A Comparison Of States Reopening and Remaining Locked Down

A number of states (like Florida, Georgia, SC, Tenn, and Ohio are reopening for businesses while others like NC, Va, and Michigan remain locked down.  There's a wide variation in these states for cases, hospitalizations and fatalities:

Click to enlarge

The ones to the left have taken first steps to reopening while the ones on the right (NC, MI, VA) have not.  As we saw in yesterday's post,  NC's COVID experience so far has been far milder than many other states. 

I ran a comparison of COVID fatalities to date vs flu deaths in the 2018-2019 season and saw an interesting result:

While the US as a whole has had more COVID deaths than flu deaths as have some states like Michigan,  in other states like NC and many of those opening, the seasonal flu has accounted for many more fatalities than COVID so far. 

The COVID experience seems to be very different from state to state.  If you only read reports of the nation's experience and of your own state it seems as if both are being struck hard; but looking at the big variations between states explains how different states have taken (and will take) different approaches to managing the pandemic and reopening for business. 

An Analysis of Governor Coopers Decision to Extend the NC Stay at Home Order

Last week, I looked at the White House's Plan to Reopen America and how it applies to North Carolina.  A couple days later Governor Cooper held a briefing where he said that we needed more time to slow the spread of the virus and extended the Stay at Home order to May 8th.

In this post, we'll look at the data they shared that drove that decision and some other details that might suggest a different approach. Before I do that, I want to acknowledge that many people are anxious to get our economy up and running again and many are frustrated with Governor Cooper.  I think its important to recognize that he is in a very difficult position-- more difficult than I think he ever dreamed of experiencing when he was sworn in.  He has spent most of his life in politics and history will pretty much judge him by what he does this year. He's having to make decisions without the information you would want to feel certain you are choosing the right path.  Whether I agree with him or not, I trust he believes what he is doing is protecting the people and hospitals in our state.

The Criteria and Data
Governor Cooper looked to Secretary of Health and Human Services Dr. Mandy Cohen to lay out the criteria and data they are looking at.  Dr. Cohen used similar criteria to the ones from the White House that I described in my earlier post:  COVID Symptoms from our Flu Surveillance system,  14 Day trends in new cases and positive test %ages, and the potential load on hospitals. 

For the first criteria was about hospital visits with flu-like symptoms that would indicate potential COVID cases. 

The Red line has been trending down for more than 14 days, so the Dr. Cohen confirmed that this criteria had been met. 

The second criteria is about the number of new cases per day.  We're looking for that number to show a 2 week decline. Here is their graph from the NCDHHS dashboard site: 

As in my analysis, this trend line is curving up.  I think some additional context would help, though.  Here is our trendline of new cases compared to Georgia, Florida, and Michigan (Ga and Mi are the closest to us in population. Fla has twice our population).

While we are slightly curving up,  our curve has always been much flatter than our peers so even fairly small changes in counts - about 100 cases a day over the 14 days -- looks like s significant increase when other states are still far from achieving our daily case rate.  

The third Criteria is about the trend in positive testing.   Here, the NC DHHS offers this chart:

Although the criteria is for 14 days,  they plot the trend line across 21 days to create the dashed 7 day average line. Also they appear to be using different data on testing from that logged by the COVID tracking project, which pulls data several times a day from the DHHS site -- I've been using the COVID tracking projects' data throughout my research.  Here's what my graph looks like:

The lighter lines are the calculated trend lines.  The blue line at 10% is trending slightly down over 14 days, but the red trend line is slanted upwards. Because the positive rate was much lower 3 weeks ago, it is driving the rolling average upwards.   It's also a good idea to look at the rate of testing against the positive testing rates:

Here the positive test rate is the blue line while the number of tests is the red line.  You can see we had a major dip in test rates mid-month that we've since recovered from.  When the number of daily tests drop, tests are normally targeted to the most likely cases so the % positive also goes up, as you see it did here from April 17 to the 20th.  That means that the bump in positive testing that is driving DHHS' trendline up probably has nothing to do with the actual COVID cases but is just the effect of having a shortfall in cases. Also, the fact that our positive test rate has gone down as we've nearly doubled the number of daily tests in the past week should be a very promising sign.  

