The Data Strategy Behind Northwell’s Aegis Partnership
Northwell Health, New York State’s largest healthcare provider, recently announced a joint venture with Aegis Ventures to launch and scale artificial intelligence-driven companies to address healthcare issues. During a recent webinar presentation, Marc Paradis, vice president of data strategy for Northwell Holdings and Ventures, the for-profit arm of the nonprofit Northwell Health System, discussed this partnership in detail and spoke about the importance of data quality to success.
New York-based Aegis Ventures is a “startup studio” that partners with entrepreneurs and industry leaders to originate, launch, and scale transformative companies. Aegis and Northwell said they would assemble medical, technology, and business resources to create a company creation platform for healthcare innovation. Aegis Ventures said it plans to invest at least $100 million of seed-stage funds through the platform to catalyze a significant multiple of that amount from the venture capital and investment community.
In a conversation with Dale Sanders, chief strategy officer for Intelligent Medical Objects, Paradis said that at Northwell he is responsible for assessing all of their investments from a data science, machine learning and AI standpoint, as well as for setting data strategy for the health system as a whole.
He first drew attention to a few things that make Northwell’s data assets unique. “Our service area of over 11.5 million lives is arguably the most ethnically diverse, socio-demographically diverse, culturally diverse, economically diverse and genetically diverse population in the U.S. and most likely the world,” he said. “Our scale as a system is really enormous. We treat more than 2 million distinct patients annually. One of my favorite facts is that 1 percent of all births in the U.S. occur in a Northwell hospital or facility.”
Why is that important to AI efforts? Because data is the raw material of AI, Paradis stressed. “There is, in fact, no artificial intelligence without high-quality data. For the past four or five years, the importance and necessity of fair and unbiased AI has become clear, and that really requires broadly representative data for that AI to be trained on. In order for the AI to really work, it's got to be implemented and validated in real-world settings. It can't just be toy data sets or academic data sets; it has to be real-world data collected from real clinical workflows.”
In addition, in order for health systems like Northwell to commercialize AI, they’ve got to deliver quantifiable value, Paradis said. It is easy to make cool idea AI demos, but it is more difficult to make products out of AI models, he said, in part because these models need to meet a whole variety of different criteria. They need to be reliable, generalizable, unbiased, equitable, explainable, impactful, scalable, and adaptable, he said. “The good news is that we do know how to assess, remediate and actually deliver all of these in order to build well-engineered products,” he said. “It takes a lot of highly skilled people putting in a lot of intense efforts supported by state-of-the-art technology and infrastructure.”
So where does the money come from to support that work? Paradis described the typical route: Companies are usually started with the help of friends, family and fellow founders. “We get some initial promising results. We then go out to the angels and the accelerators. We raise some seed money, then we start hiring more people, we start building more demos,” he said. “We raise more money, usually by a series of VC rounds, then we build and scale that AI product. And eventually, we raise enough money that we can go out and do an IPO backed by an investment bank. And then we've reached the scale and the impact where we can change the world for the better with our well-engineered productized AI. That's the paradigm. But the ugly truth of startups is that most don't survive. In fact, only about one in six survive to a Series C, and of those only about one in six get acquired, so the odds are stacked against you.”
Paradis said these AI startups always need to be thinking about whether they are solving a problem their customers care about. “You can have the best product in the world, but if you can't get traction, you're going to be relegated to margins of that market and you're likely to go out of business. The traction gap can't be escaped, and I would argue much, much worse by the siloed whipsaw funding models that we currently use,” he said, “so we really have to find a better way to solve this problem.”
He said Northwell is seeking to solve this problem by what it calls horizontal slicing of its investment strategy, in which it aligns with three uniquely differentiable assets: the scale of its data, the depth of its clinical expertise and the breadth of its healthcare implementation platform.
“We identify internal gaps in care, operational inefficiencies, or administrative needs that have real impact and cost associated with them, making sure that those gaps or inefficiencies generalize to other healthcare institutions and are not just Northwell problems,” Paradis said. “We find partners with AI-enabled technologies, products or services that have the potential to solve these gaps or inefficiencies or needs. If no suitable partner exists, then we look to our own internal innovation.” Northwell then invests its data, expertise, and healthcare implementation platform in return for equity or a revenue share.
Northwell uses what Paradis calls a repeatable, scalable, transparent, evidence-based pipeline that delivers these well-engineered AI algorithms with broad market appeal and that solve problems that providers and patients actually care about.
This helps demonstrate in a methodologically rigorous way that the AI algorithms actually work as part of clinical workflows, and that they deliver hard outcomes and measurable value in the real world, he said.
The partnership with Aegis, Paradis said, is bringing their joint resources together — the clinical expertise of Northwell and the commercialization capabilities of Aegis — to develop AI companies that address issues in healthcare quality, health equity, and many of the other crises that exist in healthcare.
“The intent is to bring together clinicians, researchers, patients and payers to identify the most pressing problems that that need to be addressed today,” he added. “As part of this, Aegis intends to invest at least $100 million of seed-stage funding through the partnership to catalyze a significant multiple of that amount from the venture capital and broader investment community, ultimately accelerating the speed with which new life-saving AI solutions can reach the patient bedside.”
