The “Heinsberg Study” on Covid-19: Another Case of Insufficient Study Design?

Achim Bayer, professor at the Kanazawa Seiryo University in Japan, experiences the Corona pandemic from an Asian perspective. In this guest post he critically analyzes the German “Heinsberg Study” on the incidence of the Corona virus, compares it to figures from Austria and makes some recommendations on how to present such studies in the media.

On April 9, four expert researchers of Bonn University Hospital have presented the results of their much-anticipated study on the mortality rate of Covid-19. The study was conducted in the municipality of Gangelt belonging Heinsberg county, one of Germany’s most affected areas, in which considerable restrictions are still in effect. One of the main purposes of the study was to ascertain the rate of undetected infections with SARS-Cov-2. This rate is crucial for determining the much-disputed mortality rate among infected persons.

As a reminder: At present, the “case fatality rate” is calculated based on the number of deaths among persons tested positive for SARS-Cov-2. However, there has been some doubt whether this ratio could not be too high, considering that some persons infected with SARS-Cov-2 do not show any, or only mild, symptoms and thus see no need to take a test.

A preliminary report on the study was released on April 9. Nonetheless, the two-page document which shows a remarkable scarcity of numbers and facts. According to the researchers, Gangelt county has 12,569 inhabitants, of which 400 households participated in the study. Based on the testing for SARS-Cov-2 antibodies in blood samples, the researcher assume that about 15% of the county’s population has been infected. This leads to a case-mortality rate of 0.37 percent.(1)

Unfortunately, important numbers remain unmentioned in the document. Assuming that 15% of 12,569 inhabitants have been infected would, by inference, lead to an absolute number of 1,885 cases. If 0.37% of 1,885 persons have died, this leads to the conclusion (or rather, the guess) that there were about 7 fatalities. Furthermore, the researchers report that 600 households were randomly chosen and asked to participate in the study, but only 400 participated (the exact number of persons remains unmentioned). Since 400 is only two thirds of the randomly chosen group, it may be asked whether the motivation to participate in the study may not be due to less-random reasons. For example, it may be possible that people who have already been through the disease may be less fearful of contracting SARS-Cov-2 from the research team.

It is further remarkable that the researchers to not mention the number of formerly known cases or the assumed mortality rate before this study was conducted. Since one of the main objects of the study was to ascertain the percentage of hitherto undetected cases, the researchers’ silence on this point is rather disconcerting.

It may be worth mentioning here that another randomized study was conducted in Austria. I this case, 2,000 randomly chosen candidates were contacted among 9 million Austrians, and 1,544 agreed to participate. Here, only saliva samples were taken which can detect the acute disease Covid-19.Given that only about 9,000 of 9 million Austrians were (before the study) assumed to presently have Covid-19, it was to be expected that researchers would find only one or two active cases among the 1544 participants. There was further a high probability that the study could, produce false results if, by accident, not two but four persons were tested positive. Unsurprisingly, the study led to the result that there were “between 10,200 and 67,400” active cases in Austria, most probably 28,500. (2) These uncertainties could have easily been avoided had the study been limited to the counties which have been affected the most. For example, with a test capacity for 2,000 participants, it would have been possible to test 5% of the inhabitants in a county of 40,000 inhabitants.

The above numbers show that there is still much to be done in terms of study design and communicating with the public. It is probably due to the poor communication of test results that Business Insider, for example, reports that “Only 44 people — or 0.37% of people who lived in [Heinsberg] district — have died from the disease.” (3) Nonetheless, 44 is the number of people who died in Heinsberg district (254,322 inhabitants), of which Gangelt is only one of several municipalities.

Considering that these questions currently move not millions but billions of people, it is rather stunning that science is conducted in a way which will hardly impress the public. Only a part of shortcomings pointed out above are the responsibility of the researchers, who are experts in their fields. Rather, one gets the impression that policy makers have not yet realized the crucial importance of reliable, solid research results. At a time when billions of Euro in GNP are lost every day, it would be desirable to have ongoing, broad, well-planned and well-funded research projects on a persistent basis.


Every presentation of study results should contain at least a set of the most basic data:

  1. Total number of persons to be analyzed (cohort). For example, 9 million Austrians, 12,569 inhabitants of Gangelt.
  2. Number of samples (actual participants in study).
  3. Number of randomly chosen participants, number and percentage of those who refused to participate.
  4. Probable reasons why chosen subjects refused to participate.
  5. Probable percentage of infected persons among those who refused to participate (rough estimate).
  6. Was an incentive (such as money) offered to participants?
  7. If test were conducted per household, were persons of the same household counted differently from persons living alone? What was the key (/percentage) for this calculation?
  8. Number and percentage of persons receiving intensive care.
  9. Number and percentage of persons who died in intensive care.
  10. Estimate on the number of unreported deaths while ill.
  11. Was there a time-limit on counting SARS-Cov-19 infected persons as Covid-19 deaths (for example, 14 days after last test)?
  12. In how many cases was the infection confirmed post-mortem only?
  13. To what extent are the numbers (percentages) yielded by the randomized study different from the previously assumed numbers?
  14. How reliable is the testing method? For example, are coronavirus antibodies specific for SARS-Cov-2 or could the indicate a previous infection with another coronavirus?



Note: Title image by MiroslavaChrienova on


Die Diskussionen hier sind frei und werden grundsätzlich nicht moderiert. Gehen Sie respektvoll miteinander um, orientieren Sie sich am Thema der Blogbeiträge und vermeiden Sie Wiederholungen oder Monologe. Bei Zuwiderhandlung können Kommentare gekürzt, gelöscht und/oder die Diskussion gesperrt werden. Nähere Details finden Sie in "Über das Blog". Stephan Schleim ist studierter Philosoph und promovierter Kognitionswissenschaftler. Seit 2009 ist er an der Universität Groningen in den Niederlanden tätig, zurzeit als Assoziierter Professor für Theorie und Geschichte der Psychologie.

Schreibe einen Kommentar