Skip to Main Content
Page banner

Being Assessed
Thanks for your vote
This voting milestone has now ended, and the idea has progressed to a private milestone
Return to idea list
Thanks for approving the idea
This milestone has now ended, and the idea has progressed to a private milestone
Return to idea list

Increase Covid Testing by 3, 9 or 15 fold


Mass Testing for Covid 19 – Welch Protocol

This depends on the actual level of Covid 19 in the community. Particularly useful where you expect the incidence of positives within a group to be low, such as all inhabitants and staff in a care home, NHS staff or mass testing of the population in general

Results to date show that roughly 1 in 4 of those tested are positive, so far these are mainly patients with symptoms. At 1 in 4 you might as well test individual samples – see below

3 Fold increase in testing, 3 Fold Reduction in Key Reagents

Assuming this is a lower ratio when testing key workers, for simplicity assume it is 1 in 32

Swabs taken from 32 individuals.

 Each person’s set of swabs are extracted in such a way that the samples can be subdivided for 5 subsequent tests.

Looking at what I could find of the protocol, you can go back to the original swabs and repeat extraction.

( If this can’t be done i.e. if the whole of each person’s swabs need to be taken per test then take 5 sets of swabs per person. Various other scenarios are possible e.g. if the swab extracts can be halved then take 3 sets per individual)

This argument applies if there is only one positive in the group

1)      Combine all 32 sample extracts and do a PCR test if its negative you’ve cleared 32 people with 1 test.

If positive split the samples into 2 lots of 16 swabs combined.

2)      Test both groups.  - one of the groups of 16 would be cleared

                             you’ve done 3 tests and cleared 16 people

3)      Take the group of 16 you know to be positive and split them into 2 groups of 8

Test both groups. - one of the groups of 8 would be cleared .

                             you’ve done 5 tests and cleared 24 people

4)      Take the group of 8 you know to be positive and split them into 2 groups of 4

Test both groups. - one of the groups of 4 would be cleared .

you’ve done 7 tests and cleared 28 people

5)      Take the group of 4 you know to be positive and because two groups of 2 and 2 groups of 1 now amounts to 4 tests so you might as well test the 4 individually                     

you’ve done 11 tests and cleared 31 people      

You’ve got the last one who you know is positive

9 Fold increase in testing, 9 Fold Reduction in Key Reagents

If you want to screen the general public and assume higher positive ratios, testing for 1 in 64 would only add 2 more tests, total 13 tests, and require 1 more set of swabs or swabs extract, total 7 to clear 63 people

16 Fold increase in testing, 16 Fold Reduction in Key Reagents

If you assume 1 in 128 positive ratio, you again add 2 more, total 15 tests,and require 1 more set of swabs or swabs extract , total 8 to clear 127 people.

Note – you have to test the two groups at each stage because you cannot be sure that there is only 1 positive in the group – if there is more than 1 there may be more splits and more testing but it will still be less than individual testing of swabs.

Saves reagents but may lengthen the time to get an individual result but you can do so many more per day.

Depending on the test for Antibodies you could use a similar approach looking for a positive test for absence’

I’m not a statistician who might indeed improve this protocol.

Have you validated this method, if so, how and what were the results of the validation?​

Not validated - depends on the statistics. I'm not a statistician , I'm a Chemist.

How quickly could this be deployed and what are the dependencies?​

Simply a method

What is the likely production volume?​


What are the risks and barriers to using this at scale?​

Validating and rolling out a standard methodology and when to use it. 

Who are you already partnering with on this?


You will need to login to post a comment

Sonia Gandhi Apr 19, 2020

Status label added: B

Reply 0

Geoff Baldwin Apr 20, 2020

You also need to care about sensitivity here. If you are adding 32- or 64-samples to a single qPCR, then you are using 1/32 or 1/64 of the original sample and will therefore reduce the amount of positive RNA present. This is a risk that would need to be validated. See also the discussion around sensitivity for RNA extraction free testing.

Reply 0

Alexander James Phillips Apr 28, 2020

Agreed. This is a genius suggestion but I don't think decreasing sensitivity is acceptable unfortunately. Really nice lateral thinking though! Keep it up! :)

Reply 0

p courtney Apr 30, 2020

several papers out now on this suggesting that sensitivity is not so much of an issue (up to a point) since PCR gives such leverage.
2 from germany and one from israel:

Reply 0

p courtney Apr 20, 2020

nice idea - see references here

Reply 0

Quentin Hanley Apr 20, 2020

I support this approach.

This is a pooling related proposal. There are others:

There are already papers indicating pools of 32 can work and suggest 64 are feasible:

Other fully published work.

In this one, there is a link to an R-script that attempts to estimate optimal pool size for any expected prevalence.

There is additional code in support here:

Reply 0

Bev Matthews Apr 21, 2020

The idea has been progressed to the next milestone.

Reply 0

Bev Matthews Apr 25, 2020

The idea has been progressed to the next milestone.

Reply 0

Bev Matthews Apr 25, 2020

Hello Anthony, the reviewers are interested to hear more when you have further data. please do come back at that point and we look forward to hearing from you. bev

Reply 0

Anthony Welch Apr 30, 2020

Hi Bev, replying to your request for further data. As a retired Chemist I am not in a position to acquire more data. I only advocate this approach particularly in the screening of larger groups where low incidence of positives is expected and frequent monitoring is desirable.
Examples -
Groups of NHS staff, police, firebrigade, ambulance drivers etc.
Supermarket staff
Public service workers
Groups of employees returning to work
Sports Teams and staff for competition behind closed doors.

i.e. many people in a single test. I rely on the experts who fully understand the testing procedures to decide if the idea is viable, e.g. I don't know what the minimum level RNA has to be at before it may be missed during testing.

Reply 0

Charles Ravarani May 5, 2020

We have developed a web-application (at that aims to make the adoption of the pooling strategy easily accessible to laboratories. It fits into existing pipelines and tries to mitigate the concerns of sensitivity by automatically integrating replicates in the protocols it produces.

We have a working application and are looking for collaboration to help assess the feasibility of the approach. We are open all feedback in order to develop further requirements that make adoption as easy as possible in more labs.

Reply 0

Tom Jordan Jun 27, 2020

Status label added: D2

Status label removed: B: One to watch

Reply 0

Tom Jordan Jun 27, 2020

Hi everyone,

Thank for you sharing this and for the helpful discussion in the comments section. We have shared this solution with colleagues leading on sample pooling. They will be in contact should they have any questions.

Kind regards,
Tom Jordan
Testing Methods team

Reply 0