956 0.62 0.53

Illinois 5/29** 2219 1013 0.46 23760 20743 0.87 0.53

New Jersey 6/15** 1008 464 0.46 16890 25546 1.51 0.30

New York 5/28** 1628 920 0.57 42842 62579 1.46 0.39

Washington DC 5/29 129 84 0.65 383 1350 3.52 0.18

Ohio 5/29** 575 401 0.70 8355 11497 1.38 0.52

Pennsylvania 6/4 733 507 0.69 8929 9061 1.01 0.68


This group includes some of the most populous states in the broad northeast and midwest. In most places, counts are way down, and this area also includes states with leading per capita testing rates. All of these states experienced a decline in daily case rate in the past two weeks, many accompanied by a decrease in testing.

Some caveats and comments:

  • The New York-New Jersey epidemic has seeded most of the USA SARS-CoV-2 cases, and these two states have accounted for >25% of all USA cases and >1/3 of all USA deaths. Their cases peaked in mid-April and have continued to steadily decline since then, with daily case counts down 8-10 fold from the peak. The epidemic is well under control.

  • Illinois peaked in early May, and has steadily decreased since then.

  • Washington DC has kept their infections largely under control

  • Ohio and Pennsylvania also seem to be steadily declining.

Stable Infection Rates

These states demonstrate relatively stable infection rates since opening.

Date Cases Testing Adjusted

State Opened Open Now Ratio Open Now Ratio Infection Ratio

Washington 5/31 208 166 0.80 4849 2699 0.56 1.43

California 5/12 2134 2597 1.22 51011 59236 1.16 1.05

Michigan 6/1 423 1038 2.45 8955 22840 2.55 0.96

All of these states have experienced increase in daily case numbers, but these increases have been accompanied by a commensurate increase in daily testing rate, so it is not clear that they should be considered surges. All three of these cases are interesting.


Some caveats and comments:

  • Washington experienced the initial introduction of SARS-CoV-2 from Wuhan, but never surged the way New York did. The state has maintained relatively stable and slightly declining daily cases since mid-April, also with relatively stable testing rates. The June 9th moving average of test counts are strongly influenced by 3 consecutive days of very low counts which inflated the adjusted infection ratio. But looking at the graphs, it seems like "stable infection rate" is the right classification.

  • California is an interesting case, as there has been a steady linear increase in daily case counts since late March, that has been matched by a similarly steady linear increase in daily testing counts. There is some evidence of trouble spots in some counties, but by and large California seems to have a stable, not growing, and not shrinking, infection rate.

  • Michigan is also an interesting state, since behind New York and New Jersey it has probably been the state most devastated by the crisis, and the manner in which its governor rolled out and communicated its lockdown caused some controversy on the right. The daily case counts and deaths have steadily declined since the peak in mid-April, accompanied by a 7-8 fold increase in testing. However, since June 4th, there has been a sudden 3-fold jump in daily cases in the past 5 days, accompanied by a 2-fold jump in testing. This bears watching and begs explanations. Some might say it is a "post-opening surge", but this does not hold water -- Michigan's stay-at-home order does not expire until June 12, and while the governor opened some businesses and restaurants on June 1 and all throughout the state by June 8, it seems unlikely that this would have produced a surge of infections that led to symptoms, testing, and test results by June 4. So it will be interesting to see if it continues and seek explanations.


Growing Infection Rates

Of the states just recently or not yet lifting restrictions, there are three showing some problematic indications of a potential surge.

Date Cases Testing Adjusted

State Opened Open Now Ratio Open Now Ratio Infection Ratio

New Mexico 5/31 133 180 1.35 5677 5197 0.92 1.47

Oregon 6/5 37 93 2.51 2363 2773 1.17 2.14

Puerto Rico 6/15 79 178 2.25


These places all have relatively low case counts, but indicate upticks that could suggest concern is warranted.

Comments:

  • New Mexico lifted its stay-at-home order on May 31, but had allowed retailers, offices, and houses of worship to open at limited capacity on May 16. Many of the cases have been clustered in the 3 northwest counties on and near the Navajo nation. Statewide, New Mexico has seen a steady increase in daily cases that have roughly coincided with a commensurate increase in testing for the past two months. However, in the past week or so, there is evidence of a possible uptick. The timing coincides with the uptick in Arizona and Utah, and may be related.

