独家优惠奖金 100% 高达 1 BTC + 180 免费旋转

Reopening USA one county at a time

The dinner table conversations across America have now moved on from the availability of ventilators and PPE equipment to relaxing restrictions and re-opening businesses. This is a great sign indeed! But, where do we go from here? I explore some indicators and interpret how that will pave the way to getting Americans back to work safely. Based on my analysis I find that there is a huge diversity. While one size fits all policy will not work for the entire country we can still find some t-shirt sizes. In this article I attempt to classify the counties based on these indicators and identify which ones can attempt to relax restrictions more aggressively than others.

My personal relationship with SARS-CoV-2 started back in February when some of my colleagues just barely made it out of China. Some others, thankfully, stalled their China trip as the country came to a grinding halt. Since then I have been following the spread of this virus closely. But the inspiration to do my own analysis started when my prodigious high-schooler cousin started his own personal thesis of what was going in the world. Another inspiration came from necessity — as I tried to figure out the best time to move my parents, who are in their 70s, from San Diego to our place in Santa Clara. Finally, in the early phases of shelter-in-place back in mid-March I was too tired of having breakfast with worldometer.com. That was it — having been in the data and analytics world for 20+ years I decided to dive headlong. I started doing my own analysis then got a little bolder, started sending those to some of my friends. Finally, my wife and friends encouraged me to write about it. And, that is the genesis of what follows. Though carefully collected and analyzed, this is still work in progress and is not peer-reviewed.

Looking at the moving average of a select few countries, as seen in the chart below, there is definitely some good news for the USA. It appears that we are well past our first peak. We had reached our peak of new cases per day on April 10th and have been on a slow decline since. It’s worth mentioning at this point that Italy, Spain and France, some of the worst-hit countries in Europe, are well past their first peak and have already started relaxing restrictions. While India and Brazil are on the rise, I wish them good luck in their battles with this invisible enemy.

This is a slight digression from the main topic but an extremely important one. I want to highlight that ‘Shelter-in-place’ and ‘lockdown’ just works! Also, I suspect, in the absence of vaccines, we will be faced with this question soon as we consider reopening. Epidemiologists knew this all along. And, of course, it makes sense, but I promise the following illustration should remove any doubt, from your mind permanently.

In both cases, New York City and Italy, the estimated rate of spread dropped really fast after the restrictions were introduced. In the case of New York City, the impact of restrictions is remarkable! In just under a week the estimated rate of spread went from a whopping 26 to 2.8; almost a 10x reduction. I cannot imagine the loss of life this would have cost if they were just a week late in declaring those restrictions. Or, the lives that could have been saved if this was done a week earlier.

Coming back to the main question we are looking to answer in this article — how to reopen. It is hard to imagine that the USA as a whole can have the same kind of restrictions throughout the country. Firstly, unlike some of the European countries, the USA does not have a national lockdown. In the US each state is determining its own course of action. This makes sense because each state is at a different place in its battles with this virus. The question of reopening the country as a whole is simply moot. We have to look at each state individually.

From the three charts shown below, we can see that states have a huge variance in the number of cases. New York, Illinois, California, New Jersey and Massachusetts have a large number of cases while West Virginia, Hawaii and Montana have a pretty low number of cases. Now let’s look at the 2nd chart showing the contribution of new cases from each state to the national 7 day rolling average. Here you can see that some of the aforementioned states also have a relatively high rate of new cases each day. For example, New York alone accounts for close to 10 percent of daily new cases. Illinois has raced ahead of New York on this metric just recently and by itself contributes more than 10% of new cases. Combined, New York, Illinois, California, Texas and Massachusetts account for about 40% of new cases each day. The 3rd chart reveals the trend of new cases in the top 20 states since it had 100 cases per day. Here we can see that the rate of decline of each state, even the 5 big ones, is very different. While New York is on a steep decline, new cases in Illinois are still increasing — at an alarming rate I might add.

Finally, according to one White House guideline, to relax restrictions, the state or region must have a downward trajectory of documented cases within a 14-day period. If you consider the chart below, only a few states have achieved this and one of them is New York with still more than 2500 new cases per day. While New York as a state has done an amazing job of reducing the number of new cases in the last 14 days there are still some heavy hotspots. They should tread carefully as they consider relaxing the restrictions in the state.