The White House recommendation is that a state should be seeing either drop in new cases or in positive testing over 14 days.  Even though NC does not meet the criteria for new cases, it would still pass if there was a decrease in positive testing rates, which depends on whether you use 21 days for the 14 day trend, and how you interpret the effect of the dip in tests. 

Before moving the the next criteria, here's a graph of our comparison states and their positive test rates for context:

Click to enlarge

North Carolina has had a consistently relatively low positive test rate throughout the COVID pandemic.

The next criteria is the ability to treat all patients without crisis care. The White House didn't define a particular measurement to gauge this ability.  NC DHHS provides this graph of current hospitalizations in NC. 

This is current hospitalizations, so as people recover and are discharged, the number will grow and shrink.  Dr. Cohen pointed out how this line is also trending upwards, indicating we have not crested the curve.  While this a fair read, but should be looked at alongside DHHS' report on hospital occupancy:

 This shows that more than 1/3 of hospital beds in the state are empty.  The 6836 beds in this graph is almost 15 times the number currently being used by COVID patients.  Even if we suddenly saw a 5x surge in new COVID hospitalizations (and the line looks like its flattening) we would still not come close to overwhelming hospitals.  If you follow the DHHS link above, you'll see that almost 4 out of 5 ventilators are also currently unused. 

Again for context, here is the hospitalization trend in other states:

Florida and Georgia report cumulative hospitalizations, not current ones. Current counts are probably about half that shown here, but we are again much lower than comparable states. 

The final criteria from the Governor is for adequate testing and PPE to meet North Carolina demands.  Testing was behind national average but is now catching up.  DHHS provides an interesting look at PPE needs and supplies for the state.  I don't have a way to analyze this data, but given Burlington NC is the headquarters of Labcorp, one of the two private companies doing most of the COVID testing in the country and that NC's long history of textiles should give it relative advantages in getting PPE and tests for the state.  For example, the Nonwovens institute at NC State is producing enough material to make 500,000 masks a day. 

So after looking at the 5 criteria, only the first one from the Flu surveillance data was judged positive enough to pass:

As I said at the beginning,  this is a difficult job and I trust Dr. Cohen is using her best judgement and the counsel of the best people she can find before making her recommendations.  I believe honest people can disagree with them, but I respect her efforts and leadership. 

The differences come from how you look at the graphs.  Governor Cooper sees lines sloping up and is insisting on more time to slow the virus.  Others may look at the graphs and data of NC compared to other states and conclude that we have already benefited from one of the slowest COVID growths in the country.  In addition to the graphs I showed above, North Carolina has one of the lowest case rates and death rates in the country (37th and 39th respectively).  Those doing better (besides TX) are less than half our size.  

Because we've had a relatively low COVID curve to begin with, I believe our leaders should take small upward growth trends in context.  Knox county Tennessee recently released a plan for opening their community.  They also have had a very flat curve and made this observation:

An Increase in Cases is Anticipated Due to Initial Low Case Counts Through the joint effort of citizens in “flattening the curve,” the number of active COVID-19 cases has stayed far below the capacity of the health care system. This has provided the region with time to increase supplies of personal protective equipment (PPE), hospital surge capacity, testing capacity and the ability to surge contact tracing should it be needed and appropriate. As we locally advance through the phases, we anticipate an increase in active cases at each phase due to the low number of initial active cases. The focus of the benchmarks outlined in this document is on assessing the local ability to manage an increase in cases while preventing the unobstructed growth of transmission. Our low initial active case counts will likely mean we will not obtain a downward trend throughout the phases of the reopening process. Our community demonstrated success in flattening the curve before it truly started. Due to this initial success, future phases will result in increased numbers of active case counts. This alone is not a reason to revert to a previous phase or not advance to the next phase.

