CrisisEngine: delivering rapid responses to global health emergencies

CrisisEngine is a computational system using real-time data to generate multiple simulations, modelling policy interventions in global health crises.

An agent-based platform, it tests potential responses, enabling rapid mitigation and minimising adverse consequences. CrisisEngine proved itself during a 2020 COVID-19 outbreak in Melbourne, Australia and is now ready for scale-up.

Dr Jason Thompson

Senior Research Fellow

Transport Health and Urban Design Research Laboratory

Melbourne School of Design

The University of Melbourne

The past 12 months have highlighted that in countries where public health decision-making has been poor, compromised, or indecisive, social and public health outcomes have been disastrous.

Implementation of effective policy has been hampered by lack of scientific evidence that is timely and available to decision-makers in a form that can be readily translated into policy and action.

COVID-19 has highlighted a failure of the science/policy interface – with interactions between scientists and decision-makers struggling to keep pace with the urgency, complexity, and dynamism of the crisis at hand.

Decision-makers facing novel, emergent public health threats must respond quickly to support population safety and recovery. Reflective examination of observed outcomes, implementation of trials and systematic reviews through ‘normal’ scientific processes take time and risk delaying the roll-out of effective control measures.

There is currently no universally accessible, authorised, decision-support system informed by transdisciplinary collaboration and community engagement which promotes optimal policy implementation.

Our vision is to inform better decision-making amid unfolding public-health crises by scaling up and delivering an agent-based technology platform that can easily access and integrate big data while synthesising multi-layers of information. 

Our solution will provide evidence that keeps pace with the ever-changing world.

CrisisEngine is a platform for civic leaders from local to national levels. These might be senior policy-makers in government public health units, or municipal officers in remote townships.

Its friendly user interface and flexible computing requirements mean that dynamic crisis management options are available to anyone who needs them. Open software and public datasets allow colleagues to share progress, so that no idea is left unexplored or unused.

CrisisEngine directly addresses the needs of decision-makers to be better informed about the likely consequences their actions will have on population health and welfare.

Our team is already collaborating with a range of city-based and international networks and knowledge hubs. We have a close working relationship with health authorities and State Government in Victoria, Australia, arising from the CrisisEngine input into COVID-19 responses.

Success in this area has triggered invitations to consult and collaborate with public health units in other states and overseas.

We are working to expand these connections, which are critical to understanding the unique needs and challenges of potential client agencies.  

We are also building a global ‘Challenge Champions’ network , outlined below.

The public good provided by CrisisEngine is its support for real-time policy decision-making grounded on rapidly changing evidence. In the face of a global health emergency, the system supports decision-makers’ best efforts to secure societal health, wellbeing and economic prosperity of their populations.

CrisisEngine puts the tools for better decision-making and crisis resolution free-of-charge in the hands of policy-makers. It also puts knowledge in the hands of citizens, enhancing the capacity of societies to chart optimal pathways through crises.

It produces scientific information that can be easily adapted for the public, providing objective, transparent evidence to explain potentially controversial mitigation or prevention policy actions.

Additionally, and contrary to some other approaches, our model benefits from, and is built to withstand, disparate perspectives and interrogation. It does not promote any ideological perspective or goal (for instance, in coronavirus terms, elimination versus suppression), but, importantly, can be tuned to generate strategies that achieve desired outcomes (zero cases, for instance). Established as an open-source effort, we welcome feedback from the global health and scientific community to ensure we can continually improve the model and ensure its relevance.

In August 2020, the 6.7 million people living in the Australian state of Victoria experienced a second wave of COVID-19 transmission that produced an exponential increase in cases and resulted in the enforcement of strict ‘Stage-4’ lockdowns.

CrisisEngine1 modelling demonstrated the likely efficacy of strategies that would curb the spread of new infections, but it was its use in strategising the exit from lockdown where it was potentially most valuable. It showed the government how case numbers could get low – and stay low – for the long-term.

The modelling enabled Victoria’s Premier and Chief Health Officer to explain the risks, costs and long-term benefits of extended public health measures, stimulating widespread support to reach minimal community transmission and easing the political process.

The Victorian outbreak was controlled within four months and has not thus far re-surfaced. Comparisons with other Europe and US suggest thousands of lives were saved.

Once extended and made available for use throughout the world, CrisisEngine2 will have a large and direct impact on populations within each jurisdiction using our solution, and, because of the connected world, a potential impact on global citizens in all countries.

