COVID-19: MONITORING THE NUMBER OF NEW DAILY CASES

When we take a business-like approach to managing the COVID-19 pandemic, the principal objective is to lower the number of new cases.  We therefore need to monitor that figure, preferably:

• Daily, especially if cases are rising
• Previous day’s figure available by 8am each morning, so any action can be taken that day.  Not just 7-day averages
• 24-hour period defined to reliably achieve that reporting schedule
• Covering all people who have contracted the disease.  That is those with symptoms, those not yet with symptoms (pre-symptomatic) and those never to show basic symptoms (asymptomatic), including children
• Focused on the community, known as "pillar 2".  Hospitals, care homes and other institutions need to be managed separately
• Analysed down to the smallest geographical area possible

The data currently available is nowhere close to these ideals.  But let's see what is available, and how we can still use it.




 
DAILY TARGETS

We can compare an area’s daily infection rate to a target expressed as “new daily cases per million population”.  Now more commonly reported as per one hundred thousand, to be more comparable with the size of towns and cities. 

That also allows us to compare areas and compare countries.  The latter to assess which are ‘safer’ countries or regions, and which are ‘riskier’ requiring immigration controls for travellers from those countries or regions .

Professor Chris Whitty suggested in April that we should be down to less than 1000 cases a day .  With England having a population of some 55 million, and assuming he was thinking across the entire population, man woman and child, that is some 18 cases per million.  Or some 2 cases per hundred thousand.  

One hundred thousand is the size of places such as Lincoln, Great Yarmouth and Watford.  A large city such as Leeds is some 800,000, with metropolitan areas such as London and Manchester being split into smaller areas.

The Independent SAGE group go a step further, suggesting a ZeroCOVID strategy  which is to “control” COVID-19 to a very low level that is virtually “elimination".   “Independent SAGE would suggest that a seven day rolling average of one new case per million population per day could represent ‘control’ in Scotland, Wales, Northern Ireland and [each of] the regions of England.”  That’s 0.1 case per hundred thousand over wide regional areas. 

Whilst #ZeroCOVID is desirable, that is extremely difficult to achieve, and arguably unnecessary unless you can get new infections down to literally zero like New Zealand.  They can only keep zero by very strict immigration measures, which would be impractical for England. 

A more reasonable target for England would be Professor Whitty’s 2 new daily cases per hundred thousand, which is known as #NearZero.  At that level, risk would still be very low.  In combination with immigration controls and a good public education campaign, it might mean say 10 new local outbreaks a week across the country, typically assocuated with a single factory, pub or other venue.  That would be reasonably easy to deal with.  The risk of any single person getting COVID-19 would be very low, and certainly acceptable given all the other risks in life.


CURRENT INFECTION LEVELS IN ENGLAND

No data is produced yet to the standard set out in the introduction above.  Data is either incomplete, lagged by a week or more, and/or not available reliably at a local level.

There are a number of organisations which are monitoring infection levels in England, typically universities.  The three of most interest, providing data publicly, are:

• The ONS (Office for National Statistics) with their “Coronavirus (COVID-19) Infection Survey” which they call a “pilot”.  This is a randomised study, across everyone in England, be they symptomatic or non-symptomatic, and across all walks of life. It is therefore the most reliable data, but is only produced weekly.  The samples are too small to provide reliable regional analysis, and indeed the 95% confidence spread for the national figure is rather wide.  Let’s hope the sampling is substantially increased..
• The King’s College London study using the ZOE app, downloaded by some 4 million people across the whole UK. The data is again reported weekly, based only on symptomatic people, and only for people who can afford a smartphone.  Some regional analysis is available.  The data is therefore inherently only a subset of the total who have caught the disease, seemingly needing to be multiplied by 2, but is useful for trends.  As the data is symptomatic, it inherently lags about a week behind new infections.
• Public Health England publish data from the NHS Test and Trace programme.  This is the only data produced daily, but as it only reflects people actually tested, it is only the tip of the iceberg of true new cases, needing to be multiplied by about 5.  However the data does have the advantage of being produced daily, analysed by nearly 7000 local areas.  But as community testing is only if the person has symptoms, the data also lags by around a week

Let’s look at each of these in turn.


ONS “CORONAVIRUS (COVID-19) INFECTION SURVEY”
 
As noted above, this data is most representative of the English population, and is the most up to date. However the sample size could do with being larger, both to make the total more accurate and to provide reliable regional analysis. 

This is taken from their latest report (which will change as each week goes by):

So to be 95% confident, daily cases could be anything between 2,200 and 8,100, and to be 100% confident would be even wider.  However the 4,200 level makes sense by comparison to the other data sources, and is a good working figure.

They acknowledge that infections have doubled since mid June:

KING’S COLLEGE LONDON STUDY USING THE ZOE APP

As explained above, the number of daily cases is inherently understated, and lags actual infections by around a week.

