Hacker News Reader

Comment by ttul | Hacker News Reader

I would love to know how Germany has been able to discover a sufficiently accurate antibody assay before many excellent groups in other countries. Finding some antibodies is one thing; making it mean something is another. Does the presence of a particular antibody imply that someone is now no longer infectious to others? Are they themselves now immune to reinfection?

These questions are unanswered AFAIK so I am curious as to what the Germans are actually learning.

View on HN

To my knowledge, the tests that are carried out in Munich use a so-called ELISA device. My vague understanding is that this tests reacts with a specific protein of SARS-Cov2 and has no known cross-reaction with other harmless corona viruses. Disadvantage of the test is probably that it is quite complex and not much capacity is available. Here in Munich, however, about 3000 households are now randomly selected and tested in order to estimate the real number of infected persons. This is the press release of the munich university. DeepL.com works great to translate german texts to english: https://www.lmu-klinikum.de/aktuelles/pressemitteilungen/mun...

My vague understanding is, that the tests can also test positive for antibodies from other similar viruses. They have a significant false-positive rate that you should not rely on a positive („you are immune“) result but overall they will provide a big picture approximation how many citizens might have already resistance. They are not good enough for a rumored „immune certification“.

Currently we have no idea how many people have had covid. We have a good idea how many have died, but no country has managed to do the kind of testing that would give a good indication of how many people have been infected. Do 2% of people getting it die, or 0.2%, or 0.02%? Testing 1 million people at random across the country will give good results of not only how deadly it is, but also how likely it is to spread in different areas of the country.

We have a pretty good idea it isn't 0.02% because there's an upper bound on the infection rate (around 75%) and the number of bodies is known and in all cases exceeds 0.3% of the maximum possible infections.

How did you come up with 0.3% number of bodies / maximum possible infections, aka total population? I'm trying to wrap my head around this whole mess...

Take Lombardy, hardest hit region in Europe. 1130 deaths / million, or about 0.1% death rate. All German states are under 100 deaths per million, which is less than 0.01% death rate. Sanity check: Bayern, 1.2k deaths/13M people ~ 0.01%.

Of course, this is a lower bound estimate assuming all people in a given region have got through the disease. Which is definitely not true. To figure out how bad is going to get we need the total number of infected people. Hopefully the random testing strategy will shed some light on that number. But at least we start [much] lower than 0.3%.

https://covid19.quersive.com/chart?c=M&e=IT-Lombardy&e=DE-BW...

In NYC, 0.16% of the population's died from it, so that's a hard lower bound.

Thanks!

> They are not good enough for a rumored „immune certification“.

And how about being tested negative, and a few weeks later being tested positive? I suppose that would improve the accuracy.

I can't comment on the test itself, but there is a wider point here: don't test individuals, test populations.

Most countries are only testing people who arrive at hospital with serious conditions (and billionaires and nba players!?). That's pointless because the treatment is the same whether they have COVID-19 or some other respiratory infection.

If we tested populations and did so regularly, you would know whether lockdown was working, whether you could afford to partly lift it and how and when to reinstate it.

Right now, we are all sitting at home with no idea if this is working or not. In a few weeks, we will start roping thing with no idea whether that is a good idea or not. We'll just guess.

You can do the tests before you know how to interpret the results...

Also, when you currently have no knowledge, even potentially inaccurate guesses are a big upgrade...

Eh not really, these aren’t measurements these are for the most part binary tests the current assays seem to trigger on AB’s for multiple coronavirus strains and even worse other common viruses.

Without knowing the specificity and sensitivity of this specific test you can’t say any “inaccurate guesses” based on it is an upgrade.

Say the test has really poor sensitivity but 100% specificity then sure it might not identify a lot of past infections but it won’t produce false positives so which ever figures come out you can extrapolate of them fairly easily.

If it has terrible specificity then it’s a whole other problem unless you have additional tests that have high specificity that you can use to exclude false positives.

But if you want to know "what percentage of my population have had coronavirus in each town and age group", then these tests, combined with research to find out the sensitivity and selectivity, will answer that question.

Most of the big decisions governments have to make depend on only aggregate statistics after all...

AB testing yes, this specific test maybe which is why the GP was asking about the assay.

The current AB tests were found to be insufficient.

Reply on news.ycombinator.com