The Way I See It

The media (and the UK government earlier today) is calling Covid-19 a once-in-a-century pandemic, but it might also be a once-in-a-century series of interconnected chaotic global responses to a problem that hasn't been adequately defined.

Many countries have introduced draconian measures in an attempt to counteract the disease, but as far as I can tell, there just isn't enough reliable evidence as to the numbers infected, or continuing to become infected, for governments and modellers to make such monumental decisions or to monitor their impact.

If the pandemic dissipates, whether that's on its own or as a result of social distancing and lock-downs, the impact of it may allow the world to return to something like "normal", whatever that will mean for the next decade or so. However, if the pandemic continues to spread around the world, how will those making the decisions be able to ascertain whether they've made a positive or negative impact when the consequences of long-term lockdowns are unknown and vaccines or affordable treatments take years to develop and test properly.

The data collected so far (or at least the information I've been able to find), on the number of people infected and how the epidemic is evolving don't appear to be particularly reliable statistically speaking as it's probably safe to assume that some deaths and the vast majority of Covid-19 infections are being missed due to the limits on testing. Therefore, it's impossible to state with any degree of confidence if we are failing to capture instances of coronavirus by a factor of three or 300 as few countries, if any, can test a large number of people and so don't know how prevalent the virus is in a random sample of the general population.

This creates a tremendous amount of uncertainty about the risk of dying from Covid-19 as the reported fatality rates (according to the World Health Organisation it's 3.4%) don't just create fear and panic, they're completely meaningless. As far as I can tell, those who have been tested for Covid-19 are disproportionately patients of some description with severe symptoms and bad outcomes, and since few health systems have unlimited testing capacity, this sample selection bias will only get worse.

There has been only one instance of testing an entire, closed population, that of the passengers and crew of the Diamond Princess cruise ship where the case fatality rate was 1.0%, but it needs to be noted that this was a predominantly elderly demographic for whom  the Covid-19 death rate is thought to be higher than the general population.

Taking the Diamond Princess mortality rate and applying it across the age structure of the UK, our death rate of people from (or possibly with) coronavirus would be 0.125%. However, we need to caveat that as this calculation uses what I consider to be extremely thin base data (7 deaths from 712 out of almost 4000 on board) the true death rate could be one fifth of that (0.025%), or five times higher (0.625%).

We also need to consider the possibility that some of those infected may die later, and that cruise ship tourists may not be truly representative of the rest of the developed world (i.e. their lifestyle, on or off ship, may mean they have different frequencies of chronic diseases) so if we say double those base estimates we get a varying case fatality ratio somewhere between 0.05% and 1.25% across a country like ours.

That difference indicates the risk of what is being done as a population-wide fatality rate of 0.05% is lower than seasonal influenza, which if it is the true proportion, the world's global lock-down and its likely social and financial consequences is not at all rational.

Now of course the Covid-19 case fatality rate isn't likely to be that low, and there are many common-cold-type viruses that have case fatality rates up to 10% when they infect the elderly, particularly if they're residents of nursing homes, but these “mild” coronaviruses may be implicated in several thousands of worldwide deaths every year without the vast majority being tested for their existence.

We've had successful influenza surveillance systems for decades, and typically something like 20% of laboratory tests confirm its presence in the samples sent for testing, while the estimated number of annual influenza-like deaths in the UK is between 4,500 and 11,000, therefore a positive test for coronavirus should not mean necessarily that this virus is always primarily responsible for a patient’s demise.

If we take a mid-range guess from my Diamond Princess analysis of say 0.3% as the case fatality rate of individuals with Covid-19, and assume that 15% of the UK gets infected (7,000,000 people), that translates to about 30,000 deaths, but without anybody knowing there was a new virus this would just get buried in the annual noise of deaths due to an influenza-like illness.

According to the Office for National Statistics, this year's death rate remains below its five-year average, and a long way under the levels of 2017-18, but even if the 2,100 were to be added to that total we might casually just note that this season's flu seems to be a bit worse than average, and the media coverage would be proportionate to that which accompanies the marriage of two minor celebrities.

It's obvious from the empty shelves in our local supermarkets that many people are worried that the 104 deaths from (or with as it should probably be more accurately labelled) Covid-19 in the UK as of right now is going to increase exponentially to 520, 2600, 13,000, 65,000 ... and that the pattern will be similar all over the planet, but is that realistic? At what point will the curve flatten as our Chief Medical Officer keeps talking about every time he's on TV?

Statistically speaking, we can't answer those questions without knowing the current prevalence of the infection in a random sample of a population, then  repeating the exercise at pre-defined time intervals to estimate the incidence of new infections, but unfortunately that's information we don’t have.

So in the absence of data, we've got prepare-for-the-worst reasoning which has led to lock-downs and social distancing without knowing if these measures have any sort of positive impact. Closing the nation's colleges may reduce transmission rates, but it may backfire if students socialise anyway, and the school closures might mean children spend more time with susceptible grandparents, or disrupt their parents ability to work.

And looking further forward, with the caveat that I know next to nothing about virology, might school closures also have the potential to reduce the nation's chances of developing long-term herd immunity in an age group that is apparently free of serious disease?
"Flattening the curve" to avoid overwhelming the NHS as our country is apparently attempting to do is admittedly conceptually sound because it theoretically means that other common diseases and conditions can be adequately treated. However, if the epidemic does overwhelm the health system and these extreme measures have only modest effectiveness, flattening the curve could make things worse by crippling the NHS for longer instead of a short, admittedly acute, period of time.

The bottom line is that we simply don't know how long lock-downs and social distancing can be maintained without irreversibly crippling our economy, and  destroying the fabric of society for generations, possibly even to the extent of civil unrest and long-term mental health issues.

The very minimum we need right now is unbiased prevalence and incidence data, and that means not prioritising testing for those suspected of infection, but a true random population sample so that governments can make evidence-based decisions, not ones based entirely on theories and models.

The most pessimistic scenario I've had on my TV this week, and one that my maths mean I really don't subscribe to, is that Covid-19 might infect 60% of the population before we develop a herd immunity and if only 1% of infected people die, there'll be more than 40 million deaths globally. And since that would make Covid-19 as deadly as 1918's Spanish Flu (although in this case the deaths would be mainly the elderly or people with pre-existing conditions), this is the justification for lock-downs and social distancing, making them not just desirable, but imperative.

Hopefully, rather like back in 1918, life as we know it can continue, at least in something resembling civilised societies, but lock-downs of months and years have consequences we can only guess at, and the lives of billions, not just millions, will be impacted, so we really ought to have some data before continuing with these plans to jump off the cliff just in case there is a chance of actually landing somewhere safe.