Staged Alert Systems for Mitigating Epidemic and Socioeconomic Risks

Guiding policy responses to emerging and reemerging pathogen threats using optimal staged alert systems that balance community health and economic priorities through an equity lens.

Lauren Ancel Meyers, Ph.D., Professor, Departments of Integrative Biology and Statistics & Data Sciences; Director, UT COVID-19 Modeling Consortium

We will build robust, data-driven strategies for triggering non-pharmaceutical interventions to mitigate infectious disease epidemics and ensure the integrity of healthcare systems while minimizing socioeconomic harms and disparities.

During the COVID-19 pandemic, government agencies worldwide developed staged alert systems that monitor data streams and trigger changes in non-pharmaceutical interventions based on selected metrics. These systems aim to preserve both economic activity and healthcare capacity by enacting costly restrictions only when absolutely necessary. However, implemented systems vary considerably in complexity, metrics, and policy levers. France’s four-stage system uses COVID-19 incidence and ICU occupancy. New Zealand, Singapore, South Africa, and the UK have systems using three to five alert levels that track various combinations of COVID-19 incidence, hospitalizations, death, and available hospital capacity. Public-facing dashboards of alert systems rarely specify their underlying design. 

Our project will be the first to: (i) systematically evaluate staged alert systems worldwide in terms of their design, implementation, and impact on morbidity and mortality overall and across socioeconomic and racial/ethnic subgroups, (ii) build a flexible and rigorous framework for designing and validating such systems prior to and during emerging threats, and (iii) derive principled staged alert systems for mitigating COVID-19 and other epidemics post 2021.

This solution will serve the needs of communities and governments worldwide by addressing two key questions asked when facing novel epidemic threats: (i) 'What data should we track?' and (ii) 'Based on the data, when should we act?' Our staged alert system has added value in Austin as clear, consistent, and quantitative triggers allow city leaders to effectively communicate health and safety guidance to the public and help individuals, schools, businesses, and other organizations anticipate and respond appropriately to the shifting risk landscape. Moreover, it has had the desired impact of ensuring sufficient healthcare capacity to provide high standards of care for pandemic and non-pandemic needs while limiting the duration of restrictive measures. 

Looking ahead, this project will provide similar benefits to decision makers and communities worldwide as they navigate unforeseen challenges of COVID-19 and other pathogen threats.  We will provide retrospective insights into the successes and failures of implemented alert systems in 2020 and pioneer a simple and versatile approach for designing data-driven alert systems that are transparent and effective in managing competing health and economic risks. We will pilot this approach by identifying key indicators and designing robust alert systems for mitigating COVID-19 post 2021 in communities worldwide.

Our project will yield public goods stemming from each of our three key aims. First, we will publish the findings of our retrospective analysis of international staged alert systems, which will provide critical insights on the diversity and effectiveness of implemented government policies in combating the COVID-19 pandemic both for the general population and also for mitigating disparate impacts from the pandemic. Second our team has a track record of publishing open-source code software that we have produced, as referenced in Duque et al. 2020 and Yang et al. 2020, and we will publish the methodology, findings, and code stemming from the development of the modular “plug-and-play” version of our optimization framework that encompasses a wide-range of pathogens, resource constraints, possible data streams, and policies. Finally, we will develop a freely accessible dashboard that will allow individuals to compare optimized staged alert systems across a wide-range of pandemic scenarios to understand their potential implications and indirect effects on socioeconomic factors and equity. The dashboard will be key for promoting and socializing the idea of staged alert systems for the post-2021 COVID-19 response, seasonal influenza epidemics, and future pandemics.

Our solution targets communities holistically. For policymakers, optimized staged alert systems provide a data-driven and evidence-based approach to transparently respond to future epidemiological threats. Through retrospective analysis, we found Austin mortality rates may have been 50-110% higher had the city followed the trajectory of the other major cities in Texas, showcasing the framework's utility. We will host tabletop exercises for policymakers to simulate future pandemic responses to show how policies can impact pandemic trajectories. For the general public and institutions such as businesses, schools, and religious organizations, staged alert systems provide the ability to prepare and adapt to changing dynamics that are communicated by the alert system. For example, Austin-area businesses tracked hospital admissions in December 2020 and were ready when the city moved to the highest level of restrictions which reduced business capacity. Finally the results from our retrospective analysis will identify staged alert system shortcomings and pitfalls that can help inform future responses. Specifically, we will investigate how alert systems contributed to the disparate impact of the pandemic with a particular focus on vulnerable populations. We will follow this work by designing geographically resolved models that allow for targeted interventions to prevent inequity during an emerging pandemic.

