tl;dr: A method to get fully reliable testing for large numbers of people while reducing the number of tests: combine PCR with pooling and antigen testing.
I’ve just read an article in Scientific American (also an article in the Atlantic) about the issues involved in testing for the Covid-19 virus. While the idea of “testing at home/office” with an “instant coffee” test should probably be addressed separately, the very cheap, easy antigen test could be part of a combined testing regimen that could save substantially on the number of more expensive PCR tests needed to be run.
I’ve undertaken to describe (here) a combined testing process, which is (or should be) obvious to anyone in the field, which would allow full diagnostic PCR testing accuracy for everyone tested while substantially reducing the actual number of PCR tests run.
The above flowchart roughly describes the process. It uses the standard flow-charting method from IT, and could be partly automated (in my opinion) with proper design of sampling/testing equipment.
The key here is that the “antigen testing” method tends to be less reliable, especially for false negative outcomes. At the same time, when a number of people’s samples are “pooled”, combining a small amount from each sample into a “pool” that is mixed and subjected to PCR testing, a negative outcome will be reliable for everybody. A positive outcome, on the other hand, just means at least one sample in the “pool” is positive.
The process depends on “pooling” samples that already had negative outcomes from the “antigen testing”, while those with positive outcomes (from “antigen testing”) are simply tested individually. Additionally, all those that are part of any “pool” that tests positive will be tested individually.
Samples can be collected as saliva individually, labeled with bar-coding, then transferred to testing centers, where they go through automated testing processes.
The sample containers would have to be designed for easy mass-production, and to be easily handled by automatic machinery in the testing centers, including removing small amounts for pooling and individual testing.
Tracking of samples can be automated, computer-driven, as can the extraction and “pooling” of small amounts of negative samples.
I did this mostly as an exercise in communicating process design, since it should be obvious, in my view, to anyone in the field.