Joachim Burghartz is the director of the Institut für Mikroelektronik Stuttgart (IMS Chips) and the former director of Dimes at Delft University of Technology.

4 November 2020

I’m back in school thanks to corona. I’m attending live lectures in biology, medicine, pharmacy, statistics, law, psychology, social sciences, public media, material sciences and more. Why? Because our freedom is currently severely impacted by precautionary measures imposed by our political leadership. Those measures are based on figures and numbers. As an engineer, I feel obligated to verify those. I looked at some of those figures in Germany. They raised questions, which I’d like to share with the readers of Bits&Chips.

Let’s take a look at the corona app. After its release in June, about 17.5 million Germans downloaded it, or 20 percent of the German population, when not considering multiple downloads. Beware, these are just downloads; it’s not actual usage. While the German Robert-Koch-Institute (RKI) doesn’t publish about that, we know from the similar Swisscovid app that only about 60 percent of the downloads is being used. This lowers the number of likely corona app users in Germany to 12 percent.

According to the Oxford Study, app usage should be at least 15 percent to have any noticeable effect. But even at 20 percent, the reproduction factor R could only be lowered by 0.08, which is within noise level. Is the low usage a matter of lack in public spirit and discipline? No, it’s not. Only 60 percent of the Germans has a smartphone capable of running the app. Most people above age 65 don’t own a smartphone. Why did our political leaders not consider this before pouring millions into the corona app? Weren’t the seniors the ones we wanted to protect from that virus?

I also learned that the corona PCR test results, which are used to back current political decisions, may not be reliable. One problem is that the test isn’t free of error. Even under lab conditions, sensitivity (ie the percentage of infected people correctly identified as such) ranges between 97.7 and 98.8 percent and test selectivity (ie the percentage of healthy correctly identified as such) is about 98.6 percent. In real life, the effective sensitivity is 70 percent and selectivity is 95 percent, says Dagmar Lühmann from UKE in Hamburg.

At a low prevalence of infected persons among the population, this creates issues. The risk threshold for precautionary measures in Germany is 50 within 100,000. With a sensitivity of 70 percent, only 35 of the 50 infected people will be tested positive, while for 15 of them, the test will give a false negative. Those 15 actually infected ones would feel secure and be a risk for others to get infected. With the selectivity of 95 percent, 4,997 of the 99,950 non-infected people will falsely be tested positive and be quarantined for no reason.

The figure of merit of the PCR test (essentially, expressing the usefulness of the test) is the ratio of the 35 predicted cases out of the pool of 50 actually infected ones to the 5,032 total predictions of positives. For real-life conditions, that figure is as low as 0.7 percent, and even under lab conditions, it only improves to 4.7 percent, while obviously 100 percent is the target.

This relates to the low prevalence of 50 out of 100,000. Hendrik Streeck, a virologist in Bonn, the initiator of the Heinsberg Study, already pointed this out in April. The high prevalence of 15.5 percent in the Heinsberg region close to the province of Limburg pushed the PCR figure of merit to 94.7 percent.

My conclusion is that the current intense testing of people returning home from vacation, though without any symptoms, not only doesn’t make any sense but even leads to wrong and misleading conclusions. The focus must be on the ill and hospitalized people, says Streeck.

So, my appeal to you as my colleagues, who are used to deal with numbers and facts in your daily engineering work, is to continuously look at those corona figures and explain them to your relatives, friends and neighbors, so that we all can participate in a democratic debate, always leaving room for different opinions, but never for misleading numbers and figures.