Independent Healthcare AI Equity Auditing
Is your healthcare AI working equitably for every patient?
EquiAudit was founded on a simple observation: health organizations are adopting AI tools at speed, with almost no independent way to know whether those tools work equitably for every population they serve.
The EquiAudit Approach
We apply a rigorous, evidence-based methodology to help ensure AI tools perform appropriately across the demographic, socioeconomic, and geographic characteristics of the population they serve.
1. Data Provenance & Representativeness
Who was in the training data — and does that population reflect the demographic, socioeconomic, and geographic reality of the community this tool will serve?
2. Model Development & Validation Equity
Did the vendor follow a sound equity methodology when building and testing the system — including fairness constraints, algorithm selection, and how they handled performance gaps during development?
3. Performance Equity Across Populations
Does the system produce consistent outcomes across the demographic, socioeconomic, and geographic characteristics of the population it serves — or do performance gaps emerge for specific groups?
4. Vendor Equity Literacy & Governance
Does the vendor have the accountability structures, named responsibility, and demonstrated commitment to identify and respond to equity concerns — before and after deployment?
5. Clinical Workflow & Deployment Context
Does the clinical environment this system operates in create conditions that systematically advantage or disadvantage specific patient populations — through data gaps, workflow design, or resource constraints?
6. Community & Patient Voice
Are the communities affected by this system represented in how it is governed and operated — through patient experience data, community health worker input, cultural and linguistic accessibility, and alignment with the National CLAS Standards?
7. Regulatory Compliance & Documentation
Does the organisation's documentation demonstrate equity due diligence against applicable regulatory frameworks — including FDA AI/ML guidance, EU AI Act, CMS requirements, National CLAS Standards, and applicable state legislation?
Strategic Outcome: Every engagement produces plain-language findings that boards, compliance officers, and procurement panels can act on immediately.
About Melinda Kirrane
The problem I kept seeing:
Health equity has been one of public health's most persistent and urgent challenges long before artificial intelligence entered the conversation. The disparities are documented, entrenched, and stubbornly resistant to change — in maternal mortality rates, in chronic disease management, in access to specialist care, in the quality of pain assessment across racial lines. These are not new failures. They are the accumulated consequence of systems built, often without intention, around assumptions that did not hold equally for everyone.
What concerns me now is not that healthcare is adopting new technology. Innovation in healthcare is necessary and overdue. What concerns me is the pace at which AI tools are being deployed into clinical and administrative settings without first asking a foundational question: does this work equitably for the population it will serve?
Why this matters now:
Bias encoded into an algorithm doesn't announce itself. It executes — consistently, invisibly, at scale — long before anyone notices the pattern in the outcomes. And the data it generates trains the next model. Today's disparity becomes tomorrow's infrastructure.
Regulators have noticed. The EU AI Act requires independent conformity assessment for healthcare AI. The FDA has issued bias guidance for AI medical devices. CMS is scrutinising AI claim denials. States are legislating. The question for every health organisation deploying AI is no longer whether to take equity seriously. It is whether you can demonstrate that you did — before something goes wrong.
What EquiAudit does:
EquiAudit provides the independent assessment layer that is currently missing from almost every healthcare AI deployment. Drawing on frameworks including the National CLAS Standards, Bhatt and Gopaliya's (2026) pre/in/post-processing methodology, and FDA and EU AI Act compliance requirements, we conduct structured, evidence-based equity audits across the full AI deployment lifecycle. We tell organisations not just whether a problem exists — but what it is, how serious it is, and what they need to require of their vendors to address it.
The independence principle:
The value of an assessment depends entirely on the independence of the evaluator. EquiAudit has no commercial relationships with AI vendors. We do not offer technical remediation services that would create a financial interest in finding problems. We are the human in the loop — the independent expert judgment that automated tools cannot supply and that internal teams cannot credibly provide for themselves.
​
EquiAudit does not advocate against AI adoption in healthcare. Our role is to help organisations implement AI responsibly, transparently, and with confidence that systems are performing appropriately across the populations they serve.
Credentials & Experience
EquiAudit is grounded in clinical expertise and public health leadership. Our founder brings over a decade of experience in government health IT and clinical research to ensure rigorous, evidence-based assessments.
The Expertise Behind EquiAudit
Our Founder, Melinda Kirrane, has spent over a decade at the intersection of public health policy, health IT modernization, and equity strategy — working directly inside the government health systems that are now the primary adopters of AI tools.
At the Centers for Disease Control and Prevention, she served as Senior Technical Project Manager for the GRASP program — overseeing data systems including the COVID-19 Data Tracker and redesigning the United States Cancer Statistics dashboard. She understands how federal health data infrastructure actually works, which is precisely what determines whether an AI tool built on that data will perform equitably or not.
As a Public Health Practice Lead and Senior Public Health Consultant, she led public health consulting engagements for numerous state health agencies — work that included electronic health record planning, disease surveillance modernization, behavioral health data systems, and health accreditation readiness. She also served as Vice Chair of the Diversity, Equity, Inclusion, Belonging and Access Council, leading the organization's equity strategy at executive level. She has worked directly with the agencies and the data environments that healthcare AI is now being deployed into.
Her academic foundation spans global health policy and clinical practice: a Master of Public Health in Global Health from Imperial College London's WHO Collaborating Centre, and a Bachelor of Science in Nursing from Johns Hopkins University — where her research examined care engagement among women living with HIV in underserved populations. She also brings direct NHS experience from her role at Southwark Council in London, where she led the borough-wide Child Death Review process and the Healthy Weight Strategy serving diverse urban communities.
EquiAudit's assessment methodology draws on all of it — not just the academic frameworks, but the lived operational knowledge of what health systems actually look like from the inside.
Let's talk about what you're building
Whether you're a health system evaluating a vendor, an AI developer seeking independent validation, or a government agency navigating new regulatory requirements — a 20-minute conversation costs nothing and might change how you think about what you're deploying.