Modell validation based on immunity-related positive testings

Note: This is an automated translation (using DeepL) of the original German article.

Recent, new data sources provide an update on the evaluation of current system behavior in the COVID-19 crisis. Another important aspect is the verification of the model goodness of the Austrian population model and the application to the COVID-19 spread. The so-called validation, i.e. the comparison between model results and reality, is the be-all and end-all of model building and a continuous process, which is supported in particular by the availability of new data sources. Within the last few months, the recording of reinfections in the Epidemiological Reporting System (EMS) as well as new data from the electronic vaccination record have provided additional information. With the help of focused data analysis and comparisons to model results, these data sources can now be used to test existing hypotheses on the protective effect of recovery and vaccination and to formulate new hypotheses.

The current study addresses two core questions:

  • How long does immune protection against (confirmed) reinfection last after recovery?
  • How effective is a partially/completely completed vaccination series with respect to the vaccine manufacturer against a confirmed infection?

The aim of the study is to analyze (1) how well (compared to real data) the agent-based COVID19 model of the Vienna University of Technology and dwh GmbH already mapped the respective protective effects before incorporating the new data and (2) how to improve the mapping.

The study allows for three key conclusions: First, the calibrated simulation results show that recovery probably provides protection from reinfection for longer on average than the 180 days previously considered in the "green passport." This in no way represents a change in the position that vaccination is recommended and useful even for convalescents. Rather, the result serves as a hypothesis to be tested by physicians in order to better understand the complex interactions. Especially in the communication with vaccine-critical groups of people, it seems important to create maximum transparency in all directions in order not to have to react a posteriori.

On the other hand, evaluations of vaccine effectiveness indicate that both partially and fully completed vaccination series show very manufacturer-dependent protective effects against confirmed new infections. This shows an unexpectedly low protective effect of partially completed vaccination series as well as a comparatively low effectiveness of vaccinations from manufacturer Johnson&Johnson/Janssen. For the latter, the data show a vaccine efficacy of at most 50%, which is significantly lower than the values of the other fully completed vaccination series.

Third, the strongest protection against confirmed new infection is seen in at least singly vaccinated and recovered, which should be taken as an unambiguous vaccination recommendation for recovered, and doubly vaccinated.

The study is particularly limited by data quality of the E-vaccine passport and EMS, as well as model uncertainty in the distribution function for immunity protection after recovery. In addition, the conclusions of the study are linked to current epidemic events and, in particular, the currently prevalent viral variant.


Detailled description of project incl. first results. German version, Sept. 28th 2021 (corrected 30.9.2021)