How to know which numbers to track for Covid-19
Knowing if something should be treated like a puzzle or a mystery was the subject of a previous edition of Business Unusual and used the example of the US CIA in working with the intelligence they collect.
In a puzzle scenario, you need all the pieces of information to make a picture and typically some are missing. Finding the missing info allows you to solve the problem.
A mystery, however, has all the information you need, in fact, it has more than all the information you need, it includes information that may be wrong or distracting.
We tend to see most problems as puzzles when they are more often mysteries.
May 2020 marks the 200-year-anniversary of Florence Nightingale’s birth, she is correctly associated with founding the modern nursing profession but her work with statistics has had a much greater impact and is being used now to understand how best to respond to this pandemic.
The challenge with this novel coronavirus is that we do still need to learn a lot more about it which makes this a puzzle problem, but it is also a mystery as everyone from experts to citizens are not sure exactly which sets of information would prove to be the most helpful to both understand what is happening and the to determine what the best course of action would be.
Is it about testing or deaths or both?
US President Donald Trump, following harsh criticism for not doing enough testing, shared a graph of the total number of tests for Sars-Cov-2 in a list of countries including those that were hardest hit. The graph showed that the US has done the most tests of any country in the World. The graph was accurate but did not address the issue his critics had raised. There are many more Americans than Italians or Spanish or British, so while his graph was accurate it was not reflecting the key test, how many tests were done in relation to the size of the population. That is a very different looking graph and still, that assumes you only need to test someone once for the virus when there is a risk every day that they could be infected so a more accurate reflection of how adequate testing is would be daily tests per 1000 population and in that case, President Trump would need to report that America is able to test less than one person per thousand population. South Africa, you will notice tests even less.
For the constant criticism from the Do Nothing Democrats and their Fake News partners, here is the newest chart on our great testing “miracle" compared to other countries. Dems and LameStream Media should be proud of the USA, instead of always ripping us down! pic.twitter.com/8AwnPCNchF— Donald J. Trump (@realDonaldTrump) May 5, 2020
So, do we need more testing? It depends, we know that until we have a vaccine or most people have got and recovered from the virus it will continue to spread. We can’t actually stop it and the best efforts to slow it also do significant harm to the economy.
Total tests per 1000
Total tests per 1000 per day
NICD testing total testing numbers per 100 000 in May 2020
A potential set of numbers to track
The objective is to prevent as many deaths as possible while doing the least harm to the economy. The best way to prevent death is to give everyone that needs the most critical care access to the best facilities we have. There is a finite amount of space to treat those critically ill and it is determined by taking the total numbers of critical care space and subtracting the typical number needed to deal with all the current life-threatening situations.
Total high care beds less already used high care beds equals the capacity to deal with Covid-19 cases.
From that number, we need to determine the percentage of infected people that will need high care. For this illustration, we can say we have 100 high beds and that 10% of those infected will need high care. If there are 1 000 infections then we will use the full capacity.
Now we need to determine how to determine when and if we can maintain infections at 1 000 or less.
If you count how many tests you do each day as a percentage of a 1000 population and you set the target to be 1 per 1 000 thousand, then you will be making a big assumption about how many people actually have the infection, but if you compare how many each day tested positive from the total tests per day then a low percentage would indicate a more accurate indication that a high percentage.
Let’s assume we find 10% of cases per day are positive for Sars-Cov-2 then it suggests that one test per 1000 is giving a fairly good indication of how many cases there may actually be.
If the percentage of positive cases per day is increasing then we can assume the virus is spreading faster and that we should do more to limit the spread and avoid exceeding the capacity to deal with those with a severe reaction.
If the numbers are stable or declining it suggests the spread is slowing or stable and we can consider options to reduce the lockdown restrictions.
At the same time, we should track the deaths as a percentage of positive tests to determine if the ratio we assumed at the beginning is accurate. Should the number of deaths be higher than assumed we need to do more to lower the number of infections, should it be lower we can consider reducing the lockdown.
Is everywhere the same?
The scenario above assumed the whole country was the same when it is likely very different at least in the early stages. In early May, the Western Cape had more than 50% of the total infections in the country and Cape Town had over 90% of the total cases in the Province.
This suggests that the efforts to reduce the number of cases in Cape Town should be the highest priority while the potential to ease the restrictions in other parts might make good sense.
But this is only true if there is confidence in the number of positive cases in relation to the actual number of cases, this is the next challenge.
The best way to test
Tests take time and cost money and there is a limit to how many tests can be completed every day.
If your capacity to test was limited to one per 1000 per day, then how should you decide on which person to test.
You could make it a random test, but 1 per 1000 might be too low to make a random test accurate, instead, you could test the person that indicated a potential infection from a screening test. We can do many more screening tests in far less time and at a much lower cost. You are more likely to get a positive test from screening with the most criteria being met. This is a good option but there is another than when done in conjunction with screening could be the most effective - contact tracing.
If you screen everyone who came into contact with someone that was infected and then tested all those that met the criteria you should see more positive tests as a percentage of total tests but you should also have a more accurate reflection of actual cases.
The assumptions above focussed on what data we can collect to describe the situation as it is now (in effect the virus does not often show symptoms for five to seven days and testing times could take as much as three days or more to be returned) or rather as it was about a week ago.
The next step to make a prediction about what will happen next. This is called modelling and can be very difficult as there are some many variables to consider. For this illustration, we will use the theoretical models. The typical model focusses on the rate of spread. This is the R0 (R-nought) which is a number above or below 1. At R1 each infected person will infect one more person. This is a stable spread which if maintained will decline as more people are infected and recover. It would be great if we could achieve this, but considering in early May there were 11 000 total cases in a population with pressure to ease the lockdown it would be unlikely to not have it be above 1. You might think that if it reaches two it would also not be that bad, but it makes for a disastrous scenario. It occurs when every infected person infects two people that then infect two more. In a perfect scenario at R2 all 58 million South Africans could be infected in four months.
This is exponential growth in that it grows faster as more people become infected, looking at the current increases we are likely to see it accelerate in areas with higher concentrations before slowing and the pick up over time in areas with lower concentrations. If the areas that experience rapid increases are able to move to higher lockdown levels it should see the numbers stabilise sooner. By lowering the lockdown in areas with low numbers should doctors to respond to the worst affected areas while keeping the economy going in other parts.
This is a major simplification though and is intended to illustrate just how complex the situation may be and what factors should be considered when trying to make the best decisions. There is a risk factor to how fed up people are to adhere to restrictions on their movement but its effect will more likely be managed if we can communicate clearly what we are trying to achieve, how we are trying to achieve it and whether we are succeeding in doing so.
For now, the best thing to do is stay home, take precautions and continue to learn more about how to assess the numbers to make this less of a mystery.
One of six women chosen earlier this year may be the first to walk on the moon.Read More
Can Uber and Airbnb survive being disrupted themselves?Read More
They say a picture is worth a thousand words. If the picture is of you, it may say even more.Read More
The greatest risk to your privacy may be determined by what you willingly share.Read More
Theatres across the world have gone dark, but you can still support and watch the arts while at homeRead More
A business tool for older generations, a staple for younger onesRead More
A collection of stories about how to do business in very unusual timesRead More
Tech companies are looking to take data collection to a new level and it’s not to sell ads.Read More
SpaceX is looking to fund its trips to Mars by building an internet companyRead More
It is an issue every year, but with Covid-19 still spreading, this flu season will be an unusual one.Read More