Berkeley Launches Agile Metabolic Health and Open Platforms Initiative

Agile Metabolic Health aims to revolutionize the patient experience for millions of people with diabetes, catalyzing a new frontier of open-source, data-driven personalized healthcare.

Experts at UC Berkeley and UC San Francisco have begun working to improve the treatment of diabetes and metabolic health using a first-of-its-kind open-source platform with broad societal benefits.

The Agile Metabolic Health project is being launched by Berkeley’s College of Computing, Data Science, and Society to illustrate how open platforms can transform healthcare. The ability to acquire, analyze and share safely and securely health records and data from wearable devices in one place would enable doctors and others to better understand and treat patients. 

This initial project could help the more than 38 million Americans who are diagnosed with diabetes. But the college’s vision is much bigger. It aims to do much more than improve diabetes care. What the college is building in partnership with developers from Project Jupyter, the 2i2c organization and The Commons Project is a platform to revolutionize the entire health ecosystem.

“Health technology has been driven by proprietary, targeted, siloed approaches. It's a patchwork, and it just isn't working,” said Ida Sim, co-director of the UCSF/UC Berkeley Joint Program in Computational Precision Health (CPH) and a UCSF professor of medicine. “We're committed to building an alternative to transform healthcare.”

Now is the time for a new era in digital health. The Covid-19 pandemic exposed a disconnected health system with massive gaps in data usability and access. Recent advances in AI together with an open-source approach offer untapped potential to help solve issues for patients and providers and catalyze the creation of healthcare startups to advance the ecosystem.

“The healthcare sector is complex, highly regulated and labor intensive, but also in need of change that uses our best research knowledge while lowering costs and empowering clinicians and patients,” said Fernando Pérez, an associate professor of statistics, co-inventor of the open-source platform Jupyter, and incoming faculty director at the Berkeley Institute for Data Science (BIDS). This launch “is a signal of the maturity of our tools and the quality of our team.”

Led by Sim and Pérez, the Agile Metabolic Health project is being built, fostered and supported by the college’s Open Platforms for Powering Science and Society initiative. The open platforms initiative will learn from this initial project, leveraging its outcomes to create new platforms and strategies that similarly transform data capacity around basic science, climate and sustainability, and human welfare and social justice. 

It builds upon a track record by Pérez and others, harnessing open platforms to maximize the scale and impact of work or research. For example, Nature named Jupyter one of the 10 computer codes that transformed science. Open-source tools have also revolutionized and expanded access to higher education by powering JupyterHub, which supports more than 10,000 students monthly at Berkeley alone and is used in hundreds of higher education institutions worldwide.

“Agile Metabolic Health will have an extraordinary impact on the lives of people dealing with diabetes and help clinicians harness the power of data-driven patient care in real-time,” said Jennifer Chayes, the college’s dean. “Open Platforms for Powering Science and Society will multiply that impact to address healthcare, climate change and other complex societal problems.”

“Civilizations are built on infrastructure, and open platforms are a new type of infrastructure to build our future,” Chayes said. “This is only the beginning.”

These innovative endeavors are made possible by a visionary partner who made an anonymous philanthropic investment. 

‘A new kind of medicine’

Diabetes is an impactful lens for medical and computing experts to address using open-source code, science and platforms. The chronic disease, a leading cause of death in the U.S. and an illness that disproportionately affects people of color, results in too much blood sugar in a person’s bloodstream and can harm the heart, kidneys and vision.

Medicine as currently practiced can seem like a blunt instrument for treating patients with diabetes or related problems. For example, doctors often treat patients based on three-month averages of sugar or glucose levels in their blood. But with wearable devices and other technological advances, doctors today could use real-time data on glucose, physical activity and sleep to gain a more precise understanding of a patient’s experience with the illness for better treatment.

That’s a transformative opportunity to improve patient care. A doctor with access to this kind of data could find that new or worse symptoms for a patient could be improved by getting better sleep instead of prescribing a costly drug with potentially negative side effects.

“That's the kind of conversation I can't have right now, and it's a conversation with you about your sleep and your exercise and its impact on your blood sugar,” said Sim, who is a practicing primary care physician. “This is a new kind of medicine we’ve never been able to do before.”

