TheEconomistNovember20th 2021
Graphic detail Covid-19 testing91An Immensa
cock-up
A
mongthetoolshealthofficialshave
used to fight covid19, the impacts of
lockdowns, masks and vaccines have been
wellstudied. Testing, whose effect is hard
to measure because it coincides with other
pandemiccontrol policies, has received
less attention. But thanks to a grim natural
experiment, that is now starting to change.
In September covid seemed to come to a
halt in southwestern Britain. Overnight,
the positivity rate of pcrtests near Bath
and Bristol fell from 3% to 1%. Nowhere
else in England enjoyed a similar decline.
By midmonth a local volunteer group
said this apparent dip might stem from
faulty tests. A few weeks later, the group
and local scientists asked Britain’s Health
Security Agency (ukhsa) to investigate.
On October 12th the ukhsashut down a
lab run by Immensa, a testing firm. The
agency said that the firm had incorrectly
told 43,000 infected people that they were
virusfree. The cause of the error remains
unclear. Immensa declined to comment.
As soon as the lab was closed, the re
gion’s reported covid case rate soared. A
spokesperson for the prime minister said
that the lab errors did not cause this surge.
On November 14th, however, Thiemo Fetz
er of the University of Warwick released a
paper showing that they probably did. It
has not yet been peerreviewed, but offers
firm evidence that accurate testing does in
deed slow covid’s spread, by letting infec
tious people know that they should isolate.
To estimate what might have happened
were it not for the snafu, Mr Fetzer built a
“synthetic control”: a group of areas whose
prior rates of vaccination, testing, and co
vid cases and deaths matched those of the
13 most affected regions. The difference
was stark. From September 2nd to October
12th, the areas in question recorded 13,000
fewer positive tests than the control. After
wards, they registered 21,000 more.
This implies that each faulty test may
have led to 0.61.6 extra cases (the range re
flects uncertainty over how many people
who tested positive received earlier false
negatives). Based on Britain’s casefatality
rate, this translates to 4001,100 deaths.
Surprisingly, this toll is not the highest
that Mr Fetzer has attributed to technical
glitches. In 2020 a spreadsheet error pre
vented Britain’s statistical service from re
porting 15,000 covid cases to contacttrac
ers. In an earlier study, Mr Fetzercalculat
ed that 1,500 people died as a result.n
False-negative covid-19 tests in Britain
led to thousands of extra cases→ Reported covid-1 case rates in part of Britain fell and rebounded abruptlyShareoftestsreportedpositive,%
2021Difference in rate of positive tests from
period with false negatives to month after→Thesurgeincasesimpliesthatfalse-negativetestshelpedcovid-1spreadConfirmedcovid-19cases,differencefromexpected†
13 most-affected*localauthorities, 2021Maximumnumberofadditionalcasesattributabletoeachfalse-negativetest‡
13 most-affected*localauthorities*SentmostsamplestoImmensalab †Basedondatafroma groupofregionswithsimilarpriorratesofvaccination, testing
andcovid-19casesanddeaths ‡Assumingthatnoneofthepeoplewhoreceivedfalse-negativeresultssubsequently tested
positive Sources:“Measuringtheepidemiologicalimpactofa falsenegative:evidencefroma naturalexperiment”,by
T.Fetzer,2021;OfficeforNationalStatistics-1 000-50005001,0001,5002,000Jun Jul Aug Sep Oct NovTotaldifference
fromexpected
-12,972Immensalabreports
falsenegativesTotaldifference
fromexpected
20,853012345BathBristolSedgemoor CotswoldsSwindonAverage
Eachfalse-negativeresultledtoas
manyas1.
additionalinfectionsPopulationin2020,’000
100 200 5000246810Jul Aug Sep Oct NovRestofEngland13 most-affected*
localauthoritiesImmensalab
reportsfalse
negatives95%
confidence5kmENGLANDImmensa
labBathBristolTheseareassentthegreatest
shareofteststoa labthat
reportedfalse-negativeresults←Before After →More positive results