This perspective could also apply to much of North Carolina. 

What is the White House Plan to Reopen America, and How Does It Apply To NC?

Last Week, the White House went public with its plan to reopen America.  The plan pairs a set of criteria that determines if an area is ready to start reopening and three phases of reopening once that criteria is met.   So the timeline would be:

1) Meet the criteria
2) Go to Phase 1 reopening
3) Wait at least 14 days to see if the criteria is met again
4) Go to Phase 2 reopening
5) Wait at least another 14 days and see if criteria is met
6) Go to Phase 3

The Criteria
The Criteria for reopening looks at three areas:  Signs of COVID symptoms,  Confirmed COVID cases, and the level of strain on hospitals and healthcare workers.  Let's look at how NC fares with this criteria:

COVID Symptoms
While the US performs far more COVID tests than any other country, we know there are many cases that we haven't been able to test.  However, for decades we've had systems to track the growth of the flu and most doctors and hospitals regularly report to the CDC when they encounter flu-like symptoms.  Now that the seasonal flu is largely behind us, that reporting system gives us a good look at how many people are going to the doctor with COVID-like symptoms.  There are two different reporting systems.

The FLUView system is a great way to monitor the flu in a normal year.  When doctors take in a patient with an Influenza Like Ilness (ILI), they report it to the CDC and it shows up on this chart.  The criteria is for a decline in the past 14 days.  Here's the view for North Carolina:

The graph has three peaks.  The first is Flu A, the second Flu B, and the third is COVID.  From this graph, the number of COVID cases reported at doctors' offices have been in steep descent since mid March, so the criteria here is met. 

The other system is used by hospitals to report flu-like symptoms. Its not state by state but grouped into Regions (NC is with AL, FL, Ga, KY, MS, SC, and TN).  The hospitals report on cases where the patient has COVID symptoms or - COVID-19 Like Illnesss (CLI).  The criteria calls for a 14 day decline.  Here's the graph for our region.

The graph shows visits peaking at week at (ending Mar 22) and in decline since then for at least the last 3 weeks so the criteria here is met. 

COVID Cases
The Next Criteria is for COVID cases.  This criteria looks at the number of daily new cases to see if they're in steady decline.  It also looks at the rate of positive testing to see if the % of daily positive tests is decreasing while the number of tests stays constant or is increasing.

Here is North Carolina's new daily case chart:

This is for the last 14 days.  You can see the lighter trend line is angled slightly up. Here is our testing for the last 14 days. 

This red line is the number of positive tests.  The lighter trend line shows that the positive tests are in decline.  The Blue bars are the number of tests per day.  

The criteria here calls for either a 14 day decline in positive tests <or> a decline in daily cases.  The daily cases is trending up and does not meet the criteria.  The positive testing is in decline and would satisfy the criteria, but with room for argument.  The decrease in the positive test line comes from a spike on April 7/8 that coincides with an unusually low number of tests those days-- it looks like some tests got delayed and came in on the 10-12th.   The criteria specifies that the decline in positive tests has to happen while the number of tests stayed steady or increased.  The blue trendline for testing looks very flat, so the decline should meet the criteria, but in a few days after that blip is not in the 14 day window, it may look very different.  However, technically, it appears NC meets this criteria.  

Hospitals
The third criteria has to do with hospitals and it also has two parts.  The first is to be able to treat all patients without crisis care.  NC looks to have plenty of excess capacity for hospital and ICU beds currently according to NCDHHS.

While the criteria is not very specific, it seems reasonable to conclude that NC Meets this criteria.

