State HIE data quality failures that scale with every new connection
State health information exchanges process over 15 billion transactions annually, but fewer than half of states enforce minimum data quality standards on contributing sources. Without scored trust at the record level, HIE scale amplifies bad data instead of solving fragmentation.
State health information exchanges now operate in all 50 states and the District of Columbia, according to ONC's 2024 Health IT Dashboard. They collectively route billions of clinical messages per year. Yet the fundamental question most HIE governance boards cannot answer is simple: how good is the data flowing through the pipe?
The answer, across most state HIEs, is that nobody knows. And that is the problem.
The volume illusion in state HIEs
| | Value (%) | |---|---| | Hospitals that send data electronically | 96 | | Hospitals reporting received data is usable | 73 |
State HIEs measure success by transaction volume. Indiana's IHIE processes over 100 million transactions annually. New York's SHIN-NY network connects more than 70 participating organizations. These are real accomplishments in connectivity. But connectivity without data quality governance is a pipe that moves contaminated water faster.
A 2023 ONC data brief found that while 96% of non-federal acute care hospitals electronically sent health information to outside organizations, only 73% reported that the information they received was usable without manual intervention. That 23-point gap between sending and receiving usable data is the data quality crisis hiding inside every HIE throughput metric.
Volume without quality scoring creates a specific failure mode: clinicians learn to distrust HIE data and stop using it. A 2022 JAMIA study found that provider satisfaction with HIE data dropped significantly when records lacked structured medication lists, problem lists, or allergy documentation. The pipe works. The data does not.
What state HIE governance typically requires, and what it misses
Most state HIE participation agreements specify technical requirements: HL7 v2 message formatting, C-CDA document structure, connectivity to a master patient index, and compliance with the state's consent framework. These are necessary. They are not sufficient.
What these agreements almost never specify:
The Sequoia Project's Carequality framework and CommonWell Health Alliance both define technical interoperability requirements. Neither defines what constitutes a trustworthy record. The distinction matters: interoperability is a transport problem, trust is a data problem. Solving one does not solve the other, as we covered in CommonWell and Carequality: why health data networks have not solved trust.
The C-CDA quality problem at HIE scale
The Consolidated Clinical Document Architecture is the dominant document standard in state HIE exchange. It is also, as documented by ONC's own C-CDA Scorecard, frequently malformed.
ONC's Scorecard evaluates C-CDA documents across multiple rubrics including vocabulary conformance, vital sign units, and medication coding. In practice, many documents score poorly. A 2019 analysis published in the Journal of the American Medical Informatics Association found that 35.5% of C-CDA documents from a national sample had at least one critical error in medication encoding. These are not edge cases. These are the documents flowing through state HIEs every day.
The C-CDA problem compounds at scale. When a state HIE aggregates documents from 200 contributing organizations, each with different EHR vendors, different documentation practices, and different levels of clinical informatics staffing, the result is a repository where document quality varies wildly and no consumer of the data can distinguish a high-fidelity discharge summary from a skeleton document without manual review. We analyzed this pattern in detail in The CCD document quality problem: hidden trust failures in care summary exchange.
TEFCA raises the bar, but not on data quality
The Trusted Exchange Framework and Common Agreement, administered by the Sequoia Project as the Recognized Coordinating Entity, establishes a national floor for health information exchange governance. TEFCA's Common Agreement specifies participant obligations around patient matching, record location, and privacy requirements.
TEFCA is a significant governance improvement. It standardizes consent policies, requires participant vetting, and creates a dispute resolution framework. What it does not do is score the data that flows through it.
A Qualified Health Information Network under TEFCA is obligated to ensure technical connectivity and policy compliance. It is not obligated to verify that the clinical content within a C-CDA document is complete, current, or concordant with other records for the same patient. TEFCA solves the trust-of-the-network problem. It does not solve the trust-of-the-record problem.
This distinction will become critical as TEFCA scales. The ONC 2024 Interoperability Roadmap update projects TEFCA participation growing to cover the majority of US healthcare organizations. More participants means more data volume, and without record-level quality scoring, more volume means more noise.
For health systems preparing for TEFCA's requirements, the data foundation question comes first, as we covered in TEFCA and the interoperability imperative: what health systems need to prepare.
Patient matching failures multiply data quality problems
State HIEs depend on probabilistic patient matching to link records from different contributing organizations. Without a national patient identifier, every HIE builds its own matching algorithm using combinations of name, date of birth, Social Security number fragments, address, and other demographics.
The Pew Charitable Trusts' 2018 report on patient matching estimated that match rates across organizations range from 80% to 95%. That means between 5% and 20% of records either fail to link to the correct patient or link to the wrong patient.
A false negative match, where records for the same patient are not linked, means a clinician sees an incomplete picture. A false positive match, where records from two different patients are merged, creates a safety hazard. Both failures are invisible at the transaction level. The HIE reports a successful exchange. The data is wrong.
The 2020 OIG report on patient matching in Medicare found that approximately 1 in 5 patient records had at least one demographic error that could affect matching. Scale that across a state HIE with millions of records from hundreds of sources, and the cumulative error rate becomes a structural reliability problem.
