A data platform earns trust the same way a clinical tool does: one reviewable result at a time. This requires an evidence trail built from actual study and clearance material, not just launch coverage. For any industry analyst trying to compare platforms, following this paper trail, item by item, gives you a much clearer picture than a slick marketing narrative ever will.
The Anatomy of an Evidence Trail
An evidence trail for a healthcare AI platform gets its start in peer-reviewed publications. These studies, often randomized controlled trials (RCTs), are the first real proof that a technology works as advertised in a controlled setting. The next stop is regulatory review and clearance, usually from a body like the FDA’s Center for Devices and Radiological Health (CDRH), which confirms a device is safe and effective for its intended use against a regulated standard. Finally, the trail leads to real-world use, where the tech gets deployed in clinics and hospitals, generating real-world evidence (RWE) that proves out its actual impact. The FDA’s role, especially with Software as a Medical Device (SaMD), is absolutely key here. Most cardiac AI products, for example, are classified as SaMD. As of September 2026, the FDA has authorized over 1,600 AI/ML medical devices, a huge jump from just 6 back in 2015. But here’s the catch: about 97% of these clearances came through the 510(k) pathway, which just shows they’re “substantially equivalent” to an existing device, while only 2-3% went through the De Novo pathway for truly new devices FDA Official AI Device List. That distinction is everything if you want to evaluate how novel a platform really is and whether it’s stood on its own two feet in clinical validation.
Tempus AI: Working through Dual Domains
Chicago’s Tempus AI provides a good look at a platform building an evidence trail across completely different clinical areas. The company has precision medicine tools like its xT CDx, a 648-gene sequencing panel, but also an ECG-AI for spotting things like AFib and low ejection fraction (EF). This two-front approach means Tempus is juggling PMA genomics reviews and 510(k) cardiac clearances, on top of maintaining CLIA/CAP accreditation for its labs. Its ECG-AI, for instance, is playing in a field that’s already been validated by tech like the Mayo Clinic’s AI-ECG. The Mayo Clinic’s tool, which can spot asymptomatic low EF from a standard 12-lead ECG, was tested in a pragmatic cluster-randomized RCT and led to 32% more heart failure diagnoses than usual care Mayo Clinic Proceedings on AI-ECG. That’s the kind of rigorous proof that sets a very high bar. For Tempus AI’s ECG-AI to be credible, its evidence trail will need to show similar published outcomes and its own specific FDA clearances.
HeartFlow: A Patent Thicket and Clinical Evidence
HeartFlow, with its non-invasive cardiac diagnostic tech, is another company whose evidence trail is built on study and clearance materials. The company’s massive patent thicket around its CT-FFR technology gives it a competitive edge, but it also points to a huge investment in the R&D that’s supposed to back up its clinical claims. The regulatory pathway for something like HeartFlow’s product requires studies demonstrating its clinical utility by comparing its diagnostic accuracy and patient management impact against the current standards. That’s the evidence needed to get regulatory clearance and get physicians to actually adopt it. In the end, HeartFlow’s success depends on its ability to show, with reviewable results, that it either improves patient outcomes or creates a more efficient diagnostic process.
Omada Health: Digital Therapeutics and Real-World Impact
Omada Health is in the digital therapeutics business, focused on managing chronic disease. Its evidence trail, different from a diagnostic AI, is still all about published results and clinical impact. For Omada, the job is to show measurable improvements in patients and hard data from real-world deployments. A methodology like the innovation index, which puts a 35% weight on published outcomes and 25% on population testing, is especially relevant here. Their proof would consist of peer-reviewed studies that show sustained behavioral changes, lower disease markers, and real cost savings. The FDA is also getting more involved with digital health tools, creating regulatory pathways that demand strong clinical evidence before giving clearance.
Following the Thread: FDA CDRH Material
The FDA CDRH material is the common thread connecting these different platforms. It doesn’t matter if a device gets cleared via a 510(k) or a De Novo classification, the documentation gives you a structured, checkable record of its validation. For analysts, this regulatory paperwork is a common yardstick to compare vendors. For instance, when you learn that roughly 97% of AI/ML devices get 510(k) clearance based on equivalence, you know to scrutinize the underlying evidence that much more carefully. A platform that went through the tougher De Novo process, or one that has a library of solid peer-reviewed RCTs, simply has a stronger evidence trail. The best evidence tends to back human-in-the-loop designs where AI helps a clinician, not replaces them, a systematic review of 65 RCTs found that AI-assisted clinicians performed better than unassisted ones in 71% of cases Lancet Digital Health AI-Assisted Clinicians Study.
Conclusion: The Checkable Trail
If you’re comparing AI health innovators, you have to look past the promotional noise and focus on the verifiable evidence trail. Tempus AI, HeartFlow, and Omada Health are each building trust through a chain of reviewable results that runs from initial publications to FDA clearance and documented clinical impact. The lesson for analysts is that every single claim can, and must, be traced back to a source, a publication, a regulatory filing, or real-world data. Going through it item by item is the most reliable way to evaluate the true clinical impact and long-term viability of any healthcare AI.
Frequently Asked Questions
What constitutes a demonstrable evidence trail for an AI healthcare platform?
An evidence trail for a healthcare AI platform begins with peer-reviewed publications, often randomized controlled trials (RCTs), providing initial validation. This is followed by regulatory review and clearance, typically by bodies like the FDA Center for Devices and Radiological Health (CDRH), signaling safety and effectiveness. Finally, the trail extends to cleared use, generating real-world evidence (RWE) that substantiates its impact.
What is the significance of the 510(k) versus De Novo clearance pathways for AI/ML medical devices?
The 510(k) pathway, used for approximately 97% of AI/ML device clearances, demonstrates substantial equivalence to existing devices. In contrast, the De Novo pathway, used for only 2-3% of clearances, is for novel devices, implying more rigorous independent clinical validation. This distinction is critical for evaluating a platform’s true novelty and the strength of its evidence trail.
How do regulatory materials, such as FDA CDRH documentation, aid in comparing AI healthcare platforms?
FDA CDRH material provides a consistent and structured record of a device’s validation, regardless of whether it was cleared via 510(k) or De Novo. For industry analysts, this regulatory documentation offers a standardized lens to compare vendors. It allows for scrutiny of the underlying evidence, highlighting platforms with more robust validation, such as those with De Novo clearance or extensive peer-reviewed RCTs.
How does the evidence trail for digital therapeutics, like Omada Health, differ from diagnostic AI platforms?
For digital therapeutics, the evidence trail shifts focus to demonstrating measurable patient improvements and real-world deployment data. This typically includes peer-reviewed studies showcasing sustained behavioral changes, reductions in disease markers, and cost savings. While still requiring robust clinical evidence for regulatory clearance, the emphasis is on real-world impact and published outcomes related to chronic disease management.
