VBC Evidence Pyramid: Vetting Your AI Health Vendor’s Claims

The promise of artificial intelligence in health is vast, yet its true value in a value-based care (VBC) framework hinges entirely on demonstrable, quantifiable outcomes. For health plan executives and clinicians alike, the critical question isn’t whether an AI solution exists, but rather, what level of evidence underpins its claims of cost reduction and improved patient health. As Eric Topol has often emphasized, the rigorous validation of digital health tools is paramount to their integration into clinical practice and reimbursement models.

Our VBC Evidence Pyramid serves as a crucial framework for evaluating AI health platforms. This hierarchy ascends from the foundational, least rigorous forms of data to the gold standard of scientific validation. At its base, we find pilot data and anecdotal reports. Moving upward, we encounter observational studies, followed by prospective cohort studies. Higher still are randomized controlled trials (RCTs), culminating at the pinnacle with peer-reviewed RCTs published in reputable journals. For any AI health tool to genuinely participate in value-based care arrangements, and to justify the financial commitment from payers, it must demonstrate evidence at the upper echelons of this pyramid. Tools without robust, peer-reviewed outcomes data cannot credibly claim to reduce costs or improve patient outcomes in a manner suitable for VBC contracts.

Framework Requirements for Outcomes Evidence

For AI health platforms to be considered credible partners in value-based care, they must publish specific types of evidence that ascend the VBC Evidence Pyramid. This starts with transparently presented pilot data, which, while foundational, offers only preliminary insights. More robust are observational studies, which can identify correlations and generate hypotheses for further investigation. Prospective cohort studies follow individuals or groups over time, offering stronger evidence of association between an intervention and an outcome. However, the true inflection point for VBC credibility begins with randomized controlled trials (RCTs). RCTs, by minimizing bias through random assignment, provide the strongest evidence of a causal relationship between an intervention (the AI tool) and a health outcome or cost reduction. Crucially, for VBC arrangements, these RCTs must be peer-reviewed and published in high-impact medical journals. This external validation by the scientific community ensures methodological rigor, data integrity, and unbiased interpretation. Without this level of evidence, claims of AI healthcare cost reduction or improved financial performance remain unsubstantiated and unsuitable for inclusion in outcomes-based contracts. Lisa Rosenbaum has consistently highlighted the importance of rigorous clinical trials for novel medical technologies, a standard that AI health solutions must meet.

Comparative Application: Evaluating AI Health Platforms

When we apply the VBC Evidence Pyramid to prominent AI health platforms, a clear picture emerges regarding their readiness for value-based care. The landscape is diverse, with varying levels of commitment to rigorous outcomes research.

• HeartFlow: This company stands out for its robust clinical evidence, particularly its peer-reviewed RCTs published in journals like JAHA and JACC. Their technology, which uses AI to analyze CT scans for coronary artery disease, has demonstrated significant clinical utility and has been the subject of multiple studies validating its impact on diagnostic pathways and patient management. This level of evidence aligns with the highest tiers of the VBC Evidence Pyramid, making a strong case for its inclusion in outcomes-based contracts. HeartFlow clinical evidence publications
• iRhythm Technologies: As a leader in cardiac arrhythmia detection with its Zio XT patch, iRhythm has also invested in clinical validation. While much of their early evidence might have started with observational studies, they have progressed to publishing data in peer-reviewed journals, demonstrating the efficacy of their long-term ECG monitoring. This commitment to publishing outcomes data moves them firmly into the upper-middle tiers of the pyramid.
• Hinge Health: Focusing on musculoskeletal pain, Hinge Health has published numerous studies, including RCTs, demonstrating reductions in pain and surgery rates. Their evidence base, often peer-reviewed, positions them well within the higher levels of the pyramid, supporting claims of AI healthcare cost reduction through reduced surgical interventions and improved patient function.
• Omada Health: Known for its digital diabetes prevention and chronic disease management programs, Omada Health has consistently published outcomes data, including some peer-reviewed RCTs, showing sustained weight loss and reductions in A1c levels. This evidence places them in the upper-middle to higher tiers, making a compelling case for their role in VBC models targeting chronic disease management.
• Noom: While widely popular for weight management, Noom has significantly expanded its evidence base, now boasting over 40 peer-reviewed scientific articles, including recent randomized controlled trials (RCTs) demonstrating sustained weight loss. This progress moves them into the middle to upper-middle tiers of the pyramid, strengthening their claims for VBC models.
• BetterHelp and Calm: These mental health and wellness platforms offer valuable services. BetterHelp has published outcomes reports and peer-reviewed research demonstrating symptom reduction and clinical improvement. However, for both platforms, a consistent track record of peer-reviewed randomized controlled trials (RCTs) demonstrating significant, long-term clinical outcomes or AI health financial performance in a VBC context remains less prevalent. They generally reside in the lower to middle tiers of the evidence pyramid.
• Commure: As a platform company focused on healthcare infrastructure and applications, Commure’s evidence often relates to operational efficiencies and integration capabilities rather than direct clinical outcomes from specific AI interventions. While crucial for the ecosystem, their evidence profile is distinct and typically does not fit neatly into the VBC Evidence Pyramid for direct patient outcomes, though their underlying technologies may support solutions that do.