We intend to provide a framework for staged alert systems that yields actionable policies for effective navigation of future pandemics, and more effective mitigation of extreme flu seasons and of disease associated with the evolution of SARS-CoV-2 variants.  We will work with colleagues around the world, including at the CDC, Oxford, and members of Trinity Challenge to disseminate our methods and alert systems and work with government bodies to tailor and implement the framework globally. Specifically we will use the next year to collect the necessary data from five countries (United Kingdom, France, New Zealand, Singapore, South Africa) to evaluate their implemented staged systems. From these findings we will produce rules of thumb for staged alert systems that we will integrate into our plug-and-play optimization framework. Finally, we anticipate that in the wake of COVID-19, countries worldwide and jurisdictions within those countries will perform tabletop exercises in preparation for future pandemics. We have worked closely with the Texas Advanced Computing Center (TACC) to produce Web-based software tools including public-facing COVID-19 Dashboards. We will develop a public-facing dashboard that can simulate several pandemic scenarios and allow policymakers and the general public to better understand how policies can shape pandemic trajectories.

First, we will measure success through broad dissemination of our findings, including direct feedback from, and communication with, government bodies that enacted staged alert systems during COVID-19. We will conduct user surveys of current and future potential users to assess the usability and interpretability of the current dashboards. Second, we will carry out retrospective analyses of Austin’s staged alert system alongside those implemented in other states and countries, which will inform our design of future systems. Specifically we will evaluate how the number of stages, policy responses, and specific thresholds contributed to preserving healthcare capacity, reducing costly days in lockdown, and reducing morbidity and mortality. Using these metrics we can quantitatively assess the impact of the staged policies to evaluate the performance of our optimized system against others used around the world. Ultimately, we will measure success through uptake of our system in jurisdictions worldwide, including engaging policymakers in testing our system through tabletop exercises and user surveys. We will leverage these relationships and our framework to help develop staged alert guidelines for post-2021 COVID-19 surges as vaccinations progress and variants continue to emerge, which will provide more real world confirmation of the impact staged alert systems can have.

From our experience with the prototype system in Austin, we summarize barriers in four points: (1) Persuading decision makers to adopt new technology is a challenge, especially if there is comfort with existing practice. Adopting a new solution requires investment of people and time in overcoming a learning curve. (2) Coordination between different stakeholders can be difficult. For COVID-19, there have been continual clashes between the City of Austin and the State of Texas, which have differed in evaluation of the pandemic and the corresponding objectives. (3) It can be challenging to collect necessary data from healthcare agencies. For example, different hospital systems may not share their capacity with the city. (4) The policy may not yield desired results upon implementation in different parts of the world due to different cultural, religious, and behavioral reasons. Engagement in low-stakes tabletop exercises provides a means to overcome (1). In this way, building confidence in the system among a subset of stakeholders provides a platform for conversations involving (2) and motivation for government leaders to engage area hospitals to overcome (3). Ultimately, success with (4) hinges on city and public health leaders socializing the system in a real-time environment.

1. The University of Texas  at Austin
2. Santa Fe Institute
3. Northwestern University
4. The Chinese University of Hong Kong, Shenzhen
5. Oxford University, Coronavirus Government Response Tracker

The primary barrier we face will be access and buy-in of policymakers towards our proposed solution. While we have a pilot solution with the City of Austin, scaling our solution and expansion to other regions will not be possible without working with policymakers around the world. We have developed a network of policymakers and public health officials within the United States, but believe that the Trinity Challenge will provide an opportunity to develop our framework into a truly international solution for future pandemics. Access to the founding members provides a truly global group of organizations that can help make inroads with policymakers around the world. Scaling our solution will also take more research to develop a flexible optimization framework, computational power to generate global solutions, and technical expertise to develop accessible dashboards for policymakers and the public alike. The resources made available through the challenge will fund the expertise necessary for scaling and expanding into other jurisdictions.

We are happy to partner with any of the Trinity Challenge Member organizations. We have long and ongoing relationships with researchers at Northeastern University and The University of Hong Kong, so we would find it useful to partner with our collaborators at those institutions to iterate on our solution. We also believe that international foundations, specifically the Bill and Melinda Gates Foundation could provide an important partnership so that we can harness their expertise in interacting with policymakers around the globe. We would also be interested in  partnering with an organization with economic expertise such as the London School of Economics and Political Science, so that we can properly account for the full economic costs of proposed staged alert systems. Behavioral Insights would also be a strong partner to bridge the science of the alert system with its implementation and adoption. Finally, we plan to continue our current relationship with the Oxford Government Policy Tracker group to generate policy indicators around the globe which can inform our analysis.