In the first year, we plan to advance model design, construction and deployment across high performance computing clusters to improve the speed of models, the solution space they can explore, and the scope of health crises they can address.

Across the first three years we will develop a global network of institutions and individuals – named the Challenge Champions network – that will amplify the rollout of the system and demonstrate its value.

The Challenge Champion network will build on approaches and expertise already mobilised by the University of Melbourne. It will look to existing data-intensive city networks to effect knowledge, models and science-policy interaction capacity across different countries. This will aid capacity, data exchange and validation, and enhance public-policy discourse.

We will also engage with health and policy data observatories, partnering with key science advice institutions able to support both modelling and policy engagement as 'translation hubs' for the program.

CrisisEngine1 allowed policy-makers to explore a huge range of possible scenarios, far beyond what was previously possible and at a much faster pace. CrisisEngine2 will take us to our ultimate goal of ‘real time’ policy modelling and decision support. 

Impact for target populations will be monitored in three ways.

First, we will measure the depth and integration of the Challenge Champions network across 20 governments. We will invite and embed Challenge Champions in individual countries, and health jurisdictions in each country. We will then monitor levels of geographic and population saturation of the network, and depth of engagement of individual representatives. Ongoing engagement will be critical to maintaining endorsed Challenge Champion status.

Next, we will monitor the use of the modelling framework in each health jurisdiction and ask Challenge Champions to report on the utilisation and translation of modelling advice into policy and interventions. This data collection exercise will feed back into a central repository of reports and case studies for use by other countries, jurisdictions, and members of the network.

Finally, we will monitor official health outcomes and mitigation measures reported in each country through scraping existing, standardised global health datasets, such as Institute for Health Metrics and Evaluation, WHO Global Health Data Observatory, The Oxford Stringency Index, and individual country and jurisdictional reports.

Potential barriers CrisisEngine faces are:

Financial: Most of our work to date has been unfunded or minimally supported through government, development banks, or philanthropic sources. We seek funding to scale up the central architecture of the solution so that it can be deployed globally in a consistent, replicable, and robust fashion.

Technical: Our solution for Year 3 relies on the successful transfer of our existing agent-based platform into the Everest CrisisEngine2 framework, and the provision of large-scale computing resources to support global use. We believe that with additional resources and effective partnerships we will be able to easily overcome these potential barriers to progress.

Human: Scaling up relies on the establishment and maintenance of the Challenge Champions global network, which will be supported to use and advocate for CrisisEngine2. It is crucial that the CrisisEngine solution move beyond a technical tool to one that reflects and adapts to the local context, and has the instrumental, conceptual and political utility described above.  With the support of Trinity Challenge members, and our effective delivery of a valuable product, we will be well able to overcome any potential challenges relating to this.

The Behavioural Architects

The goals of the Trinity Challenge are exactly aligned with ours. We too are a global, collaborative effort, with widely respected contributors and partners from across academia, philanthropy, and the private sector. The Challenge can connect our team to a vast array of international interdisciplinary experts. This is vital to our vision. The long-term view of The Trinity Challenge – to help countries develop a means for responding to current and future health crises – is also central to CrisisEngine.

The legitimacy and international visibility of our partnership with The Trinity Challenge will assist to attract expertise to solve potential technical challenges associated with the project when we move into the CrisisEngine2 development phase.

The Trinity Challenge’s profile and presence will assist in building the ‘Challenge Champions’ network, which is central to our goal of reaching as many countries and health jurisdictions as possible.

The start-up financial assistance to develop the CrisisEngine2 system and its network architecture is crucial.

Our team is committed to maintaining the ‘ownership’ of this solution in the public domain and for a public good. The Trinity Challenge’s commitment to similar principles means we are excited to collaborate.

There are three levels of partners we will engage with:

• development,
• end-use, and
• supporting.

Development partners will be networks, institutions, research laboratories, data warehouses and observatories, policy experts, businesses, students and individuals who can provide valuable input and expertise into the ongoing development of the CrisisEngine solution.

End-user partners will consist of members of the Challenge Champions network. At three years, we expect coverage in 20 different contexts with 20 discrete Challenge Champions. The next step will then be an expansion to potentially thousands of users. At that stage the development of new partners, new users and new policy support modelling products will become the focus of our growth activity. End users will then include all levels of local, state, and national governments.

Supporting partners will include multinational funding agencies and the sponsors of international networks, who recognise the value of our solution and can provide ongoing support for the CrisisEngine framework and the Challenge Champion network.