The final criteria is for "Robust testing program in place for at-risk healthcare workers, including emerging antibody testing".  There's no official data I can find for North Carolina for this criteria.  If anyone reading this has a sense of how we're doing on testing health care workers, please comment or let me know. Officially for the moment, I would say it is unknown if North Carolina meets this criteria. 

To sum up, here's how NC makes out with the criteria:

	Criteria	NC Status
1	Downward Trajectory of Influenza-like Illnesses (ILI) reported in 14 day period	MET
2	Downward trajectory of COVID-like syndromic (CLI) cases reported withing 14 day period	MET
3	Downward Trajectory of documented cases within a 14 day period	UNMET*
4	Downward trajectory of positive tests as a % of total tests within 14 day period (Flat or increasing volume of tests)	MET
5	Treat all patients without crisis care	MET
6	Robust testing program in place for at-risk healthcare workers, including emerging antibody testing	UNKNOWN

While number 3 is Unmet,  the criteria calls for either 3 or 4 to be met. 

My conclusion is that we're at least borderline meeting the criteria for the first phase of reopening. The details can be seen here, but it allows for social groups of up to 10, return to work in phases, and opening of restaurants and gyms.

Once we've done that, we reset the criteria and wait until they are satisfied again under phase 1 -- so it would take at least 14 days to see that COVID is still under control -- before going to phase 2. 

For the areas of NC that count on Tourism, we need to at least get to phase 2 to avoid major economic impacts, so my hope is that we see movement in the state towards phase 1 soon. That said, the Governor has access to more information that I have from public sources, and once we start opening up, it would be very hard to get people to go back to where we are now, so I understand the need for caution.  

It may be that rather than taking the whole state to phase 1, we take certain regions of the state there first.  I have been tracking COVID in NC on a county by county basis and hope to share some analysis of that soon.  

Sweden: The Country Without A Lockdown

With most of the world opting to have national lockdowns -- even countries like Rwanda with only 136 cases and no deaths to-date -- Sweden bucked the trend and instead limited gatherings to less than 50, asked people to self quarantine, wash hands and keep social distance, but didn't order people to shelter in place and businesses to close. Yesterday we looked at the question of what happens after lockdowns and I wrote about how many preventative measures would continue even after businesses re-open.  Sweden provides an intriguing look at that scenario. 

Sweden is one of the northernmost countries in Europe. With a population of 10.3 Million -- about the same as North Carolina.  Most of the population is at the southern end of the country and about 1.5 million live in their capital of Stockholm.

They put in place national warnings and social distance advisories about the same time as other European countries. They banned travel with China in Mid-February and put in place policies limiting gatherings of more than 50 on March 11, the day of their first COVID death. The recommended businesses work from home and about half the workforce soon did.  They recommended schools go to distance learning and the schools followed suit.  They advised the elderly to limit contact with others. Their COVID testing is about the same as the US on a population basis -- ranked 10 out of 20 countries on my COVID dashboard.  They pretty much took the same steps as other countries, but never put in place a legal ban to close businesses and make their people shelter at home as we have done in NC and most of the US.  It's been about 40 days since they took a different path. While there are some ways where it seems they've had similar results to other countries, there are also important differences.

When it comes to the reported cases, they seem pretty comparable, if a little lower than other North American and European countries:

When looking at fatalities, they also seem slightly lower than average.

When looked at from this perspective, it appears that lack of a national lock down did not result in a worse situation and that the other measures to avoid COVID growth -- self quarantines, social distance, and voluntary work from home -- were sufficient to contain the pandemic.  As I've noted earlier, nearly all the countries above are several days past their peak day of new COVID cases and many are also past their peak day of new COVID deaths.  Sweden also looks to be on a similar track:

Country	Days post 100 cases	Highest Case per Day	Days Since Peak Case per Day	Highest Death per Day	Days Since Peak Deaths per Day
USA	45	33,752	5	2,482	0
Italy	53	6,203	20	951	11
Spain	45	8,271	20	961	13
Germany	46	6,933	19	333	7
France	47	23,060	12	1,762	1
UK	42	8,681	5	980	5
Canada	34	1,600	10	123	1
Sweden	34	726	7	170	0

While the results look promising,  there are some important differences with Sweden's experience to consider.

