Vaccinations are coming – and AI is needed more than ever (Part 2)

As COVID-19 infects more people worldwide than ever before in its nearly year-long rise to global pandemic proportions, hope is also emerging. Several vaccine trials have proved safe and effective, and vaccination programs are moving forward globally that could eventually end the pandemic.

This makes our Pandemic Response Challenge with XPRIZE even more timely and relevant than when we launched it in mid-November. The reasons:

• The need to adapt intervention plans as vaccinations roll out. It may take months, perhaps years, to vaccinate the world population and eradicate COVID-19. Throughout that time, the world is vulnerable to the disease, and intervention plans continue to be crucial in containing and mitigating it.
  
  Vaccinations will complicate intervention plans by creating a continuously changing situation. The plans will need to be adapted to the level and distribution of immunity in different regions – and public health officials have little experience coping with such added complexity. This challenge is exactly what the XPRIZE competition is about: The techniques developed in the competition can be instrumental in finding effective intervention plans simultaneously with ongoing vaccination efforts.

• The complex logistics of managing the vaccination programs themselves. Vaccinating billions of people around the world, many for the first time, is a formidable logistical endeavor. It is, however, a decision problem that can be defined by variables that describe the context, the actions to be considered and the outcomes that must be optimized. The techniques developed in the Pandemic Response Challenge can be applied to overcome this obstacle, as well.

Accounting for dynamic variables

Fortunately, the challenge is already set up to take into account vaccination effects to a limited extent. The Oxford University COVID-19 Government Response Tracker dataset used in this competition and updated daily contains an intervention category Health Level 7 (H7), coding for ongoing vaccination efforts (i.e., the extent of population being vaccinated, such as essential workers, clinically vulnerable groups, elderly groups, others). Remarkably, this data is available immediately as soon as any such actions are taken in any country – there is no history to be encoded into the dataset first.

How can the vaccine information be taken into account in the competition entries? Technically, the H7 information is background data, similar to factors such as obesity rates, demographics, weather and other supplemental data. Note that although H7 is listed similarly to intervention actions in the Oxford data, it is actually not an intervention that can be prescribed at will (e.g., full vaccination of everyone may be desirable but not possible). Instead, H7 indicates a level of existing vaccination efforts, similar to other background variables. Predictor and prescriptor models can use it as additional input, and change their output accordingly.

Note, however, that similar to other background information, H7 cannot be updated once the predictor or prescriptor evaluation period has started in the competition – it only forms an informed starting point for the rollouts.

While it is possible in this way to prepare for changing H7 data, the vaccination information is unlikely to affect the course of the competition significantly. The amount of data will be limited for several months: Only a few countries will be able to start their vaccination programs, vaccinate only a small subset of the population, and it will take several weeks for the vaccinations to become effective. The effect will probably first be seen in deaths and hospitalizations, and months later, in the number of cases. By the time our competition concludes in February, vaccinations will be just beginning to have a discernible effect.

Gaining knowledge

However, there may be an opportunity to get an early understanding of such effects in some cases, and the competition platform can be used to demonstrate them, even if it isn't part of the competition. Since the competition allows specialization for particular regions, it may be possible to utilize vaccination information in a few selected regions to do better.

For instance, the UK started its vaccinations in early December, and if its program progresses rapidly enough, data may exist on the first few stages by the time the prescriptor evaluation stage begins. At that point, it may be possible to retrain predictors for such regions with the latest H7 information before the end of Phase 2, and thus characterize and utilize their effect. 

Since H7 data is likely to change within a few months, the evaluation horizon for such entries may be condensed. If the opportunity arises quickly enough, such optional evaluations may be possible, allowing us to learn more from the competition, and thus prepare for possible deployments after the competition.

Managing the vaccination rollout

The Pandemic Response Challenge is intended to encourage the development of technologies that are helpful not just in the intervention plan for COVID-19, but also for the decisions needed for solving grand challenges caused by future black-swan events, as well. This includes future pandemics, natural and man-made disasters, even global warming.

However, the most immediate issue is to solve the second challenge mentioned above: How should the vaccination programs be designed to maximize effectiveness at minimal cost? Different countries have different resources to carry out vaccination programs, spanning manufacturing, distribution, storage and administration of the vaccines. They have dissimilar demographics, mobility and culture, requiring different targeting, delivery mechanisms and public information campaigns. Available vaccines will differ in cost, availability and ease of distribution, and different countries will implement vaccinations at different times.

The prescriptions resulting from the Pandemic Response Challenge may include public information campaigns, vaccination requirements and checks associated with different activities. Provided that the competition is successful, the winning technologies could be adapted for this second challenge. The context, action and outcome variables may differ, but the same idea of learning to predict the outcomes, and using the predictions to learn prescriptions, still applies.

Whereas the world was unprepared for COVID-19, scrambling to contain and mitigate it, we can learn from it. Efforts like the XPRIZE Pandemic Response Challenge should allow us to develop the tools we need to do better in the vaccination phase of the pandemic.

This post was adapted from a blog that originally appeared on the XPRIZE site.

Learn moreabout the Pandemic Response Challenge with XPRIZE, a $550k, four-month challenge that focuses on the development of data-driven AI systems to predict COVID-19 infection rates and to prescribe intervention plans.

Take a look at the finalists, the winners and the technology behind the Pandemic Response Challenge, and read Parts 1, 3, 4, 5, 6, 7 and 8 of our blog series.