Work packages

• WP 1.1 Advanced data integration and access
• WP 1.2 Development and implementation of cloud-based information systems

Recent key results

• Trustable data sources were identified and a cross-domain metadata standard was developed.
• Automatic data access pipelines have been written in order to permanently feed the common COCAP database
• Various QA procedures to identify errors, gaps, typos, plausibility checks, were implemented based on common standards.
• The data warehouse is made accessible through a web-API and corresponding results can be directly stored in the data warehouse.
• The infrastructure is designed to be extensible for automatic workflows for data analysis and model simulation, which will be integrated within the 2nd project phase.
  
  

Links to repositories

• Data warehouse: https://web.app.ufz.de/cocap/Germany/cases
• pyCOCAP: https://pypi.org/project/pycocap/
• Julia package: https://codebase.helmholtz.cloud/cocap/COCAPData_JuliaInterface

Principal Investigator

Work packages

• WP 2.1 Socioeconomic drivers for COVID trajectories
• WP 2.2 Statistical Methods for Interpretable Causal Effects

Recent key results

• To identify key features in excess mortality in comparison to infection numbers we compared national conditions across countries using Isomap dimensions to study differences in excess mortality and reported infection rates during the pandemic in 2020 and 2021. These two outcomes differ significantly: reported infection rates are higher in wealthier countries, while excess mortality rates are higher in middle-income countries.
• The study of socio-demographic influence on infection numbers and excess mortality at the subnational level showed that at the regional level, socio-demographic factors were not useful for predicting health outcomes due to the varying introduction times and wave patterns of the pandemic. Instead, we tracked weekly excess mortality at a local level and identified five clusters in Europe with similar patterns. This is the first empirical description of COVID-19's spread in Europe, enabling better future analysis of countries' resilience and responses. Our initial findings suggest that the timing of a country’s response is more important than socioeconomic factors, explaining conflicting conclusions in previous research.

Principal Investigator

Work packages

• WP 3.1 Coupled epidemiological-socio-economic model
• WP 3.2 Core nation case studies and rate parameter trajectories

Recent key results

• We developed a coupled epidemiological-socio-economic model and associated methods for fitting the coupled model to data via nonlinear least squares minimization (Calabrese et al. 2024). Specifically, we successfully combined an SIR-type epidemiological model with a time-varying disease transmission rate, a behavioral model capturing the autonomous response of individuals based on the average degree of risk aversion in the population, and a sector-specific economic model for Germany. The coupled model serves as the centerpiece of the COCAP project, and its development required substantial integration of efforts across the COCAP consortium. A key feature is that the model facilitates scenario analysis by transferring NPI effects from a reference nation to a focal nation. As a first proof of concept (Calabrese et al. 2024), we took Germany as our focal nation during Spring 2020, and New Zealand and Switzerland as reference nations with contrasting NPI strategies that we compared with Germany’s NPI response. Our results suggested that, while New Zealand's more aggressive strategy would have yielded modest epidemiological gains in Germany, it would have resulted in substantially higher economic costs while dramatically reducing social contacts. In contrast, Switzerland’s more lenient strategy would have prolonged the first wave in Germany, but would have also have increased relative costs. This case study served to demonstrate the potential of our approach to yield novel, multifaceted comparisons of response strategies across nations.
• A new modelling framework to assess critical characteristics of newly emerged diseases was developed (Demers et al., 2023). It overcomes the limitations of typical ordinary differential equation compartmental models by decoupling symptom status from model compartments to allow a more realistic representation of symptom onset and pre-symptomatic transmission. Looking backwards, this framework clearly shows that even the original variant of SARS-CoV-2 was extremely difficult to control, while later variants (delta and omicron) were substantially less controllable, with omicron representing a “perfect storm” of the three key factors (basic reproduction number, the mean latent period, and the degree of non-symptomatic transmission).
• A new general Bayesian estimation framework for models with dynamics that change at particular points in time represents a significant advance over existing statistical methods for simultaneously inferring both system dynamics and changepoints. First, it allows for the use of informative prior distributions, making it easier to estimate challenging parameters without treating them as fixed. Second, the Bayesian approach accounts for estimation uncertainty in all parameters, including changepoints, and provides credible intervals. Third, it identifies when significant changes in the system dynamics occurred, helping to determine which NPI changepoints truly impacted disease dynamics. Finally, though we developed this framework with our coupled epidemiological model in mind, the statistical framework itself is much more general and can be used for models covering a large range of dynamical systems.

