Sepsis AI: FDA Clearance vs. Exemption, An Investor’s Guide

The AI scene in healthcare is moving fast, but that speed brings a ton of promise and some gnarly regulatory headaches. If you’re a buyer, part of a health system’s AI governance team, or an investor trying to make sense of it all, you have to understand the different ways these AI-powered tools get to market. You can see this clearly in a life-or-death area like sepsis prediction, where the line between an FDA-cleared medical device and an exempt clinical decision support (CDS) tool changes everything, from the rigor of its validation to how it gets adopted and, in the end, if it keeps patients safe.

Two Paths to Sepsis Prediction: De Novo vs. CDS Exemption

Sepsis is a killer that demands fast, accurate identification. AI tools that predict sepsis risk are a fantastic idea, but how they get through the regulatory maze shows a deep split in how AI gets classified. You can see this split perfectly by comparing Prenosis’s Sepsis ImmunoScore with the Epic Sepsis Model that’s in hospitals everywhere. On one hand, Prenosis went through the full wringer and got FDA De Novo authorization for its Sepsis ImmunoScore on April 3, 2024. This was a big deal, the first AI/ML sepsis diagnostic to get this specific type of green light FDA De Novo authorization for Sepsis ImmunoScore. The De Novo path is built for new, low-to-moderate-risk devices that don’t have a clear predecessor. For an investor, this is a huge de-risking event. It’s a signal that the FDA has kicked the tires and agrees it’s safe and effective for what it claims to do. That’s a massive consideration for any SaMD (Software as a Medical Device) that actually makes a diagnostic call. The Epic Sepsis Model, on the other hand, plays by different rules. It isn’t FDA-cleared. Instead, it lives in the market under an exemption created by the 21st Century Cures Act’s Section 3060 for certain Clinical Decision Support (CDS) software. This isn’t a secret. It’s an explicit distinction discussed in academic papers like the ScienceDirect article, “Bridging the gap: aligning clinical decision support regulation with clinical practice in the era of artificial intelligence” ScienceDirect article on CDS regulation. The Epic Sepsis Model gives doctors useful information, but it’s positioned as a tool that informs rather than diagnoses, which keeps it out of the FDA’s direct line of fire for medical devices.

The 21st Century Cures Act and the CDS Exemption

The 21st Century Cures Act from 2016 was meant to get medical products to patients faster. A key part of that was Section 3060, which carved out exemptions for some software, pulling it out of the “medical device” definition. This includes CDS software functions, as long as they meet a few conditions. Generally, the software isn’t supposed to analyze medical images or signals from lab tests on its own, and it must let a doctor independently review the evidence behind its recommendations. For a company like Epic Systems, this exemption is a straight shot to market for its Sepsis Model, which is baked right into its electronic health record (EHR). It lets them deploy the tool widely without the long, expensive premarket review devices go through. The catch? This puts the burden squarely on hospitals and doctors to vet the tool’s performance and use it correctly. The FDA hasn’t signed off on its clinical efficacy or safety. This is a red flag for any health system’s AI governance team, who now have to create their own internal validation and monitoring protocols from the ground up.

FDA De Novo: A Pathway for Novel AI Diagnostics

So what about the FDA De Novo path? It’s the rigorous route for new devices that are low to moderate risk but have no legally marketed equivalent to compare against. For an AI/ML diagnostic like Prenosis’s Sepsis ImmunoScore, taking this path basically tells the world the tech is genuinely new and its safety needs a custom review. The company has to submit a complete package of data, including clinical studies, proving the device is reasonably safe and effective for its job. Bobby Reddy Jr., the CEO of Prenosis, has framed this decision to pursue FDA De Novo as a deliberate choice to seek a higher bar for validation. This kind of talk really connects with investors who want regulatory clarity and a less risky path to commercialization. A De Novo authorization also creates a whole new regulatory classification. Why does that matter? It means the next company with a similar device can use the much simpler 510(k) pathway. Going for clearance first is a strategy that shows a real commitment to clinical evidence and playing by the rules, which builds a ton of trust with health systems thinking about adoption.

Implications for Buyers and Health System Governance

If you’re buying AI for a hospital or sitting on a governance team, the difference between an FDA-cleared AI diagnostic and an exempt CDS tool is practical, not just semantic. It has real, tangible implications. An FDA-cleared device hands you an independent, third-party validation of safety and efficacy which lightens the load on your own team to run massive validation studies from scratch. Of course, health systems will always do their own due diligence, but the FDA’s approval provides a foundation of trust. This is incredibly relevant when you consider “algorithmic drift,” the tendency for AI models to get less accurate over time as the patient data they see in the real world changes. A cleared device often has a Predetermined Change Control Plan (PCCP) or a similar framework for managing model updates with ongoing FDA oversight. For exempt CDS tools, that responsibility for monitoring and validation falls almost entirely on the hospital that bought the software. These tools can be very helpful, but deploying them demands a serious internal governance structure to check their impact on workflows, patient outcomes, and potential biases. As researcher Jenna Wiens has pointed out, all AI models need external validation and transparency, regulated or not, to make sure they’re actually helping and not causing unintended harm Research on external validation of AI in healthcare.

Investment and Regulatory Field

From an investor’s point of view, a company’s regulatory strategy says a lot about its risk profile and market potential. When a company goes for FDA clearance, especially De Novo, it shows they’re committed to a high standard of evidence. This often translates to a stronger market position and more investor confidence. The clarity that comes with a 510(k) or De Novo authorization can also smooth the path to reimbursement, which is obviously make-or-break for commercial success. The fact that we have both FDA-cleared and CDS-exempt AI tools trying to solve the same problem (sepsis) just shows how much the regulatory field is still in flux. The CDS exemption allows for flexibility and gets new ideas out there, but it demands that health systems are sophisticated buyers and that AI developers are transparent. As the AI health market grows up, we’re going to see more pressure for all AI tools that affect patient care to show strong validation and prove they’re being monitored. The different roads taken by Prenosis and Epic with their sepsis tools are a perfect case study of where we are right now. Both want to improve patient outcomes, but their regulatory choices have huge implications for buyers, governance teams, and investors. The trend toward proactive FDA clearance for new AI diagnostics shows a maturing field, but the continued use of CDS exemptions means health systems have to stay vigilant and be ready to do their own validation work.