Complexity in epidemiological data

>> One of the most iconic features of the corona pandemic was the dashboard. The clear graphical representation of epidemiological data suggested to users a scientific and objective view of the infection process. On the screen you found yourself in the role of the pilot in the pandemic cockpit.
Precisely because of their accessibility, such depictions of the infection process proved to be a double-edged sword. Visualizations are an effective means of depicting complex data as faithfully as possible, depending on the purpose, or making it more usable by simplifying it. However, the prerequisite for this is that the addressee understands the underlying theory and phenomenology in order to be able to utilize the information presented to them. As a doctor, just think of an ECG or ultrasound image, which remains a closed book for the patient.
In order to understand the observed system (infection process, cardiovascular system, nuclear reactor, etc.) and ideally be able to control it, you need an adequate mental model. Otherwise, every attempt at control becomes haphazard experimentation and, in the worst case, produces disastrous results. Situations are particularly dangerous where a system behaves paradoxically and reacts hardly or even counterintuitively to control impulses. A pilot must e.g. B. know that if the flight is too slow “behind the performance curve” he is not allowed to raise the nose of the aircraft to initiate a climb - in the worst case, the aircraft will fall to the ground like a stone after a stall.
The mental model for interpreting epidemiological data has mostly remained extremely simple during the pandemic. It essentially included the (exponential) continuation of trends and an instantaneous, linear cause-effect model - more con-
clocks, more infections and so on. Now, as is so often the case, this model is not entirely wrong in a narrowly defined context, but it does have its limits. If additional factors (time, space, environmental influences, host characteristics, etc.) are taken into account, a complex dynamic emerges that can be systematically misunderstood if one does not have sufficiently granular data and, above all, an adequate understanding of theory. Of course, granular data alone is not a solution to this challenge; unfiltered, fine-grained information can also be misleading.
Such misunderstandings have abounded over the course of the pandemic and, arguably avoidably, contributed to debate polarization and suboptimal governance. The transmission protection of Covid vaccinations can serve as an example. When 2G regulations were introduced at the end of 2021, supposed confirmation was quickly found in current epidemiological data, for example: B. in the study by Maier et al. (1).
In the summer of 2022 the situation was reversed; The RKI figures showed negative vaccination effectiveness. The reaction there was remarkably helpless and tight-lipped by initially simply cutting off the vaccination effectiveness at zero and, when asked by “Welt” (2), explaining away the negative values as an “expression of statistical uncertainty”.
With the relevant theoretical and empirical knowledge (3), one would not have been surprised. Based on early available foreign data on vaccination effectiveness and epidemiological textbook knowledge, it was foreseeable that Covid infections prevented by vaccination in the short term, combined with low or negative vaccination effectiveness, would have to be “made up for”, albeit with the advantage of a significantly reduced risk of severe disease. The medium-term failure of 2G regulations was therefore predictable from the start. Of course, this was initially only a theoretical expectation, but it could have been validated and refined with the help of unbiased data and appropriate analysis, e.g. B. looks at the cumulative risk of infection and excludes previous infections as a confounding factor.
Collateral damage from the slalom of knowledge: social stigmatization and unrest caused by 2G and, ultimately, problems of justification in the face of new, only apparently surprising data.
Why such aberrations? It would be too easy to attribute misjudgments of the pandemic years to wishful thinking or even a conscious agenda. It seems more likely that several factors interact in a self-reinforcing manner. In this specific case, the following arise:
1. Insufficient awareness of the system's own dynamics with the “pendulum swinging back” in vaccination effectiveness
2. In view of the increasing number of infections, a narrowed focus on specific data on transmission protection while ignoring the “time” dimension
3. Ignoring available data from abroad
4. Data from Germany distorted by testing criteria, making it difficult to correct misconceptions and further develop understanding of infection dynamics.
Only a systematic debriefing can establish exactly why, as is often criticized, high-quality epidemiological data is often missing in Germany or wrong conclusions are drawn from available, often distorted data. You also have to think beyond surveillance instruments and improved technical solutions (digitalization). Seroprevalence studies such as MusPAD or
Immunbridge or the cluster analysis of Covid risk factors to determine vaccination prioritization by STIKO show: In principle, there is no lack of scientific competence and technical possibilities for generating and analyzing meaningful epidemiological data. The question is therefore: Was Germany's data problem also hiding a deficit in the collective understanding of the situation? For the lessons learned from the pandemic, one should not be satisfied with simple answers. <<