Investing in Layer Health
Authors: Parth Desai and Jonathon George
Clinical chart data is the most valuable data in healthcare. The insights buried within this data drive the vast majority of clinical and financial activities in healthcare. Unfortunately, 80% of this data is unstructured, limiting its utility.
For decades, we’ve invested hundreds of billions of dollars in a disparate meshwork of technologies to improve how this data is structured and used to inform contingent activities, while our clinicians spend countless hours auditing these efforts. The vision has always been to unify raw care data with its clinical and financial outcomes, so that we can map treatment decisions to their real-world impact, in real time. This vision had been an elusive one until we met the Layer Health team last year.
Today, we’re excited to announce our investment in Layer Health, a company that is leveraging breakthroughs in artificial intelligence (AI) to unify clinical data and transform chart review activities, starting with clinical registries. We’re eager to be co-investing alongside our friends at Define Ventures, GV (Google Ventures), and partner health systems (MultiCare, Froedtert & Medical College of Wisconsin).
Behind this technology is a group of world-class AI researchers and clinicians from MIT, Harvard, and leading technology organizations such as Microsoft, Google and Flatiron. Co-Founder and CEO David Sontag, Ph.D., is a recognized expert in healthcare machine learning who has built a team that combines deep engineering expertise with real-world clinical and product experience: Steven Horng, Monica Agrawal, Alastair Blake, Luke Murray and Divya Gopinath.
Market Opportunity
Clinical registries play an important role in healthcare. They are used to map care data from charts to structured outcomes, informing clinical protocols, population health and quality improvement activities as well as therapeutic and device product efficacy. Registry insights, therefore, have a significant clinical and financial impact on hospital as well as health plan and life sciences operations. However, the process of converting care data from charts into the formatted outcomes specified by registries is complex, because of how disconnected and unstructured the source data is. As a result, hospitals invest heavily in technology and services to streamline this work. Historical technology approaches have been able to automate small parts of this process, but ultimately, clinical staff use their reasoning to complete most of the work manually. This approach tends to be error-prone, time-consuming and limited to just a few high-cost specialty areas. The result? Hospitals lose control and awareness of important clinical and quality insights that could otherwise meaningfully improve care and billing functions.
The financial impact is staggering. Health systems can spend upwards of $6 million in personnel costs for care quality data reporting per hospital, a multi-billion dollar market opportunity. Properly structured chart data linked to outcomes can also positively impact clinical documentation improvement (CDI) efforts, another multi-billion dollar market. CDI optimization also means fewer medical coding errors, stronger revenue integrity programs and health plan quality activities as well as risk adjustment that are trued-up to real outcomes. Meanwhile, the life sciences sector’s growing demand for real world data (RWD) produced from charts and registries, represents an additional $2 billion market. By transforming chart review and increasing the utility of clinical data, Layer Health’s potential impact is far-reaching.
The Cutting-Edge Product
The breakthroughs are made possible by Layer Health’s cutting edge technology stack. Today, the product seamlessly combines the powerful reasoning and inference abilities of commercial large language models (LLM) with the speed, accuracy and cost-efficiency of proprietary small language models.
The product works by connecting to a health system’s back-end database and accessing raw clinical data in Layer Health’s cloud instance. The company’s proprietary small and large language models, fine-tuned on customer data, then analyze the data and provide a concise summary of evidence to populate a registry. These models are highly-performant and generate results much faster than commercial LLMs. However, in some instances complex reasoning may be required to create a final output, in which case Layer Health will sequence its own models with a commercial LLM, to produce the best end result. In either case, Layer Health’s final output achieves human-level accuracy with 65% additional time savings, at a fraction of current costs. This is critical to establish trust and prove the efficiency of their underlying technology.
Importantly, the team has also invested heavily in state-of-the-art machine learning operations (MLOps) as well as quality, safety and audit rails, which ensure their product remains performant, faster and significantly cheaper than the status quo. Layer Health’s commitment to transparency and outcomes is equally distinctive. Every insight generated by the platform can be found accompanied by links to source evidence, giving clinicians a clear audit trail that eliminates the “black box” perception of many AI products.
Layer Health’s ability to both inform and train its models on these unique clinical data linkages creates a powerful clinical reasoning foundation. Longer-term, the company can extrapolate this foundation to optimize the many derivative activities informed by clinical chart data (i.e., quality reporting, CDI). Notably, Layer Health’s technology facilitates their ability to build and ship these products more efficiently and at equivalent to better performance than the status quo.
The Team Behind It
In the dynamic and evolving world of healthcare AI, product advantages are a direct function of the technical and clinical prowess of teams. And in this department, Layer Health is unmatched. The team is purpose built to solve healthcare’s clinical data challenges.
CEO David Sontag, Ph.D., is a leading AI researcher and Professor of Electrical Engineering and Computer Science at MIT, with affiliations in the Institute for Medical Engineering and Science and the Computer Science and Artificial Intelligence Laboratory. Prior to MIT, he served as an Assistant Professor at New York University’s Courant Institute of Mathematical Sciences and was a postdoctoral researcher at Microsoft Research. His work in healthcare machine learning is widely cited and has earned him various industry accolades and faculty awards from organizations including Google, Meta, and Adobe. He was previously also the Chief Health Strategist of ASAPP.
David’s co-founder and VP of Clinical Informatics, Dr. Steven Horng, has a distinguished background in clinical informatics and emergency medicine from Harvard and Beth Israel Deaconess Medical Center. Steven ensures the product directly addresses real-world healthcare needs. David and Steven’s other co-founders include Monica Agrawal, a renowned AI researcher at Duke who previously spent time at Google and Flatiron, and Luke Murray and Divya Gopinath, whose engineering expertise provides the architectural strength for Layer Health to execute its bold vision.
The Road Ahead
As more health systems and life sciences partners adopt Layer Health’s platform, they’ll see immediate cost and efficiency gains. But the longer-term story is even more exciting: fully structured, high-fidelity data that can drive better clinical outcomes, inform better clinical protocols, reduce administrative friction, and increase revenue through more accurate billing or RWE partnerships.
We’re thrilled to be partnered with the Layer Health team on this journey to modernize clinical data abstraction and redefine how healthcare uses its data.
To learn more about Layer Health: Reach out to Parth Desai parth@flarecapital.com and Jon George jon@flarecapital.com — or any other member of the Flare Capital team — we’d love to hear from you.
Careers: You can view the company’s open roles here.
Stay in touch with Flare Capital:
· Subscribe here for more insights
· Follow us on LinkedIn, X (Twitter), and our website.
