Pandemic forecasts using real-time contact data

We use privacy-preserving and real-time contact and mobility data for a spatially-resolved outbreak analysis tool. It will inform the general public and allow targeted actions to curb future pandemics.

Dr. Sten Rüdiger, NET CHECK GmbH

The resurgence of COVID-19 cases in Europe and across the globe has challenged our resilience despite the rolling out of vaccination, limited contacts, and reduced socio-cultural interaction. Finding the right balance between economic sustainability and the tolerance of the healthcare system is challenging and requires targeted intervention. The speed and magnitude of the outbreak are influenced by the social contact pattern. Therefore, we need to understand the social contact and mobility patterns from the microscale to the global  level in order to identify major contact hubs and plan accordingly. Mobile networks and cellphones can be useful in this context to identify contacts on an individual level to track infection events as well as at a statistical level to measure social distancing behavior. Privacy risks, however, limit the adoption of proposed solutions. We want to build a privacy-first method that allows us to quickly identify risky social behavior at the population level and to fine-tune non-pharmaceutical interventions. We see an urgent need  for system that can be installed on millions of phones worldwide, record contacts between phones by GPS, use encryption methods for maximum safety, and a real-time dashboard for social contact behavior including a forecast of outbreaks.

The main purpose of NPIs in epidemics imposed by the governments is to influence social behavior, and consequently, disease transmission. Social behavior is a very complex phenomenon with dynamic and adaptive characteristics, and is highly diverse among individuals based on their personality and needs. Therefore, in the absence of sufficient information, certain assumptions on social behavior are made for designing NPIs. As the NPIs were partially successful in influencing social contacts and restricting the COVID-19 epidemic, for example in Germany, it is becoming more clear that assumptions on social behavior have to be revised by relying on further data.

Our solution offers a privacy-preserving data collection and analysis pipeline to report current and (short-term) future epidemic status and warning signals at different spatial scales. Such information is needed by governmental decision makers to adapt state- and country-wise NPIs, as well as application users and the general public to adapt their social contact and mobility in risky times and locations such as supermarket.

Our solution will benefit society in multiple ways. Aggregated and processed data will be made publicly available. Scientific understanding of contact patterns across the regions and different age groups will have broad academic interest that will result in peer-review publication. Disease propagation in networks is a well discussed topic in the scientific community. Collected data will enrich and strengthen the existing knowledge.

The data will also be used to create a user friendly dashboard. Given the city, highest contact prone zones will be identified and marked with a timeline. This will help policy makers in designing strategy and in casting interventions. A user can also be benefitted from this knowledge by avoiding crowded situations.   

We will be able to provide a full pipeline for disaster management for diseases and other events where contact data is useful. We obtain the data from cell phone apps and will analyze it for outbreak warning and NPI design. With the help from the Challenge we intend to make this solution available as public good for as many countries as possible.

The development of an outbreak is a highly complex process and is usually associated with the behavior of different age groups, demographic structure, contact frequencies, fraction of risk group etc. Control strategies, therefore should be designed accordingly. Mobility and contact data will reflect the microstructure of the society and are useful in understanding social behavior at different settings. 

Our solution will evaluate current and (short-term) future epidemic status and can be used to warn individuals to avert high risk zones. Such information is necessary to adapt state- and country-wise NPIs and are useful to the policy makers. We stress that our solution will be available to the public. The information resulting from the data-guided modeling will be made available in the data-collecting cell-phone app (“Contact guard”). Individual users can use this information to adapt their social contact and mobility in risky times and locations such as supermarkets. It is important to the success of the app and project that users will receive an incentive for their contribution.