Here is the latest graph for England.  This shows infections stable at just under 2000 a day, having dropped to around 1000 a day at the end of June.  They roughly doubled in those four weeks, but has seemingly stabilised.

That is consistent with the ZOE report, but started to rise a little later reflecting that ZOE data lags ONS.

The average for England is therefore some 3 per hundred thousand, though known to be understated.  If the ONS figure of 4.200 is used, ZOE data needs to be roughly doubled. That seems reasonable.

Here is the national and regional analysis for the period to 31 July.

For the North West as a whole, this shows some 4 new cases per day per hundred thousand (39 per million) on average, suggesting the high rate towns listed below are balanced to a certain extent by lower rate areas in the North West.


NHS "TEST AND TRACE" PROGRAMME

Public Health England publishes data [link] on new infections on a daily basis, as raw figures and expressed as new cases per hundred thousand.  This is analysed to nearly 7000 local areas, so in that respect the data is very useful, especially when looking at trends.

The data reflects tests done under the NHS Test and Trace programme around the country.  The data is not of the quality that we might hope for, though:

• Excluding people who are infectious but haven’t yet had a test, and excluding everyone who is non-symptomatic.  The data should only be considered the tip of the iceberg.
• As community testing is only on people who have symptoms, plus test processing time, the data is about a week delayed
• Community testing is called “pillar 2”, with pillar 1 being mainly in hospitals and other institutions.  It is not clear which pillar(s) the data represents
• For some of the data, it is unclear whether it represents a week or a day, or whether it is new infections or total infections.  This is very ‘disappointing’.

The figures are therefore inherently substantially under-stated, and difficult to interpret. 

The total new infections figure for England is around 850 per day.  With ONS suggesting 4200 per day (see above)), the NHS figures look as if they need to be multiplied by a factor of at least 5.  That seems reasonable in the circumstances.  The NHS data has the advantage of being the only data with local analysis, for which figures need to be multiplied by at least 4..

The top 15 towns and cities are as follows.  The rate is expressed as the number of new cases per 100,000 people, for the week to 26 July, compared to week to 19 July. It is not clear what the numbers represent:

• The ONS data above of some 4200 new infections per day for England, represents nearly 8 per day per hundred thousand on average, or 56 per week.  
• Looking at Swindon’s and Oldham’s rates, it suggests the figures below are per day, rather than per week, but this is not stated.  
• Nor is it clear whether they are pillar 1 (hospitals etc), pillar 2 (community) or both.  So the figures are only ‘representative’, subject to still needing to be multiplied by 5:

Whatever the definition of these figures, it is nonetheless clear that:

• These towns are way above the 2 per day which we should be targeting
• There must be many more towns and village areas that are above that 2 target

Here is a map of England cases from the government website.  The darker the area, the higher the daily infection rate:

Only the areas in white have low infection rates (ignoring Wales and Scotland).  It is clear from this map that much of England is above target rates.  Furthermore, as it is not clear whether tests are analysed by where performed or a person’s primary residence, staycation areas like Cornwall and the Lake District may be understating the number of cases physically there.

We can see the North West is a darker colour, with towns merging with each other for a score of 30 or more.  Reduce the threshold to the target 2 cases per hundred thousand, and many areas merge together.  That means action is needed at a national level.


IN SUMMARY

No data is available daily on a very timely basis, nor in sufficient detail for monitoring local infection rates.  However we have to work with what we’ve got.  We can make reasonable estimates, but the inherent lag in all the data is 'regrettable'.

The ONS sampled data is the most representative to cover all new infections, and is the quickest to reflect them.  But it is only available weekly, and at a low elvel of accuracy.  The latest data for the week to 26 July suggests some 4200 infections per day in England, double what it was in mid June, and is still rising.

That’s nearly 8 per day per hundred thousand on average, well above the 2 per day which Professor Whitty suggests is the level England ought to be.  2 seems to be a reasonable target to let life get back to near-normal, raise confidence to fire up the economy and get kids back to school, and minimise deaths and LongCOVID.  That means England’s infection rate is some 4 times higher than it should be, and rising.  The rise needs to be nipped in the bud and the infection level reduced down to the target.

The King’s College London ZOE data is also showing a doubling of daily infections since June.  Their data suggests that the level of new infections is just under 2000 a day.  As this inherently understated, it looks like all their figures need to be at least doubled.

The NHS Test and Trace data is showing only 850 new infections per day across England, but is also inherently understated.  It looks like their numbers need to be multiplied by 5 to fully cover symptomatic people not tested, pre-symptomatics and asymptomatics.

The NHS's regional and local data is showing daily infection rates well above 2 per hundred thousand in many areas.  The areas with higher daily rates are tending to merge into one another.  The recent lockdown across the north reflects daily infection rates of around 30 or more.  Take that threshold down to 2 where it ought to be, and it is clear local initiatives would merge with each other over wider areas.  Action is therefore needed at national level. The national action suggested is covered here.