# Independent SAGE and the “27,000 Excess Deaths” Claim

## It is implausible there would be another “27,000 excess deaths” without another wave.

Speaking on Sky News, Prof Sir David King (Independent SAGE) asserted:

What we are saying is 27,000 excess deaths are likely between now and next April if the expectation by the Chief Medical Officer is that he would be surprised and delighted if the UK is in the same place next spring.

If he’s correct we would still have about 2,000 to 3,000 new infections in England per day and that is the number of deaths that would follow from that.

Broadcasters Sky News and ITV repeated this claim on their websites. Newspapers repeated this figure, including: Evening Standard, Daily Mirror, The Independent, and Metro.

It is a very stringent assumption for infections to remain constant for nine months. The estimate implies an implausible infection fatality ratio of around 3.3%.

# Excess deaths are not COVID-19 deaths

There are many problems with this claim. First, it confuses COVID-19 deaths with excess deaths.

Excess deaths means deaths above a baseline. That baseline is often an average of past years. The Office for National Statistics uses the past five years. Institutes may use modelled baselines for their ‘excess’ death calculations.

In the latest ONS report, weekly death registrations was lower than past years. For 20–26th June (week 26), the number of registered deaths from all causes was 3% below the five-year average. These death registration statistics are for England and Wales.

Not all deaths with COVID-19 will be in excess. An epidemic disease can ‘bring forward’ some deaths. That displacement leads to fewer deaths in later weeks. Certificates may not give the correct diagnosis for all deaths. Some deaths with COVID-19 may not be diagnosed.

This is a conceptual difference between saying ‘excess deaths’ and ‘deaths with COVID-19’. Mortality displacement is difficult to quantify.

# SARS-CoV-2 infection levels

The estimate appears to rest on a very stringent assumption. The assumption is there would be between 2,000 and 3,000 new infections each day.

The basis is Prof Whitty’s comment:

I would be surprised and delighted if we weren’t in this current situation through the winter and into next spring. I think then let’s regroup and work out where we are. But I expect there to be a significant amount of coronavirus circulating at least until that time.

The Chief Medical Officer does not assert an expectation of constant SARS-CoV-2 infections. The CMO expects “significant” viral prevalence.

Are infections remaining constant? It does not appear so.

Confirmed cases continues to decline. Hospitalisations are also falling. As a measure, admitted patients lags infections.

The ONS infection survey of English private households suggests declining positive proportions. The survey is estimating the number of people that would test positive.

The latest estimate for 22nd June to 5th July was 0.03% (0.01% to 0.06%).

Estimating the actual prevalence of SARS-CoV-2 needs assumptions of test accuracy. Tests are imperfect. The survey is too small to be clear of that decline in recent weeks.

# Infection fatality ratios

Suppose there were 3,000 new infections per day. We are looking at 1st July 2019 to 31st March 2021. 27,000 deaths would need the infection fatality ratio of ~3.3%. This is very high.

It is the proportion of infected people who will die from the disease. A Nature article highlights central estimates of this ratio between 0.5% and 1.0%.

Infection fatality ratios are not a biological constant. Estimates depend on demographics, healthcare quality, and study methods.

The infection fatality ratio can change during an epidemic. Healthcare can improve. Who has the virus can also change. Estimating this ratio by age is important.

# Publishing analysis

There is no explanation on their website for how they calculated this figure. It is unclear if the claim refers to England or the whole UK.

Chief Medical Officer Chris Whitty suggests virus will continue to circulate at ‘significant’ levels; at current levels this shows at least 20,000 more people could die by April even before further rules relaxed.

The suggestion is the current level of infections implies this number of deaths. No provided analysis supports that claim.

It turns out the method was simple. The group rounded down to 100 deaths and multiplied by the rough number of days:

27,000 comes from rounding down to 100 for 9 months per day as a simple calculation.

Despite Prof King’s statements, the claim did not derive from estimated infections. This crude method is very inappropriate for estimating long-term COVID-19 deaths.

Deaths follow from cases. The most recent deaths are from past infections. An early analysis of 24 deaths in mainland China suggested an average lag of 17-19 days. This is the estimated average time between onset symptoms and death.

There is also reporting lag, between when a death happens and when it enters the reporting system. The latest infection levels does not imply the latest death counts.

Are confirmed deaths averaging at 100 per day?

No. We can look at the seven-day rolling average for the DHSC measure of COVID-19 deaths. This measure is by reported dates.

On Friday 19th June, that average was 141. A week later, it was 121. On 3rd July, the figure was 103. Yesterday (10th July): the average was 74. The PHE dashboard uses a centred rolling average, so we shift along the dates.

This average is not constant — more like an exponential decay. A constant extrapolation is inappropriate for forecasting a falling measure. Extrapolating current counts far into the future is likely to overestimate new deaths.

Published analysis should partner these kinds of claims. We can then study methods. Scientists should reflect uncertainties in statements about effects of public policy.

Without publishing analysis, Prof King and Independent SAGE let loose implausible figures. The figure does not mean what they say it means. Assuming a constant level is not appropriate for estimating future deaths.

This blog looks at the use of statistics in Britain and beyond. It is written by RSS Statistical Ambassador and Chartered Statistician @anthonybmasters.

## More from Anthony B. Masters

This blog looks at the use of statistics in Britain and beyond. It is written by RSS Statistical Ambassador and Chartered Statistician @anthonybmasters.

## Vaccine Trial Participants Randomized to Placebo Should Get Vaccine Now

Get the Medium app