XPRIZE Digital Twin

Current models of healthcare are not holistic and biopsychosocial, and fail to sufficiently account for physical, emotional, and social well-being that also have socio-environmental interdependencies.

There are many apps, wearables, and other devices that track health information, but they tend to track just one or two things, and they don’t share their data.

What are your thought on an XPRIZE competition that integrates data from biomarkers, body scans, diets, medical research, sensors, wearables, and user input to create a “digital twin” of the body – contextualized to culture, place, and individual needs and plans – that can:

  • Provide users with actionable insights to make better health decisions;
  • Democratize access to data and expertise.
  • Enable timely interventions; and
  • Improve health outcomes.

Would this be an audacious enough prize in light of current innovations?

If so, how would we judge the winning solutions?

@kkatara, @mayaelhalal, @kenjisuzuki, @mario_perez, how would you feel about an XPRIZE that challenges teams to develop a “digital twin” of the body? Do you think this is audacious enough in light of existing and emerging innovations? Would it be achievable within the usual timeframe of an XPRIZE competition (~5 years)?

CC @siddhartha

Maybe it’s too early for that. Gathering all the data from the environment (called “exposome”) that can influence health seems rather complex. Maybe the focus can be on specific organs or diseases?

Examples from companies:

I like what myfeel is trying to achieve. Just knowing when you’re stressed and making it easy to change the situation could help in more than one disease.


@mario_perez Thank you , these are some fantastic examples of companies working in these areas. Seems like there could be a lot of crossover impact in accelerating precision medicines and drug/therapy development, beyond equipping users with better data about their health.

I agree with @mario_perez. We’re doing a digital twin for part of the endocrine subsystem, and even that is really at the edge of the envelope for various aspects of medicine/AI etc. So, specific organs and diseases make more practical sense.

@mario_perez, @MachineGenes, what if we let teams decide which data to collect and integrate?

We could stipulate that the digital twin needs to integrate data from open-source medical research + user-specific data, but competing teams could decide whether to use data from biomarkers, body scans, diets, sensors, wearables, user input, etc.

The goal is to provide users with actionable insights to improve their health and well-being – as well as the health and well-being of others, with an option to donate specific data to the model.


That could work. Actionable insights from preventing disease should be the goal.

Keep in mind this would probably be a competition of ~5 years, launched at the earliest in 2022. (If it makes it all the way through the Global Visioneering vetting process and gets funded.)

We want to set achievable targets, but we also want to accelerate the field – and the worst thing would be if our prize were outpaced by the market.

@MachineGenes, @mario_perez, @sarahb, @fabienaccominotti, @key2xanadu, @yanchen, @ymedan, I would like to ask your advise on the current state of this prize idea.

Teams would be challenged to:

Develop an open-source digital twin that collects and integrates a wealth of data to provide users with actionable insights to improve their health and well-being, and prevent the need for sick-care.

Potential parameters and judging criteria include (with our questions):

  • Collect and integrate open-source medical information with user data (from biomarkers, body scans, diets, sensors, wearables, user input — teams would decide which).
    • Model must be able to integrate more data over time.
    • If teams want to bring their own non-user data, they need to share it with all competing teams. (Role for Data Collaborative.)
    • Bonus points for bringing data from underserved communities? To combat biases in health data.
  • Contextualize data to age, culture, language, place, and individual needs and plans.

Should teams be allowed to choose to which factors they want to contextualize the data? Or do we give them the criteria?

  • Predict a user’s health journey.
    • For example: “If you keep eating like this, there’s a 74% chance you’ll get diabetes in the next 15 years.”
    • Leverage continual learning architectures (AI and machine learning).

I think we should leave this open-ended, otherwise we’d have to give teams a litany of conditions we want them to predict for.

  • Provide personalized and precise insights, directly to the user in an understandable way.

I don’t think we can set objective criteria for what “understandable” means, so let the judges decide?

  • Measure and demonstrate the impact of behavior changes.
  • Simulate medical interventions.

Would it be too ambitious to include this as well?

  • Users must retain ownership of their health data and have the option to donate specific data to the model.
  • Data security.

How to phrase a judging criteria for data security?

Wellbeing is not an exclusive function of “medical information”. We now know that health is acutely modulated by social determinants. These may even provide earlier warning signs regarding health hazards and risks, before it can be actually measured, leading to more effective mitigation actions.