First, Sweden is much smaller than the other countries compared here.  Even though the graphs are adjusted for population, density matters.  Sweden's largest city has 1.5 Million people -- much less than the largest cities in these other countries.  It may be fairer to compare Sweden with its neighbors -Denmark and Norway. Each has about half the population of Sweden, but Norway's largest city has only 500K people while Copenhagen in Denmark has 1.1 million.  Here's how they compare:

	Population	Days past 100 cases	Covid Cases	Covid Deaths	Cases per 1M	Deaths per 1M
Sweden	10,230,000	41	11,927	1,203	1,166	118
Denmark	5,806,000	37	6,681	309	1,151	53
Norway	5,368,000	41	6,790	150	1,265	28

On the number of COVID cases, Sweden is very close to its neighbors, but has far many more deaths, including 284 deaths in the past two days -- almost 20% of their total. “We have had an unfortunate development, especially compared to our neighboring countries, with an introduction of the virus at many elderly care homes,” said Sweden’s top epidemiologist Anders Tegnell.  The key oversight seems to be that many of the elderly need day to day assistance and that assistance comes from the younger population.   By asking the elderly to limit contact with others but giving more leeway for their younger caretakers to be out and about,  it seems to have exposed those most vulnerable to COVID to more risk.

As we consider how we should approach ending our lockdowns and allowing more social contact,  it will be important to consider how we can continue to protect our older people and not have them pay a higher price for our decision. 

What Happens When We End The Lockdown?

As the number of COVID cases a day has passed its peak level and countries and states start to look at how to safely end their lockdowns and closures, what can we expect?  What would happen when we send people back to schools, work, stores and restaurants?

While it’s obvious to imagine that we would just see another explosion of COVID cases, the next phase will be different from March in several important ways:

1. We know how to identify it and are doing large scale testing across the country
2. Those with mild infections are unlikely to assume it's just a minor cold and go into work, school, or social settings where they could infect others. 
3. People will continue to demand safer environments from stores and public spaces that will reduce the chance of infection.
4. People will continue to keep habits of hand washing and many will continue to wear masks while COVID is present in the community

Even if we ended all the shelter at home policies, the actions and habits I've listed would still be in practice, so we wouldn't be returning to the same situation we had before.  While it may be difficult to see the dramatic effect these efforts have had on COVID because we're still catching up with it on testing,  you can see the effect its had on the flu when compared to last year:

But how can we estimate whether these remaining factors are enough to keep COVID in check?

The way scientists measure the growth rate of the pandemic is by researching samples of COVID patients and identifying how many people on average someone infected while they were contagious.  The number is referred to as R0 (and called R naught). When R0 is greater than one, the pool of active COVID cases grows (because the infected person replaces themselves and adds more beyond that).  When R0 is less than one, it shrinks (because the infected person doesn't fully replace themselves before they recover).  The acceleration of COVID seems to have stopped since we went into lockdown and did all the other things I described above, so the current growth number should be something less than 1. If we're worried about COVID cases exploding again if we lift the stay at home, what we're worried about is that COVID's growth factor would go back to being much higher than 1.  

To understand how COVID was growing before we made an effort to stop it, we can look to research.  There have been at least 137 studies done on R0 since the epidemic in different countries (but mostly China).    You can find them (as well as a nice summary of a lot of academic research into key parameters of COVID) here.  Averaging the R0 values from the 110 studies that were completed in the last two months gets us to 2.5  (most recent studies are very close to this number).  That is, in a starting situation where there is no immunity or counter measures, a person would typically infect 2.55 other people while they were contagious.  