It is evident that most AI health solutions currently reside in the lower to middle tiers of the VBC Evidence Pyramid. While pilot data and observational studies are valuable for initial exploration and hypothesis generation, they are insufficient for the stringent demands of value-based care contracts where financial performance is directly tied to proven outcomes.

Institutional Standards and Regulatory Context

The push for robust evidence is not merely an academic exercise; it is increasingly mandated by institutional standards and regulatory bodies. Journals like JAHA, JAMA, and JACC serve as gatekeepers, publishing only the most rigorously conducted and peer-reviewed research, setting the bar for clinical credibility. Professional organizations such as the ACC further endorse and disseminate evidence-based guidelines, influencing clinical practice and payer policies. ACC evidence-based guidelines for cardiovascular care

From a regulatory standpoint, the FDA SaMD Framework provides a critical lens for AI health platforms. Software as a Medical Device (SaMD) encompasses AI tools that are intended for medical purposes without being part of a hardware medical device. The framework outlines expectations for clinical validation, performance, and safety, crucial for ensuring that these digital tools are both effective and responsible. Similarly, ISO 14155:2026, which specifies requirements for the design, conduct, recording, and reporting of clinical investigations performed in human subjects to assess the safety and performance of medical devices, offers a globally recognized standard for clinical trials involving medical technologies, including AI.

For value-based care, CMS and its innovation center, CMMI, are increasingly demanding evidence of not just clinical efficacy but also demonstrable cost savings and improved population health outcomes. As value-based care models mature, the expectation is that AI health solutions seeking reimbursement or inclusion in these programs will need to present evidence that aligns with, or surpasses, the highest levels of the VBC Evidence Pyramid, often requiring peer-reviewed RCTs that directly address financial performance and patient impact. CMS requirements for value-based care models

A Checklist for Evaluating Evidence Quality

For health plan executives and clinicians navigating the crowded AI health market, evaluating evidence quality is paramount. To ensure your chosen AI health vendor can truly participate in value-based care arrangements and deliver on its promises of AI healthcare cost reduction and improved outcomes, consider the following checklist:

1. Peer-Reviewed Publications: Does the vendor have a track record of publishing their outcomes data in reputable, peer-reviewed medical journals (e.g., JAHA, JAMA, JACC)?
2. Randomized Controlled Trials (RCTs): Is their evidence primarily based on RCTs, rather than just observational studies or pilot data?
3. Outcomes Measured: Do the studies directly measure clinical outcomes relevant to your patient population and financial outcomes pertinent to VBC contracts (e.g., reduced hospitalizations, lower readmission rates, decreased medication costs, improved quality of life)?
4. Independent Validation: Are the studies conducted by independent researchers, or are they solely internal analyses? External validation adds significant credibility.
5. Regulatory Adherence: Does the AI solution comply with relevant regulatory frameworks such as the FDA SaMD Framework, and are clinical investigations conducted according to standards like ISO 14155:2026?
6. Transparency: Is the methodology of their studies clearly described, allowing for replication and critical review?

Only by rigorously applying such a framework can we ensure that AI in health moves beyond hype and delivers on its transformative potential within the demanding landscape of value-based care.