Consent governance varies by state, and so does data trust
State HIEs operate under different consent models. Some states, like New York, require explicit opt-in consent before a patient's records can be shared through the HIE. Others, like Indiana, use an opt-out model where records flow unless the patient actively objects.
The consent model directly affects data completeness. Opt-in states consistently report lower participation rates. New York's SHIN-NY network has historically faced challenges with consent capture, as documented in the New York eHealth Collaborative's annual reports. Opt-out states have higher data volumes but face ongoing questions about whether patients meaningfully understand their data is being shared.
Neither model addresses consent granularity. A patient who consents to sharing a broken arm record through the HIE may not intend for their substance use disorder treatment records to flow through the same pipe. 42 CFR Part 2 provides federal protections for substance use disorder records, but implementation varies across state HIEs, and compliance gaps create real risk.
Consent is the second-highest weighted dimension in data trust scoring, at 20%. For good reason. A record shared without proper consent is not just a privacy violation. It is a trust failure that undermines the entire exchange. We covered this in depth in Why consent governance fails in healthcare data and what fixes it.
The eight dimensions state HIEs should measure but do not
| | Value (%) | |---|---| | Provenance | 25 | | Consent | 20 | | Recency | 15 | | Quality | 10 | | Concordance | 10 | | Validation | 10 | | Breadth | 5 | | Stability | 5 |
Health information exchange trust requires measurement across multiple dimensions, not just connectivity. The eight dimensions that determine whether a record is trustworthy enough to act on are provenance, consent, recency, quality, concordance, validation, breadth, and stability.
State HIEs today measure, at best, two of these: basic quality checks on document formatting and consent status. The other six dimensions go unscored.
Provenance requires knowing exactly which system generated the record, who authored it, and what transformations it underwent before reaching the HIE. Most state HIEs cannot answer this question for any given document in their repository.
Recency requires knowing when the clinical content was last verified, not just when the document was transmitted. A C-CDA sent today can contain a medication list last updated 18 months ago. Without recency scoring, the HIE treats it as current.
Concordance requires checking whether data from multiple sources for the same patient agrees. When two contributing organizations send conflicting allergy lists for the same patient, the HIE has no automated mechanism to flag the conflict or determine which source to trust.
Validation requires confirming that coded values are clinically plausible. An ICD-10 code for a condition physiologically impossible in the documented patient population, or a lab value outside any physiological range, should be flagged before the record is served to a clinician or an AI model.
Without scoring across all eight dimensions, state HIEs cannot distinguish between a high-fidelity, fully provenance-tracked, consent-verified, clinically concordant record and a skeleton document with stale data and no author attribution. Both are treated as equivalent transactions.
What HIE data quality failure costs downstream
The downstream cost of unscored HIE data is measurable. When clinicians cannot trust HIE data, they order duplicate tests. A 2014 study in Health Affairs estimated that health information exchange could reduce duplicate imaging by 12.4%, but only when clinicians actually use the exchanged data. Trust failures reduce utilization, and reduced utilization eliminates the cost savings HIE was designed to create.
Population health analytics built on unscored HIE data produce unreliable results. State public health agencies that query HIE data for disease surveillance, syndromic surveillance, or social determinants tracking inherit every quality problem in the underlying records. A 2021 CDC MMWR report on electronic case reporting through HIEs documented significant variability in data completeness across reporting jurisdictions.
AI models trained on unscored HIE data amplify these problems. A predictive model for readmission risk that trains on incomplete discharge summaries from an HIE will learn the wrong patterns. The model does not know which records are incomplete. It treats every record as ground truth. The result is a model that performs well on training data and fails on real patients, a pattern we analyzed in Readmission prediction model bias: how training data trust affects clinical AI.
What record-level trust scoring changes for HIE governance
The fix is not better pipes. The fix is scoring every record before it enters the pipe.
A trust score on every record entering a state HIE transforms governance from a connectivity standard into a data quality standard. Contributing organizations that consistently submit low-scoring records can be identified and remediated. Clinicians querying the HIE can filter results by trust tier, seeing high-confidence records first. Population health queries can set minimum score thresholds, ensuring that analytics are built on records that meet provenance, recency, and concordance requirements.
This is not hypothetical. The DTI framework scores records 0 to 100 across all eight dimensions. A state HIE that applies record-level scoring at ingestion can publish data quality dashboards for every contributing organization, enforce minimum score floors for different use cases, and give downstream consumers, whether clinicians, public health agencies, or AI models, a quantitative basis for deciding how much to trust any given record.
The alternative is what we have today: billions of transactions flowing through state HIEs with no way to distinguish signal from noise.
The DTI Engine scores every record 0 to 100 across eight dimensions before your AI model sees it. If your team is building on HIE data for population health, clinical decision support, or public health surveillance and needs to know which records to trust, talk to the SuperTruth commercial team. Schedule a conversation or call (215) 918-4140.
Further reading:
Sources
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
DTI scores the record, not the patient.
8 dimensions. 0 to 100. Travels with every record permanently.