COVID-19 is a science of uncertainty and an art of probability

COVID-19 illustrates the difficulties society has in dealing with uncertainties and the stresses this creates

No one can say with confidence what our COVID-9 world is going to look like over the next 12 to 24 months. The rules and models of the old world no longer hold. It is this uncertainty itself that adds great stress, overlaying the fear of specific scenarios.

Times like these can bring out true leaders but also false prophets. Trying to distinguish between them is much easier in retrospect than at the time.

All great leaders instil confidence in the chosen path but there is a fine balance between confidence and false confidence. The troops do need belief in their leaders, but the leaders also have to be willing to change course in the face of evidence.

There are two quotations from American President and Allied Commander Dwight Eisenhower that I rather like. In fact, there are many of Eisenhower’s sayings that I like, but two seem particularly relevant right now.

Eisenhower once said that “plans are useless, but planning is indispensable.” I have used that quote to my students many times over the last four decades. It is worth thinking about.

A second Eisenhower quote is that “farming looks mighty easy when your plow is a pencil and you’re a thousand miles from the corn field”.

When Eisenhower enunciated those words in 1956 in Illinois, he was indeed talking about agriculture, but the relevance is universal. It came to mind when I heard TV presenter Mark Richardson opine that he was sure he himself would be a better Prime Minister than the incumbent.

Experts and the media
One of the issues right now is that there are supposed experts on both sides of any COVID-related argument. For example, here in New Zealand we have had epidemiologists arguing both for and against various levels of lockdown and community restrictions. It can be a case of choose your preferred argument and then choose the so-called expert to buttress that argument.

Part of the problem goes back to our population being poorly educated in relation to science. Hence, our media, themselves poorly educated in science, prefer simple statements of certainty. These are then buttressed by a ‘call to authority’ that the message is coming from an ’expert’.

In addition, the popular media thrives on controversy which they themselves delight in stirring up. It is their job to do so. There can also be an issue of media taking the views of specialists out of context.

This week I was asked by a major media organisation if they could interview me for an article on COVID-19. I turned them down for the simple reason that I did not trust that specific phrases would be kept in context. And I told them so.

One of the reasons why over the last 20 years I have always preferred to do live radio and TV interviews rather than pre-recorded, is that I know with a live interview I cannot blame anyone else but myself if I stuff it up. With a pre-recorded interview, the worry all the time is how a particular phrase will come across as an isolated ‘sound-bite’ of 15 to 30 seconds.

Also, the mainstream media do not usually give their commentators the right to choose the title of an article. In that regard is a welcome exception. In contrast, the mainstream media prefer ‘click-bait’ titles even if they are not always accurate.

However, I also have some sympathy for the media folk who are trying to tease out the real story. Every Government Ministry and every business corporation has its communication team tasked with massaging the flow of information to the public. I think of them as the Propaganda Departments.

We are all learners
Coming back to COVID-19, if the definition of an expert is someone who knows things with certainty, then there are no experts. Rather, there are specialists in particular areas who can bring their professional knowledge to particular aspects of the issue. But in regard to the big questions of how this virus plays out in particular contexts, we are all very much learners.

In that regard, another quote I like comes from a former colleague and ecologist at University of Queensland. On his door was the statement: “In God we trust; all others must bring data”.

A key problem with COVID-19 is that the data is only starting to emerge. Then there is the question of how it should be interpreted and the extent of any generalisations that can be made from it. But decisions do need to be made.

Despite the lack of data, there are now more than 5000 publications specifically relating to COVID-19 already in the public arena, either as peer-reviewed articles or as preprints awaiting peer review. And then there is the stream of media releases, emanating from the propaganda departments of both Big-Pharma and universities, often with no supporting information available.

I have been trying to peruse both the preprint and peer-reviewed literature on COVID-19, but I defy anyone to keep abreast of it. I scan through the titles, and if something looks particularly interesting then I download the abstract, and then, depending on what the abstract says, I download the full paper. Inevitably, in so doing I am using my own subjective lens to identify what seems likely to be valid, and how we should interpret those results.

Models are projections based on assumptions
In relation to the models of what is likely to happen, I always focus on the underlying assumptions. Often these are not explicit. In regard to models, another quote that I have used many times over the years with both students and also in my own research projects, comes from famous British econometrician George Box. He said that “all models are wrong but some models are useful”. What he was saying was that models can provide insights as to how systems work, and what are the crucial parameters, but they don’t predict the future.

To understand something of models, one has to understand the language of mathematics, including calculus and differential equations. Scientists soon learn that use of equations and statistical methods can be a pre-requisite of publication in good journals, but it is amazing how many papers get through peer review with incorrect equations and incorrect use of statistics. It also means that very few non specialists can understand those papers because they don’t understand the language of mathematics.

The evidence is murky
From the evidence I can find, there is absolutely nothing that is proven at this time in terms of a successful treatment for COVID-19. There are a number of treatments that I am following closely and that can be described as ‘promising’, but even that term can be open to misinterpretation.

