In 2026, we’ve moved well beyond ‘AI is coming to healthcare.’ Adoption is accelerating, revealing the importance of reliability and governance to these new systems. While it’s important to understand what an AI model can do, it’s even more important to be 100% confident that your health system can successfully operationalize and deploy it. That, as they say, is where the rubber meets the road, and requires attention to trustworthy patient data as a foundation for AI implementations.
Ultimately, AI is only as safe as the dynamic data quality management and governance that supports it. This leads to two essential data quality questions in healthcare AI. Is the data used to build and validate the model reliable? And is the live patient data encountered by the model actually accurate, complete, current, and connected to the right patient? It’s not enough to be ‘mostly confident’ in either answer. In 12-18 months, healthcare providers and systems can expect tighter coupling between data quality programs and AI approval processes, with unsafe or low‑quality data becoming a formal reason to block or pause AI deployments.
Where we are
The AMA's March 2026 survey [1] found that 81% of physicians now use some form of AI professionally, versus 38% in 2023. To be concise, this does not mean 81% are practicing AI-assisted medicine – rather they are recognizing the value of the technology in ordinary clinical work like summarizing research and standards of care (39%), creating discharge instructions, care plans, or progress notes (30%), or documenting billing codes, charts, and visit notes (28%).
Clearly, administrative and workflow automations have already made the jump from experimentation to routine use. But there is far more to come. Generative AI is advancing in electronic health records (EHR), while predictive AI is making strides in operational areas like re-admission risk, identification of high-risk patients, scheduling and billing. Diagnostic AI, or computer vision and machine learning algorithms, has long been embedded in medical device workflows.
Ready or not
To be AI-ready in healthcare, systems need precise, accurate, dynamically cleansed, enriched and deduplicated data. Why? Deadly AI errors may emerge from poor data quality. If AI confidently orders an incorrectly dosed prescription, the outcome could be tragic. If “route of administration” metadata about a drug dose is incorrect, AI may, for example, apply a higher, potentially lethal dose of a prescribed pill.
Other examples include incorrect risk scores; misrouted, false or missed health alerts; incorrect eligibility decisions. Downstream, this can mean financial harm, erosion of trust in the system, delayed care, wrong treatment, even death.
Duplicate patient records already create substantial business and delivery-of-care problems in both healthcare research and in the clinic. (Some analysts estimate duplicate records at 8-10%. A health system maintaining a million patient records could be dealing with 80,000 duplicates – and their related effects on cost, risk, and correctness of AI-supported decisions at scale.) Fragmented records, for example if a patient record is missing potentially contraindicated supplements or medications, can create even higher risks under AI applications.
Before an AI system can reason accurately across a patient's history, the organization needs confidence that the records actually belong to that patient. Identity resolution, record matching, and deduplication can help establish a reliable, consolidated patient record – not only for safety and excellence in overall care, but before that data becomes source material for AI processes. For example, Melissa applies matching and deduplication technologies to reconcile records across systems, and create a more reliable, unified patient view.
Some of the most significant AI safeguards may occur well upstream of the AI algorithm. Melissa tools and solutions support Project US@-compliant address validation and standardization at patient registration. An initiative of ONC (Office of the National Coordinator for Health Information Technology), Project US@ is a technical specification for patient addresses, designed to prevent variations in names, addresses, and other identifying data that create fragmented or duplicate records downstream.
Data quality infrastructure is AI infrastructure
An excellent AI model can still produce a bad result if it receives an inaccurate representation of the patient. That makes something as mundane as a duplicate record, incorrect identity, outdated address, missing medication, or fragmented medical history consequential in a new way. Ultimately, the infrastructure needed for accurate patient identity and trustworthy records is the same infrastructure needed for safe AI.
Building that foundation doesn't necessarily require healthcare organizations to reinvent their data environments. A good checklist covers end-to-end data quality, starting upstream of AI and continuing through ongoing data monitoring:
- Validate and standardize identity attributes at patient registration, including name, address, and contact data aligned to Project US@.
- Resolve and deduplicate across EHR, claims, ancillary, and affiliate feeds before those records train or trigger AI models.
- Profile the completeness and recency of the fields AI actually consumes, such as medications, allergies, problems, coverage, and contact information.
- Monitor identity breaks continuously and treat them as AI incidents, not back-office cleanup.
- Require data-fitness evidence in the AI approval packet alongside model-performance evidence.
AI readiness is more than treating data quality as an occasional cleanup exercise. Healthcare organizations need mechanisms to continuously profile, validate, cleanse, match, and monitor information as records enter the organization and change over time. Melissa's data quality tools, for example, support those functions across points-of-entry and existing data workflows, helping healthcare organizations establish confidence in their patient data before AI is asked to act on it.
About the Author
Bob Stanley, Melissa’s director of special projects, helps Melissa Informatics’ clients harness the entire data lifecycle for business, pharmaceutical and clinical data insight and discovery. Connect with Bob via email at [email protected] or via LinkedIn.