  • The case counts in Oregon have remained relatively stable over time since April, with evidence of a decrease in daily case counts in late May even with increased testing. However, starting late May, there has been a steady increase that is clearly beyond what is explained by the increase in testing rate. This also coincides with the Arizona and Utah recent increases.

  • Puerto Rico does not have testing data posted, but has shown a sharp increase in daily cases in the past two weeks and bears watching.


Low Case Counts

These states have never really had many cases at all, with case counts never above 50.


Date Cases Testing Adjusted

State Opened Open Now Ratio Open Now Ratio Infection Ratio

Hawaii 5/31 0.4 3.4 8.50 764 936 1.23 6.94

Idaho 5/30 32 40 1.25 636 1363 2.14 0.58

These are included for completeness. Nothing to see here ... move along.

CONCLUSIONS:

As many of the most populous and covid-affected states begin to open, it will be important to watch them. Given their population and relatively high case counts, they may be at most risk of reserving, so will have to be watched carefully. Most of these have steadily declining rates, or at least stable infection rates, with New Mexico and Oregon part of the western states that show some concern with recent sudden increases and Puerto Rico also important to watch


OVERALL CONCLUSIONS AND DISCUSSION

So far, there is very little evidence that lifting of lockdowns has produced a surge of infections. The only states that show some potentially significant growth in infection include some southern states (Arkansas, South Carolina, perhaps Tennessee but notably not Georgia) and some desert southwestern (Arizona, New Mexico, Utah, plus El Paso in Texas) and pacific (Oregon) states. Some specific regions within states, most notably Dallas and Houston metro areas, show concern for growing infection rates even if the aggregated state-wide numbers do not. Most places have experienced continuing declines, a steady-state of daily counts neither increasing or decreasing, or increases that are commensurate with increased testing. It has clearly not yet been the disaster that some predicted.


Some additional caveats, however:

  • Most of the most populated states, including New York, New Jersey, Pennsylvania, Illinois, Michigan, Ohio, and California, are just opening or just about to open. The large population, more densely packed cities in some regions may put them at risk for a surge. So this bears watching.

  • While state governments have lifted lockdowns, that does not mean people automatically and immediately return to previous behavior. These is evidence from cell phone data that in many locations, people are still practicing social distancing even after lockdowns are lifted. This varies from place to place as I'll comment on later. But if over time, people let down their guards and stop practicing these measures, there is a chance ofa. delayed surge. Thus, we need to remain watchful and careful.

  • While not leading to dramatic surges, many states seem to have reached a "steady-state" plateau with constant daily cases. If they remain and don't decline, this could put them at higher risk for a surge in the fall.

  • As mentioned in the introductions, confirmed cases are only one measure of infection and not the most important one. Hospitalizations are key, and we have seen increases in covid-related hospitalizations in Arizona, Dallas, Houston, and other places. This is one way to look for confirmation of whether an apparent infection surge in the cases is real or not.

  • Hospital capacity is still a key factor. Initially, the great fear of overwhelmed hospitals related to ventilator availability, but now that we have learned these are not the primary treatment tool they were initially thought to be, the threshold before hospitals are overwhelmed is higher than initially thought. However, in some places, especially rural areas, the hospital capacity is not great and even a minor surge can overwhelm the system. So these areas need to be watched carefully, and state governments need to be ready to coordinate aide to these hospitals.

  • There is some concern that there may be a surge induced from relaxed social distancing during Memorial Day weekend and the George Floyd-inspired protests and demonstrations -- the latter would not have produced cases yet. However, from looking at this week's data, there is no evidence of a Memorial Day surge around Lake Michigan, the Atlantic coast, or the Gulf Coast from Biloxi to Pensacola. The only place with potential Memorial Day effects include possibly Galveston TX and maybe Florida, although see the caveats above -- the Florida increase may be explained by increased testing. This is possibly due to the fact that most gatherings were outdoors in heat and humidity, and maybe that was enough to stifle viral spread. If so, then maybe there WON'T be a Floyd-inspired protest based surge.