There is great confusion in the public arena together with debate amongst scientists as to the overall levels of community infection. Some overseas studies are indicative of a considerable number of asymptomatic infections, but scratch beneath the surface of those studies and lots of necessary caveats can be identified.

Often, the messages about possible community levels of infection get confused between asymptomatic cases (no symptoms ever) and pre-symptomatic cases (yet to show symptoms). Also, some of the studies that talk about a high proportion of asymptomatic infections show no understanding of what are called antibody specificity levels (false positives) which have to be taken into account.

Serology (antibody tests) can be highly informative in understanding how a pandemic is evolving, and I would like to see New Zealand getting involved with a lot more vigour than currently. But antibody tests need to be interpreted by people who understand the nuances of both sensitivity and specificity, together with how they apply in specific situations.

Using Mycoplasma bovis as an example, the New Zealand eradication program would have long-failed if it were not for antibody testing, despite lots of false positive and false negatives that occur. If there had been ongoing reliance on just swab (PCR) testing, then failure would have been certain. As the program evolved and with increasing experience, much more weight has been placed on antibody testing as a key diagnostic tool. However, whether that program will succeed is still unknown.

All developed countries have a program
Across the world, there are no developed countries that I can find that do not have an active program to reduce the transmission of COVID-19. There are some that were very slow to get started such as Brazil and they are now paying the consequences. There are others such as Sweden which have no lockdown, but which do have social distancing programs that include only seated customers at bars, the self-isolation of older people, and the cancellation of both concerts and mass sporting events. Indeed, the Swedish strategy is very similar to what New Zealand will have under Level 2.

One of the concerns for New Zealand is that the Swedish death rate on a per capita basis is now 75 times that of New Zealand. Adjusted for New Zealand’s population, the equivalent number of deaths in New Zealand would now be over 1500 with many more to come. Sweden had hoped that infections would decline as population immunity built up, but there is no convincing evidence so far that this is occurring. I see a lot more problems ahead for Sweden.

In most parts of the developed world apart from New Zealand, Australia and much of North-East Asia, there is no prospect of eliminating COVID-19 in the short or medium term unless and until a vaccine is developed. They left it too late.

There is also now a broad consensus that in most of the world we have only seen the prelude. In musical terms, the more substantial fugue is still to come. There is also a developing belief that the future waves will be multiple.

In New Zealand we are so lucky that through the combination of our isolation, the late arrival of the virus, and a lockdown that came in just the nick of time, that new cases are now very low. Hopefully, we will not squander the position we have reached.

Treading the fine line
One of the key issues is that there is a very fine line between a transmission rate below ‘1’, where each person on average infects less than one other person and with this leading to virus levels declining, and a contrasting level a little above ‘1’, such that disease levels builds again. This is one of the issues that all European countries now have to be very concerned about.

In Germany, the evidence suggests that their transmission rate declined to about 0.7, but that it may have already increased again to at least 0.9 and may even be heading beyond there above the critical level of ‘1’. Lags in the system between behaviours and confirmation of infection mean that key confirmatory information arrives too late for good decisions.

In the United States, the overall transmission rate is now declining but this is only because of the dominating effect of New York, and, to a lesser extent, cities such as Detroit and New Orleans. In two thirds of American states the number of cases is still exponentiating.

Accordingly, unless there is a strong seasonal effect with summer coming on, then big trouble still looms in many parts of the USA as restrictions are removed. Unfortunately, the evidence from Brazil, Mexico, Ecuador, Singapore and an increasing number of other countries is that this virus does tolerate summer and even tropical conditions. The key issue is social distancing rather than climate.

There are also lessons to be learnt from across the Tasman where the Cedar Meats cluster in Victoria has exploded in the last few days from just two cases on 27 April to 75 cases as at 9 May. It has now spread from the meat company itself to at least one hospital, one care home, and one school. Under LEVEL 2, a similar outbreak in New Zealand could be traumatic.

There is also a tendency in some quarters in NZ to believe that an ongoing low level of infection is acceptable. What is apparently not understood by these people is that a low level of infection does not stay that way without a lot of heavy stamping just like we were doing in LEVEL 4.

There has now been enough testing in New Zealand that we can be confident that community transmission is not widespread. However, there could still be pockets of transmission, particularly among groups of young people who are the most likely ones to have very mild or even zero symptoms.

As for any moves to establishing a cross-Tasman bubble, each time the bubble expands there is increasing risk that we are exposing ourselves to new hidden embers. We would need to see zero new cases on both sides of the Tasman for at least a month before we could be confident that the embers had died out.

The science of uncertainty and the art of probability
Coming back to my chosen title for this article, it comes from a German medical friend of mine with whom I shared some adventures in the Andes of South America many years ago. In a recent communication he quoted from the famous 19th century Canadian physician and medical professor William Osler, who described medicine as “a science of uncertainty and an art of probability”. My medical friend and I both think that still applies very much today in relation to COVID-19.

Another apt quote from Osler was that he once said to his students: “Gentlemen, I have a confession to make; half or what we have taught you is in error, and furthermore we cannot tell you which half it is.”