Another set of factors are environmental, related to climate change, air and water pollution and chemicals like PFAS. So the “wealth of data” should not be limited to medical only.

Easier if you give the criteria.


That’s the end goal, motivate the user to those behavioral changes that improve health?

I guess data (if shared) should be anonymized and it should follow current legislation (e.g. GDPR). A blockchain based solution could be used? (e.g. https://medicalchain.com)

1 Like

@ymedan, @mario_perez, thank you both for your feedback!

On the point of non-medical data, I’m a little wary of giving teams X number of factors they should consider. What if we’ve overlooked something? Or what if we think something is important, but they don’t?

Maybe the middle ground would be to stipulate that they must integrate at least X number of non-medical factors and then give them a list they can choose from or add to?

Absolutely! I think my formatting there might cause confusion. “Measure and demonstrate the impact of behavior changes” is definitely in. The question is whether “Simulate medical interventions” is too ambitious or not.

1 Like

What does this mean? If teams have secured ethics approval to use patient medical data in a specific way, then it’s probably unreasonable to expect them to share the data. There are multiple layers of de-identification; again, depending on the application, full anonymization might not be feasible at the clinical end.

Also, I would not be prescriptive about integrating non-medical factors, as this would be condition-specific. Perhaps the appropriate marking on the enhancement of non-medical interventions should be left to the judges to decide.

1 Like

Sounds perfect this way!

I predict those simulations will be hard, so it can be optional?

I strongly believe that this is a good idea but may be extended to the digital twin of contaminated sites to monitor the earth (soil and water) and provide a cleaner environment where earth habitants live, eat (irrigated crops) and drink.

Here’s where we’ve landed on the exact “winning team will” statement for this prize:

  • Design a tool that can securely collect and integrate various types of data (biological, lifestyle, medical, psychological, well-being) from various sources.
  • Build a solution that:
    • Enables users to take their data with them.
    • Predicts a user’s health journey.
    • Provides concrete recommendations to improve health outcomes.

We imagine this would be a two-staged competition, possibly with different teams competing in the first and second phases. The first is data- and methodology-focused, the second is about technology and user experience.

1 Like

Here’s the latest version of this prize sketch:

The grand challenge we want to solve is that one in four deaths in OECD countries can be avoided through better prevention and timely healthcare interventions. 80% of chronic diseases are driven by lifestyle factors, such as diet and exercise. 70% of potential health gains in the next two decades come from prevention. Yet our healthcare systems remain disease-focused. People don’t have the information they need to make wiser health decisions. People don’t go to the doctor until they are ill, by which time interventions are costlier. Most drugs and treatments come in one version. Individual needs, cultural differences, and local context are insufficiently taken into account.

What’s holding us back from switching from a one-size-fits-all sick-care system to personalized and preventative care? Reliable, complex data.

Democratization and integration of data would empower patients, and enable precise and timely interventions that improve health outcomes and reduce the need for sick-care.

This prize would have two phases:

  • Building an engine that can securely collect and integrate various types of data (biological, lifestyle, medical, psychological) from various sources, and attribute ownership to users.
  • Designing a solution that:
    • Maps out a personalized health journey to help the user reach their ideal health state.
    • Enables users to take their data with them.

The data engine would be judged on:

  • Integration of publicly available, anonymized medical information with non-traditional health and lifestyle data from users.
    • Existing datasets.
    • Datasets brought by competing teams. (Must be shared with all.)
    • Data crowdsourced from volunteers.

Bonus points for bringing data from communities underserved by healthcare systems.

  • Contextualization of data to age, culture, language, place, and individual needs and plans.
  • Future-proof: Able to continually integrate data over time.

The user tool would be judged on:

  • Portability: Enables users to carry their data with them.
  • Functionality:
    • Maps out the user’s health journey.
    • Provides personalized and precise insights to the user in an understandable way.
    • Allows users to donate (anonymized) data to the model.
  • User engagement:
    • How many people sign up?
    • Do users follow the recommendations?
  • Privacy: HIPAA and GDPR compliance.
  • Data security.
1 Like

I am really interested in this from a meta-data and AI perspective. Imagine as you walk into a hospital or medical clinic, all the information already being collected from gait analysis to skin temperature, and how that could inform early diagnosis. This data is sometimes already being collected. Privacy on hospital and medical sites may be a thing of the past… even shopping centres can make decisions about you based on your physical state.

1 Like

I am really interested in the data being collected here, both passively and actively.