Scaling the ValenciaIA4COVID Pandemic Decision-Support Platform (DSP)

Policymakers facing trade-offs between public health and socioeconomic impacts will make evidence-based decisions using the ValenciaIA4COVID Pandemic Decision-Support Platform, an award-winning approach leveraging multidimensional data sources (official statistics, mobile data, non-pharmaceutical interventions, surveys, etc.) to enable mobility and epidemiological modeling, healthcare burden predictions, situational awareness and an AI-powered intervention prescriptor. 

Dr. Nuria Oliver; MIT PhD; IEEE/ACM/EurAI Fellow; cofounder and vicepresident of ELLIS, Commissioner to the President of the Valencian Government on AI; Chief Data Scientist in Data-Pop Alliance (https://en.wikipedia.org/wiki/... 

The COVID-19 pandemic has highlighted once more the importance of basing public health and other emergency response decisions on sound evidence and solid predictions, for example on the number of infections under different scenarios. The staggering numbers of confirmed COVID-19 cases in many regions of the world have put unprecedented pressure on their health care systems. Thus, most countries in the world have resorted to implementing non-pharmaceutical interventions (NPIs) --e.g. school and workplace closings, restrictions on gatherings and travel-- to reduce human mobility and social interactions, with significant social and economic costs. Navigating these trade-offs requires principled approaches to both reliably predict the number of confirmed infections under different NPI policies and to recommend NPI regimes that would achieve the optimal balance between their socio-economic cost and the resulting number of infections. This global problem is unsolved to date due to a lack of scientifically sound methods to accurately predict the impact of NPIs on a population and of granular data sources to assess the social and economic impacts of the NPIs. With today's data and analytical capacities solving this problem is feasible, and will contribute to the necessary transition to more evidence-driven policy-making, particularly during a pandemic. 

Our solution serves mainly policy makers at both national and sub-national levels. Business owners, NGOs and citizens might also benefit from the platform by enhancing their ability to plan for present and future pandemic situations, including different Non-Pharmaceutical Interventions, on the basis of evidence. 

We aim to serve stakeholders in several countries around the world with a special focus on Latin America, by scaling our successful collaboration with the President of the Valencian Government in Spain, which we have assisted since March 2020 in their policy-making via Data Science and AI. 

Our collaboration has involved daily interactions and reports, especially with the General Director for Public Policies, on the predictions of the number of COVID-19 cases, hospitalizations (general and ICUs), and deaths; analyses of the outbreaks; simulations of the impact of contact tracing and vaccinations and ad-hoc reports about relevant topics, such as the role of children in the transmission of the disease and the limitations of digital contact tracing. 

This experience has given us a deep understanding of the needs of a government navigating difficult trade-offs during the crisis, which we will combine with deep local knowledge of other geographies to develop equally successful collaborations. 

The main public good contribution of our work is to enable policymakers to make evidence-based and science-driven decisions and to quantitatively evaluate the impact of their decisions in a pandemic. 

Additionally, we contribute to increasing transparency on data access and data-driven justifications for public policies that affect millions of citizens. 

Other public value of our platform includes: 

- Intuitive visualizations and dashboards of mobility, epidemiological and healthcare occupancy predictions, citizen science and recommendations of non-pharmaceutical interventions, which are valuable not only to policy makers but also to business owners, non-profit institutions and citizens.

- The citizen survey data will also be available to citizens, so that they can understand the reasons behind the decisions adopted by their governments. We are paving the way forward towards transparent public policy decision making.

- Dissemination and outreach: we have written over 100 reports/papers, have contributed to over 70 media articles and have given over 100 talks on the topic. We strongly believe in the importance of providing high-quality, scientifically-sound information to citizens about the pandemic and its economic, social and psychological impact, the collective social behavior, the success/failure of contact tracing, the progress on vaccinations, etc.

Our solution has already provided a tremendous positive impact to the Valencian Government and the 5 million people living in the region. It has been one of the regions that has best handled the pandemic Spain, with currently the lowest cumulative incidence per 100K inhabitants (35 vs 341 in Madrid, 212 in Catalunya). The President of the region has regularly relied on our work and referred to it when justifying and communicating his decisions throughout the pandemic.

Our motivation to apply to the Trinity Challenge is to be able to expand our solution and its impact beyond Spain. Our data-driven, machine learning-powered platform could potentially impact the lives of hundreds of millions of people worldwide, enabling policy makers to better understand the current situation, make accurate predictions and simulate hypothetical future scenarios to make more informed decisions. 

The impact could be profound as it would fundamentally transform the way public policies are defined and evaluated and foster the transition towards more data-driven and transparent policy-making.  

In addition, the proposed platform would be of value to businesses, NGOs, CSOs, and citizens, as a source of reliable data and scientifically sound insights. 

Our scaling strategy relies on two connected strands. 