In relation to what we currently think we know about COVID-19, it may well be that more than half is in error. So, we do indeed need to keep learning, and acknowledge that there are no all-knowing experts. There are no certainties. As we look around the world, the importance of the precautionary principle seems evident.

About Keith Woodford

Keith Woodford is an independent consultant, based in New Zealand, who works internationally on agri-food systems and rural development projects. He holds honorary positions as Professor of Agri-Food Systems at Lincoln University, New Zealand, and as Senior Research Fellow at the Contemporary China Research Centre at Victoria University, Wellington.
This entry was posted in COVID-19. Bookmark the permalink.

11 Responses to COVID-19 is a science of uncertainty and an art of probability

  1. Howard de Klerk says:

    Sweden has registered more deaths with Covid 19 as a comorbidity – not necessarily cause of death. Your chance of dying in Sweden so far in 2020 is no higher than any other year. 276 people have died everyday this year compared to 270 average.

  2. Sam Tuck says:

    Comparing Sweden and New Zealand or even just Sweden on a per capita basis without reference to the level of transmission is misleading. Naively NZ could expect that death rate when we are about a quarter of the way through the pandemics progression, based on Swedish estimates of %infected per capita. And about four times that once everyone has had it… but we don’t know if swedens death rate is representative either as how successful Sweden has been in protecting the vulnerable is vital, we did not know if it’s a subset of healthy that make the majority of Sweden’s estimated 25% infected per capita or those that have had it are unusually old and vulnerable like New York and Italy figures suggest.. in short per capita is misleading.. I’d love to read your take on the makeup of cases that present and what we can reliably infer from those? Like make up of cases that presentsuggest risks lie with comorbidity?

    • Keith Woodford says:

      The Swedish data is complex and opaque (as are the data for many countries).
      They have a low rate of testing. Also, their deaths only include people who have tested positive to PCR and so this excludes a lot of the people who have died in rest homes without being tested.
      The number of confirmed COVID-19 deaths in seden is running at a lower figure than the all-cause mortality excess in recent weeks so this is indicative of under-reporting of COVID deaths.
      About half of the laboratory-confirmed cases in Sweden are from rest homes.
      The Swedish ‘pandemic’ started about the same time that our numbers in NZ started to kick up so it is a reasonable country for comparison with NZ.
      UK provides an alternative comparison with NZ, but that comparison is with a country (the UK) that went into lockdown a couple of weeks too late. And what a comparison that dely has created!
      Retruning to Sweden, the basis on which Sweden is assuming a high level of infecction in the general population is somewhat shaky. It can only be based on antbody tests, and appears to be an extrapolation from very early antibody tests without acknowlegement of the specificity issue. In other words, the likelihood is that they got their antibody specificity (false positive) allowance wrong, and then their extrapolation exponentiated that up.
      If the disease has indeed spread like wildfire in Sweden then we would have also expected to see a very different spread of the disease in NZ with a lot more community transmission. One way or another, all will be revealed in the coming weeks and months. But personally, right now I am glad to be in NZ rather than Sweden or indeed anywhere in Europe or the Americas.

  3. Keith Woodford says:

    I would be interested in knowing your data source.
    For all-cause mortalities, I use the site. These are official figures for the 24 participating European countries.
    It shows that Sweden had below average deaths up to week 12 of this year (i.e. pre-covid), but by week 15 had risen to a Z score or 14.38. This is a very large increase – more than 14 standard deviations above average.
    Z scores for weeks later than week 15 are also high but are based on incomplete data. My experience is that these numbers always rise retrospectively.
    In any year, under normal conditions, we would expect approximately a 67% probability of all-cause mortailty lying within plus or minus one Z score of the mean, and 95% chance of lying within plus or minus two Z scores. So a figure of 14 is huge.

  4. Yiheyis Maru says:

    Thanks Professor Keith Woodford.
    That was quite informative post. Under such deep uncertainty, applying precautionary principles and preparedness to learn rapidly are key to increase our chances of effectively navigating through the multiple crises that COVID-19 brought

  5. sallyco says:

    “If We’re All Going to Become Infected, a Lockdown No Longer Makes Sense
    Unfortunately, it has become increasingly obvious that this virus is unstoppable. Even though we have a vaccine and know for a fact that it’s coming every single year, we cannot stop influenza.”

    Strangely this was my initial feeling. We have been duped.

    • Keith Woodford says:

      You are quoting one commentator.
      It is correct that in many parts of the world, lockdown can only flatten the curve because they started too late.
      The ACSH site is itself somewhat controversial – be careful about assuming that the opinion pieces there represent fact.

      • sallyco says:

        We will never know the truth and I will continue to look for alternate views. IMHO NZ has got it wrong. I hope I am wrong for the sake of my children and grand children.

  6. sallyco says:

    It becomes more evident that the models were and are garbage, just like the global warming scam which up to now has been the biggest hoax ever. The NZ government PM is a disgrace to what democracy means.

    • Keith Woodford says:

      As George Box said many years ago: “All models are wrong. Some models are useful.” It all depends on how the models are used. The Imperial College model was actually very insightful. But then it got misused.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s