Collaborators at Berkeley, UCSF, and Project Jupyter are building JupyterHealth to help enable that future. The pioneering platform aims to streamline the secure collection of health data from sensors and medical records to develop and share digital biomarker algorithms in partnership with Duke University’s BIG IDEAs Lab, for validation and deployment in clinical care. It could gather data from a patient’s watch or a continuous glucose monitor every five minutes, send it to the platform and provide that data and related analytics in the patient’s chart. This HIPAA-compliant platform will bring together previously siloed data, enable the rapid adoption of third-party healthcare apps, and give physicians and patients real-time insights.

The Agile Metabolic Health project team hopes this new data acquisition and user interface solution will be a tried-and-tested reality in just four years. The project started acquiring blood pressure data from volunteers in September and intends to bring it into a narrow version of the platform by 2024, Sim said. They’ll do the same for glucose data over the course of the following year and expect to pilot and iterate on the platform at UCSF’s general medicine, diabetes and weight management clinics sometime in 2025, she said.

By developing open-source code, “it enables this hopefully rapid iteration between development and field testing and field validation that allows us to really drive that cycle of testing much, much faster than before, so that we can help patients quicker than we are now,” said Sim, a primary care physician at UCSF and a global leader in the technology and policy of large-scale health data sharing.

This platform would allow the medical system to more quickly harness the latest science on metabolic health treatment and function across providers and data sources.

‘We aim to change the world’

While being piloted for diabetes, project partners believe JupyterHealth could present broader opportunities to revolutionize the healthcare system for patients, practitioners and businesses. That’s why the Open Platforms for Powering Science and Society initiative will look for future opportunities to use this model to transform data-driven care for people with multiple chronic diseases, who account for 40% of the U.S. population.

The JupyterHealth platform would allow the medical system to apply the latest science including AI algorithms more quickly and could lower costs by creating universal data solutions for healthcare. It could be used across healthcare systems and for different kinds of patient and clinical scenarios. It would function as a “middleware” for the industry, like the internet for the medical world, Sim said.

The collaboration enabled by the college and including expertise at Berkeley and UCSF is one of the few ecosystems with the broad multidisciplinary excellence needed to enable development of this groundbreaking healthcare platform. This is “a complex challenge that requires the convergence of technical expertise and social understanding, along with careful navigation of policy and ethical considerations,” said Pérez. That diverse expertise is necessary to ensure “our solutions are not only innovative but also responsibly aligned with healthcare regulations and societal needs.” 

The existing team will “build tools and demonstrate their use in real-world healthcare scenarios,” Pérez said. But they are already looking for more thought partners to revolutionize healthcare in the open-source ecosystem, he said. Eventually, there could be opportunities to collaborate on developing JupyterHealth solutions around additional diseases, clinics and companies. Start-ups, for example, will play a key role in developing related commercial efforts that apply these approaches at scale, said Pérez.

That prospect is particularly significant because of the share of the American public it could help and the share of the U.S. economy it could reach. Healthcare accounts for 20% of U.S. gross domestic product and improving treatment for diseases will beget additional economic and societal benefits. Diabetes alone costs the nation $327 billion in healthcare bills and lost productivity every year.

“This is about open-source software and creating a community of many more practitioners who develop these tools and deploy them in healthcare systems with the right constraints for their needs and patients,” said Pérez. “We aim to change the world.”

‘The power of an open platform’

Open Platforms for Powering Science and Society will focus on more than healthcare. In general, open-source software can make science faster and more transparent, accessible and collaborative. That approach has proven essential to addressing central societal problems, from understanding the impacts of climate change to expanding access to education. It has underpinned groundbreaking research, including the first imaging of a black hole, and enabled a more transparent, replicable scientific publishing process. 

Pérez hopes to enable similarly cooperative, cutting-edge research projects through this initiative and forge interdisciplinary programming, community and pathways to broadly support and incentivize open science work. 

Interoperable, secure, open-source platforms like JupyterHealth could be developed and used by stakeholders, including regulators, community members and researchers, to address pressing issues like biodiversity loss and pollution, Pérez said. All parties focused on similar environmental challenges could leverage the same platform to serve their unique needs while retaining full control of their data and infrastructure, he said. This approach maximizes the value of data and information and effectively addresses concerns about data privacy.

“This combination of individual empowerment with global reach is the power of an open platform,” said Pérez.