- One, improving the platform's functionalities by connecting the 5 modules; 

- Two, establising strategic partnerships with key global organizations such as ITU (International Telecommunication Union of the United Nations), IDB (Inter-American Development Bank), Data-Pop Alliance, the World Bank; companies with relevant data such as Telefonica, Vodafone, Orange; and Trinity Challenge members (e.g. the Bill and Melinda Gates Foundation, the Clinton Health Access initiative, Google, Facebook, Microsoft, GSK, the London School of Economics and Cuebiq). We will also partner with government institutions in Mexico, Colombia, etc. Critically, we have existing relationships with most of these institutions. 

In Year 1 of our project, in addition to R&D work, we will develop strategic partnerships with key stakeholders. 

Year 2 will focus on expanding the project to additional geographies. 

Year 3 will be devoted to maximizing our impact by expanding to 5 to ideally 10 geographies in total.

As first prize winners of the Pandemic Response XPRIZE, we will continue to participate in the Global Alliance for Pandemic Response fostered by the XPRIZE Foundation. The aim of this Alliance, which complements the Trinity Challenge, is to provide the necessary data to support evidence-driven decision-making. 

The KPIs to measure success will vary according to the phase of our project. 

In Year 1, the focus will be on R&D and partnerships. Regarding R&D, the functionality of each of the 5 modules has been extensively tested and validated with real data. The prescriptor and the deep learning-based epidemiological model have been validated in the XPRIZE challenge (KPIs include Mean Absolute Error, MeanRank, Pareto domination test and computational cost). The frontend development will be done in collaboration with public policy makers. KPIs include user satisfaction/usability surveys and latency. Our target is having a working platform and commitments from 2-3 governments by year end.

In Year 2, the focus will be on scaling our platform. Our target is to have at least 3 ongoing collaborations in countries outside of Spain. We will define KPIs related to intensity of use, number of decisions supported by it, and impact in terms of number of people positively affected. 

In Year 3, the focus will be on expanding and maximizing impact. We will automate the collection of data to measure impact and will collect regular feedback from our customers. Target: effective collaborations with at least 5, ideally 10, regional/national governments.

Two main common barriers may limit our impact.

First, we will face cultural and capacity barriers, as most public administrations lack digital skills and a data-driven mindset. We plan to overcome this barrier through education 'data literacy' programs, technology support, and by actively including members of the public administrations into the workstreams of our team. Co-developing the technology as in Valencia will also be beneficial regarding potential cultural barriers. Investments in state-of-the-art data collection, storage and processing infrastructure will most likely also be needed. 

Second, there are financial barriers. We have obtained two competitive funds in Spain for ~150K euro each until March 2021 and April 2022, respectively. In addition, we won the 500K XPRIZE Pandemic Response Challenge with a 250K USD award. However, given the scope and ambition of the proposed solution, we certainly need additional financial support to be able to develop and deploy the envisioned functionalities. As winners of the XPRIZE, we are in conversations with the ITU via the Global Alliance for Pandemic Response to explore the opportunity to bring a global solution based on our work and that of other members of the Alliance.

Five research institutions in the Valencian region (Universities: Jaume I Castelló, Polytechnic València, Alicante, Miguel Hernandez; the ELLIS Unit Alicante Foundation). We closely work with the Presidency of the Valencian Government and Data-Pop Alliance. Other collaborators include ESRI, Microsoft, the Vodafone Institute, XPRIZE, Cognizant

We started as a team of volunteer scientists united by a common purpose: to help the Valencian Government fight the COVID-19 pandemic through the analysis of data and our expert knowledge. Since March 2020, we have made tremendous progress: from just having access to basic data in PDF files to having emerged as one of the world’s leading teams in the use of Data Science to support policy making during the COVID-19 pandemic. We have developed one of the world’s most accurate, AI-based predictive models and prescriptive algorithms to aid in the COVID-19 pandemics response; one of the largest online COVID-19 citizen surveys with over half a million answers and have raised over half a million pounds in competitive funding to support our work. More importantly, we have successfully helped the Valencian Government in their decision making related to the pandemic, which has led to our government’s trust in our models, way of working and expert advice.

We are strong believers in the power of leveraging data to support public policy making. The Trinity Challenge offers us an exceptional opportunity to scale and expand our work to a global level by leveraging fruitful collaborations with several Trinity Challenge member organisations.

We would love partner with several Trinity Challenge member organisations:

1. Facebook, Google, Joep Lange Institute, Tencent, Cuebiq have access to valuable human behavioral data that we would love to include in our platform. Moreover, we have been collecting citizen survey data using the Facebook ads platform. We would love to partner with Facebook to scale and expand the reach of our citizen survey. 

2. Microsoft and Google for computing resources and possibly development support

3. Dr. Evidence, Johns Hopkins, Inst. Health Metrics and Eval., Joep Lange Institute to support the design and provide valuable expert feedback on our epidemiological models and our intervention prescriptor. 

4. London School of Economics to collaborate in developing a scientifically sound methodology to assess the social and economic costs of the non-pharmaceutical interventions. These costs are a key input to our intervention prescriptor.

5. Clinton Health Access Initiative, Patrick McGovern Foundation, Bill and Melinda Gates Foundation to help us identify potentially interested regions/countries to deploy our platform, particularly in Latin America