Coronavirus: lies, damn lies, and …

11/04/2020

Reported cumulative figures stand at 73,758 cases, up 5,706 from the previous 24 hours, and 8,958 deaths - up 980 from the last report. That last figure includes 866 deaths attributed to England.

Virtually every newspaper in the nation has reported on the top-line death figures, and many others besides, especially the Irish press. Thus we see the Irish Times trumpet: "UK records 980 new coronavirus deaths in hospitals", telling us that the toll is the "worst suffered by any European country in one day since the epidemic began".

We are then advised that Britain's toll on Friday surpassed Spain's worst daily toll of 950 and Italy’s of 919, with the United States now the only country in the world recording higher daily death tolls. France's official figures, we learn, are sometimes higher but unlike most countries, they include deaths in care homes and other settings outside hospitals.

On 406 occasions however, according to Google News, papers paste in part of a Reuters report – acknowledged by only 19 of them – breaking down the English figure. Of the 866 new deaths announced on Friday, we are told, 117 occurred on 9 April while 720 took place between 1 April and 8 April. The remaining 29 deaths occurred in March, including one on 5 March.

Although so many daily papers seem to have printed this report, none seems to have understood it. Neither, it seems, does the doyen of Channel 4 News, Krishnan Guru-Murthy, who sternly tells us that the overall total of 980 "is not the number who died today". It is, we are instructed: "Just the number they announced of those who died in hospital. The real number of extra dead in 24 hours - including those not yet announced, those in care homes, at home, etc - could be hundreds more".

But, of that larger component, the English figure, we now see that only the smaller proportion relates to the 24-hour reporting period. The far greater number were spread between the preceding eight days, and 29 were stragglers from last month.

In no way therefore, can yesterday's headline figure be taken to represent a daily toll. In truth, the number of people who actually died on 9 April is (and will now remain) a complete unknown to us, the general public.

Furthermore, the implications of this spread of reporting is profound. It is almost certainly the case (even if it hasn't registered) that on most days when a daily figure has been published, it comprises deaths which occurred on different dates. Thus, there is no single, pure timeline on which we can rely, giving us some sense of the daily death rate.

This is only to be expected. Deaths will only be officially recorded once the death certificates are filled in. And for a variety of reasons, there will be delays, which means that for a given day, the certificates will only filter in over a period.

One doesn't need to read any conspiracy into this – collating such figures takes time, even without the pressure of an epidemic. But I wish it had been made clearer by the Department of Health that the figures it has been publishing are provisional and do not actually represent the daily death toll.

It might help if the great and the good of the scientific and medical establishments acknowledged the fragility of the data. But instead, we have the likes of Professor James Naismith, Director of the Rosalind Franklin Institute and Professor of Structural Biology, University of Oxford (there's glory for you), declaring: "With an additional 980 deaths reported, today represents an extremely sad day for the country".

He does at least add: "The volatility in the daily numbers due to the nature of the reporting makes confidently identifying a hopeful trend in the daily total of announced deaths very hard", but what he should be saying is that, for statistical purposes, the daily reports are completely valueless. No useful trend can be determined from the figures presented.

But then, even if we were getting the actual daily reports, they would have to be treated with some circumspection. Even though a number of people may have died on the same day, they could have acquired their infection in a window which might span a few days to a fortnight, while the duration of their individual illnesses might run for similar indeterminate periods.

On that basis, even with the best will in the world, what each daily cohort represents cannot be taken as a snapshot of the progress of the epidemic. And when one adds the third variable – the period taken officially to report the deaths – the scatter should preclude anything other than a broad brush view of what is going on.

The position might be better with the case figures, although there will still be temporal variations. There is, for instance, still the range of incubation periods to be taken into account. Then, because most patients are only tested on admission to hospital, and that will occur at different stages of the disease, there is another variable.

From an epidemiological point of view, the ideal data set would comprise figures based on the dates of exposure to the infectious agent but, in this type of epidemic that is not possible. It is possible in a single source outbreak of the type that we see in some food poisoning events, but we don't have this luxury here. At best, therefore, the epidemic curve can only be a very rough approximation.

When, as we have also seen, the basis of testing, and then the reporting, are changed through the progress of the epidemic – further errors are introduced, on top of which there are administrative fluctuations – such as the "weekend effect" with which we are now familiar.

It is instructive, therefore, that at yesterday's presser, we saw the deputy CMO, Jonathan Van-Tam, using hospital admissions and intensive care cases. These are not published openly, but Van-Tam claimed that they were falling in London.

Nevertheless, exercising some caution, he stopped short of saying the worst was over, instead telling us, "It's impossible to say that we've peaked but I suggest to you that the curve is starting to bend and your hard work is beginning to pay off". He added: "We have to take the pain now to make the gain in a few weeks' time".

Actually, even these figures – from a statistical viewpoint – have the potential to mislead. There are strong indications (more than just mere rumours) that elderly people (and especially those in care homes), who were treated in hospital in the early stages of the epidemic, are now being refused admission.

There are even some reports (unsubstantiated) that local GPs have been refusing to attend care homes where Covid-19 is reported (or suspected), without which some patients are not referred on to hospitals.

Then, as the effects of intensive care treatment become more widely known – along with the very poor survival rate from intrusive ventilation procedures – individual patients are declining such treatment, while clinicians are screening out those with poor prognoses, earmarking them for palliative care.

Without taking such changes into account, even a slackening of throughput in the hospitals could be misleading, especially if one is looking at small changes and trying to detect subtle changes that might presage a downturn in real infection rates.

And then, of course, there is the "ripple effect" to factor in – the inconvenient but important fact that, while we are dealing with a single epidemic, this comprises multiple outbreaks at different stages of development. A slackening off in the intensity of infection in one area is no guide as to the progress of the epidemic as a whole.

It may well be, therefore, that the chief scientific adviser is reading too much into the statistics when he confidently informs us that the: "death toll will continue to rise for about two weeks after intensive care numbers improve". He is now asserting that time for cases to double is now six or more days almost everywhere in the country, on which basis he detects a slackening in the epidemic intensity.

In this, logic might support him. The lockdown, however imperfect, is bound to have some effect, so it is not entirely reckless to predict that we can expect a downturn. But what neither Vallance nor anyone else can tell us is the proportion of the population which might have been exposed to infection, giving us some clue as to the rate of spread of Covid-19 once restrictions are relaxed.

From a number of different sources, however, we are getting some intimations that the exposure rate is lower than might have been expected – perhaps significantly lower than ten percent – which means that there might be plenty of fresh meat for the virus to work on once the gates are opened.

Van-Tam is not wrong, therefore, when he says: "We are in a dangerous phase still. And I need to reinforce that again to you, that this is not over". And there, although we are being kept in the dark about the daily death rate, Van-Tam may have the advantage of unofficial estimates coming through his office, based on mortuary occupation.

With that, one has the feeling that we are being prepared for worse to come, as the officials undoubtedly know far more than they are letting on. Not for nothing are we seeing 30-plus emergency mortuaries being commissioned. And we have been warned: "this is not over".