The VA health system holds records for over 9 million enrolled veterans, making it the largest integrated healthcare network in the United States. But integration on paper does not mean data provenance at scale. The ongoing Electronic Health Record Modernization (EHRM) program, workforce attrition, and legislative uncertainty are compounding a trust problem that no single technology migration can solve.
The Veterans Health Administration manages care for over 9 million enrolled veterans across 1,321 facilities. That makes it the single largest source of longitudinal health data in the United States. It also makes it one of the most complex provenance challenges any health AI system will ever face.
The problem is not volume. The problem is that no one can tell you, for any given VA health record, exactly where it came from, how many times it was transformed, whether the patient consented to its current use, or if it still reflects clinical reality.
The EHRM migration makes provenance worse before it gets better
The VA's Electronic Health Record Modernization program was supposed to replace the legacy CPRS/VistA system with Oracle Health's Cerner platform. The contract, originally valued at $10 billion, has ballooned past $16 billion. Deployments have stalled at five sites after serious patient safety incidents, and the current EHRM deployment schedule remains uncertain.
Every partial migration creates a dual-system problem. Patient records exist in VistA at some facilities and Cerner at others. When a veteran transfers between sites, their data crosses system boundaries with no standardized provenance chain. Metadata about record origin, transformation history, and consent context gets lost in translation.
This is not a theoretical risk. The VA Office of Inspector General documented cases where medication lists failed to transfer correctly between systems, creating direct patient safety hazards. If the underlying records cannot be trusted for clinical care, they certainly cannot be trusted for AI model training.
How is the Big Beautiful Bill going to affect veterans?
The Big Beautiful Bill (the reconciliation package moving through Congress in 2025) includes provisions that would restructure federal workforce spending and potentially reduce VA staffing levels. For data provenance, the downstream effect matters more than the headline. Fewer staff means fewer people validating records, reconciling system conflicts, and maintaining the human oversight layer that catches data errors before they propagate.
The bill also signals broader austerity pressure on federal health IT investments. If EHRM funding tightens further, the VA may be forced to run parallel systems even longer, extending the period where provenance gaps compound.
Why are doctors leaving the VA?
VA physician attrition hit 6.8% in FY2023, driven by compensation gaps (VA physicians earn 20-30% less than private sector peers in most specialties), administrative burden, and frustration with the EHR transition. When experienced clinicians leave, institutional knowledge about local data practices leaves with them. New hires inherit records they did not create, in systems they may not fully understand, with no provenance metadata to guide them.
This workforce churn directly degrades data quality. Every handoff between providers is a potential point where clinical context gets lost and records become less trustworthy.
What is the VA 72 hour rule?
The VA 72 hour rule requires that veterans who present to a VA emergency department must have their clinical documentation completed within 72 hours of the encounter. This rule exists to ensure timely record completion, but compliance varies across facilities. Late documentation introduces temporal provenance errors: the record says one thing, but the timestamp context tells a different story. For AI systems that depend on recency and sequence, these gaps matter.
Has the VA ever laid off nurses?
The VA has historically avoided large-scale nursing layoffs, relying instead on hiring freezes and attrition to manage workforce levels. However, recent federal workforce reduction efforts in 2025 have put VA nursing positions under new scrutiny. Nursing documentation accounts for a significant share of clinical record volume. Any reduction in nursing staff directly reduces the completeness and timeliness of the data that flows into VA health records.
Key statistics
Why provenance scoring has to come before AI deployment
The VA has announced multiple AI initiatives, from predictive models for suicide risk to clinical decision support tools. Each one depends on training data that accurately represents clinical reality. Without provenance scoring, these models inherit every migration artifact, every late-documented encounter, every consent gap buried in the VistA-to-Cerner transition.
SuperTruth's Data Trust Index scores every record from 0 to 100 across eight dimensions. Provenance carries the highest weight at 25% because it is the foundation. A record with perfect quality metrics but unknown origin is still untrustworthy. The DTI framework treats provenance as the first question, not an afterthought.
When we scored 105,000 diagnostic records for imaware, we reduced standardization time from 3 weeks to 2 hours and saved over 200 hours per month. The VA operates at a scale roughly 90 times larger. The math is clear: manual provenance verification does not work at VA scale. Automated, continuous scoring does.
What the VA needs before any model goes into production
Every VA health record that will feed an AI system needs three things: a verified origin chain, a consent status linked to the specific use case, and a recency check confirming the data still reflects clinical reality. The DTI Engine provides all three, scored and auditable.
The alternative is what we already have: AI pilots built on data no one has scored, generating predictions no regulator can fully audit, for a patient population that has already earned the right to better.
To discuss how provenance scoring applies to VA-scale health data, contact Louis Simeonidis, SVP Commercial Operations, at louis@supertruth.ai or (215) 918-4140.
Further reading:
Jason Alan Snyder
Co-founder of SuperTruth and Artists & Robots, and an inventor on the Data Trust Index patents. Twenty-plus years building technology inside Interpublic Group. He writes here nearly every day on data trust, provenance, and what AI should be allowed to act on, and publishes essays on his Substack.
About SuperTruth · LinkedIn · Substack · jasonalansnyder.com
See it in practice
EHR data scored before any AI model sees it.
DTI integrates with Epic, Cerner, and all major EHR systems.