Insights
Data trust, provenance, and what AI should be allowed to act on. Written by Jason Alan Snyder, nearly every day. Longer essays live on his Substack.
Children cannot consent for themselves, and their parents cannot consent forever. Pediatric health data governance sits in a legal and ethical gap that most health systems, AI developers, and researchers fail to address. The consequences show up in broken consent chains, unusable training data, and adolescents who avoid care entirely.
Community-based organizations hold some of the most critical social determinants of health data in the Medicaid system, yet most lack the infrastructure to prove that data is trustworthy. Without a formal data trust framework, CBOs cannot participate meaningfully in value-based care contracts that require measurable, auditable outcomes.
Healthcare AI models lose up to 20% of their predictive accuracy within 12 months of deployment because the clinical data they trained on no longer reflects current patient populations. Temporal drift in health data is not a theoretical risk; it is the primary mechanism through which AI systems silently fail in production. Scoring data for recency before it reaches a model is the only structural fix.
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.
Patient consent in healthcare is still governed by static forms signed once and never revisited. Over 80% of patients report not understanding what they consented to, and most health systems have no mechanism to track consent status across data lifecycles. Fixing healthcare consent governance requires architecture, not just policy.
Roughly 46 million Americans live in rural areas where health data is sparse, outdated, or missing entirely. Synthetic data generation promises to fill those gaps, but only if the underlying inputs carry verifiable trust scores. Without scoring, synthetic rural health data inherits the same blind spots it was designed to eliminate.
Sixty-three percent of healthcare AI models retrained in 2024 cited data quality failures rooted in unknown or unverifiable origins. Before any model touches a health record, that record needs a documented chain of custody. Without provenance, AI training is guesswork with clinical consequences.
Community health organizations collect more social determinants of health data than ever, yet most of it fails basic integrity checks. The trust gap between what gets recorded and what can actually inform clinical or policy decisions costs the U.S. healthcare system billions annually and leaves the most vulnerable populations invisible.
The Data Trust Index scores every health data record from 0 to 100 across eight weighted dimensions. This practical guide breaks down each dimension, explains why the weights are set the way they are, and shows how the DTI engine converts raw health data into a trust-scored asset ready for AI, regulatory submission, and clinical use.
When Bobby Hill's father had a stroke and was rushed to Atlantic City Hospital, none of his records were there. He had just spent months watching his family become verbal medical records for their loved ones. This is not exceptional. It is the default — and it costs lives.
The Data Trust Index is now a published, citable academic framework. Today SuperTruth released the first formal paper describing the DTI — the eight-dimensional scoring system that functions as the FICO score for health data. DOI: 10.5281/zenodo.19601616
Lung cancer symptoms often appear months or years before a clinical diagnosis, but behavioral signals in health data can surface risk much earlier. SuperTruth's VIOLET platform scores and interprets these pre-diagnostic patterns using trusted, validated data to close the detection gap.
FDA's AI/ML SaMD action plan identifies training data quality as the most significant gap in the current framework for clinical AI. When an auditor arrives, they will ask six specific questions about your training data. Most health AI companies cannot answer any of them.
NCQA's updated credentialing standards effective July 2025 tighten primary source verification timelines and add continuous monitoring requirements. For health plans, compliance is not just about having the right processes. It is about being able to demonstrate those processes are working — with an auditable, defensible record.
Pancreatic cancer kills 80% of patients within a year of diagnosis because it is found too late. Behavioral data signals, scored and validated through frameworks like the Data Trust Index, could shift detection timelines from late-stage crisis to early-stage intervention.
When a patient signs a HIPAA notice at a hospital, they authorize the use of their data for treatment, payment, and healthcare operations. They do not authorize its use in training an AI model. Most health AI companies are building on a consent foundation that does not exist.
CMS CRUSH compliance is not just about having accurate provider directories. It is about having defensible, auditable data processes. Health plans that cannot demonstrate how they know their data is accurate are exposed even if the data itself looks clean.
Most health plans say they want to deploy AI, but fewer than 30% trust their own data enough to act on AI outputs. Payer data trust starts with measuring data quality at the record level before any model touches a claim, a member profile, or a care decision. SuperTruth's Data Trust Index gives health plans a quantifiable score for every record so they can deploy AI with confidence, not hope.
60 million Americans live in rural areas. They are the most underserved population in health data systems. Not because there is no data about them. Because the data is old.
Health AI doesn't fail at training time. It fails in production, in front of clinicians, when a model makes a recommendation based on data that was stale, unsourced, or never actually authorized for AI use. These failure modes are predictable — and they all trace back to unscored training data.
Primary source verification answers one question: is this license currently active? It does not answer the question that matters for AI-driven health plans: how much should any downstream system trust this record, for this specific use case, right now?
The FDA rejects submissions with incomplete provenance, inconsistent formatting, or unverifiable consent chains. Platinum-grade health data meets every dimension of trust before it reaches a reviewer's desk. SuperTruth's Data Trust Index scores each record across eight dimensions so sponsors know exactly where their data stands.
CMS ACCESS ties federal payments directly to documented health outcomes. The maximum is $420 per beneficiary per year. The minimum is $90. The difference between them is data documentation.
Clinical trial recruitment in oncology fails most patients before they ever hear about a study. Behavioral intelligence, powered by trusted data, can identify the right candidates at the right moment and close the gap between eligible and enrolled.
