Cassandra's curse

I'd thought this was dead-obvious mid-January. I was surprised nobody had written on it, so I included it in a column published 24 January. I'd promised myself I wasn't going to write any more on Covid - that there was no point because nobody was listening. But I wrote it anyway. 

24 January, from an economics generalist who is spread thinly across more policy areas than I care to count. 

The country’s Covid testing system is likely to fall apart, quickly, when case numbers rise.

Testing labs can bundle five to ten samples together for testing. If none are positive, all is fine.

If the pooled sample is positive, individual samples need separate re-testing. When positivity rates are low, the system works well. But when positivity rates are high, pooled sampling stops working. Testing capacity drops to a small fraction of what it had been, just when it is most needed.

Headline figures on testing capacity may be more than a little optimistic. Contracting now for greater capacity, focusing on the saliva-based PCR testing (which identifies genetic material from the virus) that catches Omicron cases earlier, matters.

Nobody listened. I earned a couple minor insults on twitter for having right-wing reckons and undermining the government.  

Here's one genius who claims to be a physics teacher. 

And another.

Right.

Here's one conclusion of a fairly limp consultancy report into the predictable and predicted testing shambles:

A number of the reports included useful information however, the reports were not always delivered with significant contextual information, critical analysis or a call for action. The reporting style of the COVID-19 Testing and Supply Group appears to assume its audience had the requisite background knowledge and understanding to interpret the reports and recognise the significance of what was being reported. 

It also doesn’t recognise the significant non-COVID workload of some of the audience, such as the Director-General or Ministers who have significant other portfolios to attend to. This is perhaps reflective of the capacity limitations of the group in such a demanding environment. As a result, or in addition to this, the COVID-19 Testing and Supply Group relied heavily on verbal communication to provide context or convey information. 

This approach also assumed that the audience had the requisite background to the issues communicated and that they passed on that information effectively. For example, for the updates provided at the Daily IMT it was assumed that information passed on was being interpreted correctly, being noted, and actioned appropriately. 

The Ministry's modelling didn't pick up the problem. Reports through to the DG and Minister were not understood by DG Health or the Minister; the poor dears had too much on their plate to possibly be able to understand anything unless it were spoon-fed to them. 

Health is the whole job for DG Health. 

The problem was so freaking obvious that even I could see it. 

But apparently Bloomfield needed to have had his nose rubbed in it to be able to notice, and nobody rubbed his nose in it.  

I think we have to be blunt here. It takes only very basic numeracy, combined with very very basic understanding of how pooled sampling works, to be able to understand that pooled sampling cannot possibly continue working when positivity rates are high. You have to unpool and retest too often. Pooled sampling makes a ton of sense when you're looking for needles in haystacks. It cannot possibly work when every pooled sample has a very high chance of having at least one positive result. 

You literally need to know nothing other than this to know that there's going to be a problem. Figuring out the positivity rate at which it falls apart requires knowing the number of samples on a pooled tray in the systems currently in use. Knowing that there will be a problem at some positivity rate that is almost certain to obtain in an Omicron outbreak doesn't take that. 

I don't get how this wasn't completely obvious to anybody with passing familiarity with math and pooled sampling methods. 

This place...