To consider how R0 changed after we took action, there is the lockdown, social distancing, better hygiene, and other efforts I described.  Let's look at the impact of just the factor that people who develop COVID-like symptoms will usually self-quarantine until they can get tested and will avoid making contact with people until they are tested negative or recover from COVID.  When COVID first arrived in the US, it went undiagnosed and untracked.  Since 80% or more who get it experience mild symptoms, they likely made little effort to avoid infecting others and it spread quickly.  In mid-March when the state of emergency was declared and everything got cancelled, it would be nearly impossible to be coughing and running a fever and not suspect you have COVID -- even though at least 4 out of people tested get a negative result.  So, we would expect the amount of time a contagious person makes regular contact with their community to be much lower now.  

Other studies of COVID indicate that most people caught it from someone within 5 days of them showing symptoms.  If the average person now self-quarantines after two days of symptoms, that alone would reduce the growth rate of COVID by 60%, from 2.5 to 1.0 -- That's without any social distancing, lockdowns, or Purell.  When you include the other factors that would continue after lockdown is lifted, its very plausible that COVID cases would continue to shrink going forward.

It's important to point out that I'm NOT suggesting that we could safely drop all protective actions today.  I'm merely addressing the concern that ending the lock down would put us right back where we started.  Any model that projects the result of the end of the lockdown must factor all the other things that reduced the growth rate that will remain after lockdown if it is to give our leaders a realistic sense of their choices. 

The US May Have Crested The Peak Of New Cases and Deaths

Good morning.  Looking at this graph, it appears that we are past our peak levels for new cases and deaths. 

Click to Enlarge

I've put the scale for the new death rate to the right while the scale for new cases are to the left so you can see both curves well. Both graphs use a 3 day average to smooth the curve.  It appears COVID fatalities peaked around last Thursday and have been dropping steadily since then.  The number of new cases looks flat(ish) since April 3rd, but the peak day was a couple days ago.  Case levels can be affected by daily testing levels, but while there was a spike last week, they have been fairly consistent since then.  The fatality rates are not dependent on testing rates or other things, so when that goes down, you should feel good something is happening.

This doesn't mean the pandemic is over, but it is a good sign.  New cases per day is like acceleration.  If you were in a car with a stuck gas pedal, you'd be going faster and faster. When that curve flattens, its like you aren't going any faster, but you're still going very fast.  We are still having more cases and fatalities per day than any other country -- our car is going very fast -- but we've stopped going faster every day and are starting to decelerate. 

As it turns out, most of the other countries I'm looking at have also peaked in cases and fatalities.  I've also added a couple columns to the Current Status tab in my spreadsheet that look at other countries. 

Country	Days post 100 cases	Highest Case per Day	Days Since Peak Case per Day	Highest Death per Day	Days Since Peak Deaths per Day
USA	42	33752	2	2056	2
China	85	143	20	38	40
Italy	50	6203	17	951	8
South Korea	53	600	41	9	18
Spain	42	8271	17	961	10
Germany	43	6933	16	333	4
France	44	23060	9	1417	5
UK	39	8681	2	980	2
Canada	31	1600	7	84	1
India	31	877	2	49	3
Indonesia	28	399	0	79	2
Philippines	30	538	12	50	0
Vietnam	22	20	14	N/A	N/A
Thailand	28	143	14	4	9
Malaysia	35	258	10	8	14
Singapore	42	287	3	23	3
South Africa	27	243	16	24	2
Argentina	25	234	12	14	3
Australia	33	536	19	198	5
Chile	28	529	2	9	2
New Zealand	22	146	15	2	2

Two days is not long to be past a peak.  The US curve has been flat, but Canada, the UK, and others with 2 or less days may still be growing. The Status tab will reset to 0 if there's a new peak so feel free to check on it. In a couple days it should be even clearer that for most countries the worst of COVID is behind them for this season. 

While the US may have peaked, different locations in the US are on different timelines.  For some perspective on NC, see my last post.