Outbreaks of respiratory illness are common and are controlled by many factors (e.g., environmental, societal). The planning and preparedness of a pandemic should include settings that deliver the highest control over infection spread with minimal intervention. This requires comprehensive monitoring of social contact and mobility data with high resolution. We seek to develop a privacy-preserved software and analytical pipeline that collects and analyzes age stratified contact and mobility data. This will help to identify contact rates across age groups in different places (e.g, school, work, restaurants, supermarkets etc) with different seasonality settings (public holidays, festival, summer break etc.).  Integrating this information into existing epidemiological models will be able to assess the risk in real-time and help in planning an effective control strategy. This will also alert the user by notifying how risky a facility is to avoid a high risk crowded situation. We already have demonstrated this in Germany where we frequently informed the public about contact numbers. This method is general and is not specific to any particular virus type and can be easily extended to any future respiratory outbreaks and co-circulation of viruses (e.g. influenza and COVID-19) situation.

Mobility and contact data will be collected and managed at NET CHECK and be provisioned to the team in Braunschweig. Our success in data collection will be measured by user numbers and GPS location samples. Quality will be judged by statistics of the data and compared to the results obtained so far.

Our epidemiological model has been currently under testing in the egePan Unimed consortium (https://www.netzwerk-universitaetsmedizin.de/projekte/egepan-unimed and http://egepan.de/, https://www.aerzteblatt.de/archiv/217457). The purpose of this consortium is: ‘Development, testing and implementation of regionally adaptive care structures and processes for evidence-based pandemic management coordinated by the university medicine’. Taking Hessen, a state in Germany and some cities within Hessen (Frankfurt, Kassel, Fulda, Marburg, Darmstadt, wiesbaden) into account, participants of egePan Unimed evaluate data, forecast short-term bed requirements (hospital and ICU) for COVID-19 patients in weekly basis. This study also involves the retrospective analysis of historical data (since August 1, 2020) to evaluate the efficacy.

Integrating mobility and contact data into the epidemiological model will improve the overall quality of the prediction. Given the interventions, it will then be used to analyse the historical data to evaluate the overall performance.

The work of our two groups in Berlin and Braunschweig has made large steps in the recent year related to the employment of mobility data and epidemiological simulations. Nevertheless we are facing barriers partly on the side of data generation and partly on the effort to build up-to-date models for pandemics. 

Contact and mobility data has been collected by NET CHECK but only for the German market. To employ and improve the methodology we need data from global sources including the British and US market which could be provided by other companies including partners of the challenge. Furthermore, there is an urgent need for technological advances in encryption of mobility and contact data as well. Privacy issues are a pressing topic with companies including Google Inc.  limiting the use of GPS data in the latest Android OS versions. 

Additionally the team will be addressing the real-time link of contact data with computational models. This advancement in infrastructure will be a highly challenging barrier and help in financial resources as well as guidance from the partners is needed. Resources of the Braunschweig team are fully employed for COVID-19 related projects. However, due to the urgency and multidimensional challenges of the current pandemic and the role of Braunschweig team/HZI in analyzing/reporting the pandemic in Germany, the team is highly understaffed for future development and requires hiring experienced personnel.

NET CHECK GmbH

Helmholtz-Centre for Infection Research GmbH

Epidemic outbreaks start from local events but challenge people worldwide. Epidemics are international problems that demand solutions at the same level. Given the many challenges that COVID-19 has caused, including the rolling out and distribution of COVID-19 vaccines, it is becoming evident that the national interest of developed countries may become the first priority, neglecting the majority of the population in such a crisis. Our proposed solution that is potentially accessible worldwide can highly benefit from help of teams and collaborations with international interests, which we believe that the Trinity challenge is able to provide.

Our proposed solution will gather data from developed countries, as well as from developing countries, which will represent the micro-structure of the society and is extremely important to understand human behavior. Our method will upscale the use of data and analytics that will improve global public health by better learning the ecosystem of a society at its root level, which is a major focus of Trinity challenge. 

We would welcome any partnership with organisations providing data for our solution. For instance, Cuebiq Inc or other mobility data companies could possibly provide cell phone locations in the large scale for countries that we do not cover yet, but for which our method could be useful (including developing countries). The privacy rights need to be clarified here, but data without home and work location, that is offered already by Cuebiq Inc., could legally be provided with high space-time accuracy.

Furthermore, early-warning systems can also include symptoms data which is e.g. offered by Facebook Inc.