As an example, the chart below shows that New York City has a lot more cases than Delaware or Yates. While one might consider relaxing restrictions in Delaware, Yates or Franklin but doing the same for New York City at this point could prove to be disastrous. With this much variance in cases it’s hard to imagine that New York as a whole should reopen.

New York is not special in this regard; this variance exists in all states. So can we really reopen any state in its entirety even if they do meet that White House guideline? The problem, as you can see, is more nuanced and the state authorities will have to work with counties to consider relaxation of restrictions in specific counties. That leads us to the next question — just how might the states/counties consider reopening?

Let’s look at the readiness of the 3007 counties in the USA. Since there are so many counties, we will look at this by employing a technique called segmentation which is commonly used by marketers to determine product-customer fit. Let’s look at the states in segments based on the number of cases, 7 day rolling average and the estimated rate of spread. These metrics are necessary in determining whether a county should consider reopening, however, I do not claim this small set of metrics to be sufficient in making a conclusion.

Segment 1 — By number of cases

Grouping counties into bins by number of cases allows us to create cohorts of counties that can be analyzed further. The first chart shows that about 40% of the counties have 25 or less cases which is great news for us. One might consider reopening these counties since they have very few cases.

Since we want to know how many people will be impacted by any conclusion we make based on these segments let’s look at the population distribution in these bins using the doughnut chart shown below. We can observe that about 24 Million — only about 7% of the US population — fits into the bucket of counties with less than 25 total cases. Even if these counties are doing well on all the other metrics this is a really tiny subset of the population. However if we consider all counties with less than 100 cases, that accounts for the bins 25], (25,40] and (40,100], the total population under the curve of consideration will be about 52 Million people — that’s about 16 percent of the population; this is a reasonable number to work with. However, these counties must also be doing well on the other metrics for them to reopen so let’s look at the other segments as well.

As I was looking into the percentage of the population infected and if that made sense as another segment I bumped into some big surprises. Below are the top 20 counties by percent of population infected in that county. Shockingly in Trousdale, Tennessee the infection rate is 17%! In Dakota Nebraska more than 7% of the population is infected!! These are some unbelievable numbers. When the percentage population infected is that high you really have to ask?

I decided not to consider this as a separate segment since most counties demonstrate an infected population that is significantly less than 1%. This is great, but it was not as useful for segmentation purposes.

Segment 2 — 7 day rolling average of new cases

In the chart below you can see the segmentation of counties by their last 7 day rolling average of new cases. Once again we see over 50% of the counties have a 7 day rolling average of less than 1 new case per day. These counties are also good candidates to consider for relaxing restrictions. Conversely, I would argue that counties with over 100 new cases per day should not consider relaxing restrictions and the citizens of those counties should be made aware of the acuteness of the problem. In my opinion, 100 new cases per day is a lot and there is a grave risk that it can increase rapidly, with deadly consequences, if restrictions were relaxed. As can be seen in the case of New York enforcement of restrictions can drive these numbers down precipitously. If the residents of these counties continued on with the restrictions, in a week or two they should be able to get to a state where restrictions can be relaxed.

Now consider the total population under each of these bins as shown in the doughnut chart below. Good news is that almost 50M people fit into the bucket of counties that have a 7 day rolling of average 1 new case per day. And almost 100M people live in counties that have less than 5 new cases per day; that’s almost 1/3rd of the population in the USA. This is great news in terms of reversing the economic impact as a result of the restrictions.

Segment 3 — By the rate of spread

Lastly and most importantly we must also consider the rate of spread in each of these counties. However, it turns out the calculation of rate of spread is a very technical process. Since I don’t have a degree in epidemiology or the data to calculate it as specified in the technical documentation, I came up with a back of the envelope calculus for measuring this.

Consider the following illustration.

Day 1 : 1 person gets infected but there are no cases reported yet.

Day 7: That 1 person is sick enough to call in and get tested. Say they test positive; we now have 1 confirmed case in the community. This person is shown as a solid red ball in the illustration above. In this period he/she has infected 2 others in the community shown in a shaded orange ball.