These caveats aside, why have we not generally seen the surges many predicted when lockdowns were lifted? My opinion is suggested in my comments in the introduction. I think that many of the people who feared massive surges were thinking that all of the suppression of viral spread we had observed during the lockdowns were CAUSED by the lockdowns. That is, to them, maybe they were thinking that the "do nothing" counterfactuals considered in the Imperial College report and the recent Nature paper producing the unabated exponential growth were what would happen sans lockdowns. They didn't consider that maybe even without lockdowns people would take steps that would mitigate viral spread.


The premise in the Imperial college report (and Nature paper), or at least the premise that many in the public seem to infer from it, is that the relevant counterfactual in assessing the causal effect of strong suppression or lockdown strategies is a "do nothing strategy". This counterfactual is purported to consider what would have happened had the government proposed no interventions, but importantly also implicitly assumes that people in society will not change their behavior and take any viral mitigation steps without government edicts. While I can see the theoretical benefit of imagining what would happen in a world where everyone just ignored the virus completely and continued on their normal lives as infections, hospitalizations, and deaths from the virus accrued, I strongly believe that it is NOT the relevant counterfactual against which to assess lockdowns, or to make policy decisions, or to accurately predict what would happen once lockdowns were lifted.


This supposition leads to silly statements like Trump's claim in an April 16th press conference that the lockdowns "he instituted" have "saved 2 million lives", and leads to onerous claims that we have no choice but to keep lockdowns in place until a vaccine is available, or universal testing and tracing, or other extremely high bars. It keeps people from considering or trying to assess what damage lockdowns cause and whether they are worth it. It keeps from considering carefully whether we have learned enough to craft targeted mitigation strategies that retain much of the viral suppression effect of lockdowns but with FAR less collateral damage to the rest of society. This collateral damage is to date largely unmeasured and uncertain, but clearly enormous, and disproportionally affects the most vulnerable in society. Adoption of this mindset clouds clear thinking, in my opinion.


I think that these data I have presented here looking at what has happened since states lifted their lockdowns should be considered a refutation of that supposition, as should data from Sweden (which I will comment on in a future post).


It seems clear to me that most people understand this virus is serious, and want to take steps to avoid getting infected or unknowingly infected more vulnerable loved ones. Initially we knew very little of the virus and thus the conservative lockdown strategies, but in the first month of the pandemic as we have learned more about how the virus spreads and what steps we can take to mitigate its spread, primarily basic social distancing, especially avoiding large crowds, especially indoors, especially with loud singing or shouting, and wearing masks when avoidably near people. Even upon lifting lockdowns, most people will still practice these basic mitigative steps, and thus even without government restrictions, the viral spread is decreased, suppressed far more than the unabated exponential growth in the theoretical disease spread models. I believe this is why we have not seen surges in the US, why we have not seen extreme surges in Sweden, and why we would not have realized the predicted apocalyptic results predicted in the original Imperial College model.


This individual behavior may explain some of the variability in viral spread after lifting of lockdowns -- in general, places that have continued to practice a level of social distancing after lockdowns have been lifted have done a little better, and I also suspect that places with higher levels of public mask-wearing may be doing better as well, although to date it has been elusive to find data or studies measuring this. Individuals in certain states and cities tend to be taking basic precautions better than others, perhaps influenced by cultural differences in view of government and degree of individualism.


Thus, I call us all to careful thinking. Let's look carefully at the data that comes in, evaluate it, and try to identify the key targeted mitigation steps that can be effective in limiting spread of the virus while we feverishly try to develop vaccines and treatment strategies and characterize population spread through serology/antibody tests. And our goal should be to figure out whether we can identify certain steps that, if followed, will sufficiently limit the viral spread while keeping schools open, businesses open, people at work and living their lives -- that is, strategies that avoid the devastating collatoral damage caused by societal-level lockdowns (which I am not sure our country can bear if imposed again). If we can identify, agree upon, and clearly and transparently communicate these strategies, hopefully we can achieve a reasonably high level of compliance to them.


The key question is next time we see a potential surge, what do municipalities do? Do they reflexively shut everything down again, or can they feel confident more targeted mitigation strategies would be sufficient? This will be be essential to have our minds right as we move into the fall, which will bring cooler and drier weather and increased chance for a surge of infections.



©2020 by Covid Data Science. Proudly created with Wix.com