COVID-19: WHAT DOES THE NUMBER OF NEW INFECTIONS MEAN?

In understanding statistics of new infections, there are three aspects we need to consider:

• For what period? Typically daily or weekly.
• To what extent do numbers reflect all new infections?
• How many people are out and about infectious, but blissfully unaware?

WHAT PERIOD

For weeks, the UK Government published statistics of "New Confirmed Cases" per day, often with a useful 7-day average.  Here's an example from June:

Now it seems that quoting various statistics per week is becoming the norm in the UK.  For example this announcement by Oxfordshire County Council today, 3 September:

Per fortnight is also used, for example by the European Centre for Disease Prevention and Control from which this graphic was produced for the 14 days to 16 August (compared to 14 days to 8 August):

We need to be extra careful in understanding whether daily, weekly or fortnightly statistics are being used, and converting one to the other where needed.

This is important for three reasons:

• Comparing countries and areas
• Estimating the number of people likely to be out and about infectious without being aware of it 
• Assessing how up-to-date the data is.  A rapid surge or fall is best assessed daily by 7-day averages, and no more than weekly.  Also we need to bear in mind the period end date, and how long after actual infection.  For example testing people who are symptomatic is around a week after infection

ARE THESE ALL INFECTIONS?

Whether daily or weekly figurers are used, we need to bear in mind that official "new confirmed cases" arising from NHS Test and Trace are understated for three reasons:

• Not all symptomatic people get tested, possibly only 1 in 3 on average.  Better if testing is stepped up in hotspot areas
• Swabbing, especially self-swabbing, creates false negatives of 30%-50%
• Asymptomatic people, who can still be infectious, are at least 50% of total infections

That suggests that real infection figures are around nine times the official figures.  Comparisons to other data, notably the ONS sampling and King's College ZOE, suggest the factor is around 5.  That should certainly be considered the minimum.


ASSESSING NUMBER OUT AND ABOUT INFECTIOUS

We've previously estimated that on average symptomatic and asymptomatic people are infectious for at least 5 days.  People typically develop symptoms 5 days after infection (2-14 days), during which time they will be infectious.  Asymptomatic people are also likely to be infectious for some time whilst they fight off the virus.

Some of these people will be self-isolating because of being a contact of someone tested positive, and others will have returned from abroad and in quarantine seemingly uninfected.  But other people who ought to be self-isolating may be out and about, such as if they are on zero-hours contracts.

On balance, and until a better estimate is made, the number of people out and about infectious will be around 5 times the daily infections.  That is 25 (5x5) the daily official "new confirmed cases".

Using weekly numbers, they simply need to be multiplied by 3.5 to give a equivalent number infectious.  This is because the weekly number is 7 times the daily number not 5. 

For example, Oxford's latest rate for the week to 28 August was 23:

• Multiply weekly by 3.5 is around 80 infectious
• Based on daily equivalent is (23/7)*5*5, which also gives around 80 infectious

Note that for the previous week, when it was 41, that would have meant about 150 out and about infectious. 

Hence reducing Oxford's alert level from Amber to Yellow.  Though beware the figure for testing in week ending 28 August is effectively infections in the week ending 21 August.  That's two weeks ago.  These latest figures are certainly not the current level of infections!