Day 14: Those 2 are also now sick enough that they come into the hospital and get tested, we now have a total of 3 confirmed cases. However the 2 new cases have infected 2 others in the community in this period. So total new infections in this period is 4.

Day 21: The 4 newly infected have spread to 8 others and total cases have increased to 7.

I calculated the Rate of spread as follows

Number of new cases today = Total Cases — Previous day’s total cases

Rate of spread = Number of new cases today / Number of new cases 7 days ago

So the estimated rate of spread in this case 7 divided by 3 = 2.3. This can be observed in the pictorial illustration as well.

As you can see just from the total positive cases, provided by John Hopkins University, we can get an idea of the rate of spread. However, due to fluctuations in data reporting on several days, I have taken a 7 Day Rolling average of new cases as the metric, to calculate the estimated rate of spread.

Having explained the non-standard calculation of estimated rate of spread, let’s see how the various counties fare on this metric. In the chart below, once again we see 50% of the counties with a rate of spread of less than 1 as of this writing. This is great news because the number of net new cases are declining relative to 7 days ago or they have no new cases at all.

Now let’s look at the total population in counties where the rate of spread is less than 1 and therefore are reporting declining cases. Once again from the doughnut chart below we see that 63% of the population are living in counties that are seeing a decline — rate of spread less than 1. In fact, about 35% of the population is residing in counties that are seeing relatively rapid decline in rate of spread — counties where rate of spread is 0.75. Conversely, once again, I would argue any county that has more than 0.75 rate of spread should not consider relaxing restrictions, especially, the ones where the rate of spread is greater than 1 — where the number of cases are still increasing.

Perhaps you can tell by now that one single segment is not enough to know if the county overall is in a healthy shape and should reopen. For example, a county with a large number of total cases might have a very low rate of spread and a low 7 day rolling average of new cases, while another county might have a low 7 day rolling average of new cases but their rate of spread isn’t declining. We need to cross-reference counties across each of the segments to identify if they can really consider relaxing restrictions.

Once again since there are several counties I’ve segmented them into tiers based on how safe it might be to relax restrictions in each of those counties as determined by the segments created above. Here I am not considering the Total cases as a marker for this separation since that number is cumulative and will never decrease. Here is a reasonable set of thresholds that delineates the counties into tiers.

Based on our selection of thresholds we can see that the safest tier to start relaxing restrictions is Tier-1 followed by Tier-2 and Tier-3. Counties in Tier-3 and the Others category in my opinion should not be advised to relax any restrictions. The chart below demonstrates the percentage of population in each of these categories with about 30% of the population in Tier-1 and Tier-2. Also note that as the counties start to relax restrictions there may be another outbreak, as a result their indicators might start to drop to Tier-3 and vice-versa is also possible. Consequently, these indicators would need to be actively monitored by the local or federal public health agencies.

Conclusions

While the USA is well past its first peak of daily new cases, unlike the European countries, we still cannot reopen as a whole because of our size, population and most importantly variance in daily average of new cases across the 50 states. We know shelter-in-place has a dramatic effect, so as we start to relax restrictions, we can use this tool to kill any local surge in cases. States can consider the model of segmenting of counties into tiers, as they decide which counties qualify for restriction relaxation. Conversely, the model also clarifies that some counties should not be considered for reopening at this stage and the population of neighboring counties should avoid traveling to and from those counties. Finally, such a model can also be used to determine the distribution of tests as we prepare to reopen counties based on these metrics.

References

Add a comment

What is your view?

Send

Related posts:

SOL is listed on Tidex

We are happy to announce SOL is listed on Tidex exchange. You can trade SOL in two currency pairs: SOL/BTC and SOL/ETH. Tidex is a fast-growing, secure, and user-friendly exchange. It has a flat fee…

Escape

This first-ever interactive Agile Adventure is the gripping tale of an experienced team struggling with agile adoption. In this unique mashup of a business novel written in the gamebook format, you'll overcome common yet daunting challenges that come from using agile methods. As Jim, the agile coach, you'll learn to apply a rang

A Child of My Mother

Mothers may be the most important people in our lives. They give us the bodies we use to experience our lives. The contribution to our minds is harder to describe. Mother’s emotional rhythms are ours…