Principal Investigator

Work packages

• WP 4.1 Development of strategies and scenarios
• WP 4.2 Evaluation of strategies and scenarios & synthesis

Recent key results

• A novel framework for analyzing and modeling behavioral adaptation was developed to cope with the observed interplay of autonomous and policy-induced adaptation, being a "moving target”. The framework describes a number of behavioral mechanisms, including "compliance," which is analyzed in detail in cooperation with the LOKI project.
  An empirical application to the German case suggested that mobility patterns changed significantly due to both autonomous and policy-induced adaptation, with potentially weaker effects over time due to decreasing risk signals, diminishing risk perceptions, and an erosion of trust in the government. These findings are discussed for their relevance to modeling, laying the foundation for constructing convincing counterfactual scenarios for strategy analysis.
• In a scenario analysis for Germany we evaluate the success of response strategies. We identified four ideal type response strategies in analysing time-dependent courses of infection rates across nations:
  (1) “Zero Covid”, which deploys NPIs with the aim to reduce transmissions to zero in a defined geographical area.
  (2) “Proactive defense”, which aims to control infection numbers at a low level by staying “ahead of the curve”, e.g., through tracing and isolating every infection.
  (3) “Reactive-discretionary”, which aims to control infection numbers at a tolerable level, while ensuring the social acceptability of measures by balancing cost of NPIs in society.
  (4) “Herd immunity”, which involves minimal intervention from policy makers, often with the aim of preserving economic activities and societal liberties.
  

Principal Investigator

COCAP aims to help countries develop effective strategies for handling future pandemics, especially during critical non-pharmaceutical intervention (NPI) phases. To achieve this, we are sharing our results with various national and international institutions and decision-makers through two main activities:

• International Data Sharing: We will provide over 40 countries with access to our comprehensive, quality-assured data and model workflows for creating intervention strategies. This will be done in collaboration with leading health organizations like the WHO and ECDC, with the first release planned for mid-2024.

• National and Sub-National Collaboration: Working with our sister project LOKI, which focuses on COVID-19 in Germany, we will make the COCAP model available at the local level. This collaboration allows for precise interventions based on detailed infection and spread models. If successful, we aim to extend this service internationally, possibly through a start-up.

These efforts will enhance each nation's ability to respond effectively to future pandemic crises.
 

• Schüler L, Calabrese JM, Attinger S (2021): Data driven high resolution modeling and spatial analyses of the COVID-19 pandemic in Germany. PLoS One. 2021 Aug 18;16(8):e0254660. https://doi.org/10.1371/journal.pone.0254660
• Demers J, Fagan WF, Potluri S, Calabrese JM (2023): Testing-isolation interventions will likely be insufficient to contain future novel disease outbreaks. In review, American J of Epidemiology. medRxiv, 2023.11. 16.23298614. https://www.medrxiv.org/content/10.1101/2023.11.16.23298614v1
• Zozmann, H., Schüler, L., Fu, X., & Gawel, E. (2024): Autonomous and policy-induced behavior change during the COVID19 pandemic: Towards understanding and modeling the interplay of behavioral adaptation. PLoS ONE 19(5): e0296145. https://doi.org/10.1371/journal.pone.0296145
• Wolffram, D.; Abbott, S.; an der Heiden, M.; Funk, S.; Günther, F.; Hailer, D.; Heyder, S.; Hotz, T.; Bracher, J. E.; Schienle, M.; u. a. (2023): PLOS Computational Biology, 19 (8), Art.-Nr.: e1011394. https://doi.org/10.1371/journal.pcbi.1011394
• Gneiting, T.; Wolffram, D.; Resin, J.; Kraus, K.; Bracher, J.; Dimitriadis, T.; Hagenmeyer, V.; Jordan, A. I.; Lerch, S.; Schienle, M.; u. a.(2023): Annual Review of Statistics and Its Application, 10. https://doi.org/10.1146/annurev-statistics-032921-020240

• Calabrese* JM, Schueler* L, Fu X, Gawel E, Zozmann H, Bumberger J, Quaas M, Wolf G, Attinger S (2024): A novel, scenario-based approach to comparing non-pharmaceutical intervention strategies across nations. In Review at J Royal Soc Interface. * co-first authors. https://doi.org/10.1101/2023.09.14.23294544

• Zozmann, H., Schüler, L., Fu, X. & Gawel, E. (2024): Autonomous and policy-induced behavior change during the COVID19 pandemic: Towards understanding and modelling the interplay of behavioral adaptation. PLoS ONE 19(5): e0296145. https://doi.org/10.1371/journal.pone.0296145
• Fu X, Fan K, Zozmann H, Schüler L, Calabrese J (2024): Simultaneous identification of changepoints and model parameters in switching dynamical systems. bioRxiv. 2024:2024-01.
  https://www.biorxiv.org/content/10.1101/2024.01.30.577909v1
• Fu* X, Schueler* L, Calabrese JM (2024). Linear Noise Approximation.jl: A Julia package for linear noise approximation of stochastic reaction networks. J Open Source Software * co-first authors. https://github.com/xiaomingfu2013/LinearNoiseApproximation.jl
• Wolf, G. (2024): Pandemics, Payments and Fiscal Policy: Lessons from Four Years after the Outbreak of Covid-19, EconPol Forum 25 (2), CESifo, Munich, 2024.
  https://www.econpol.eu/publications/forum-2024-2-big-data-based-economic-insights/pandemics-payments-and-fiscal-policy-lessons-from-four-years-after-the-outbreak-of-covid-19