CommonWell and Carequality: why health data networks have not solved trust
CommonWell Health Alliance and Carequality connect thousands of healthcare organizations, but connectivity is not trust. These networks solve the transport problem while leaving data quality, provenance, and consent governance entirely unaddressed, which means the health data flowing through them remains unscored and unverified.
CommonWell Health Alliance and Carequality have connected more than 70,000 provider endpoints across the United States. On paper, the interoperability problem looks solved. Clinicians can query for patient records across organizational boundaries. Documents flow from one EHR to another. The pipes work.
But connectivity is not trust. And the health data flowing through these networks carries no score, no provenance chain, no quality guarantee, and no consent verification beyond the minimum HIPAA floor. The result is a national data exchange infrastructure that moves documents without any assurance that those documents are accurate, complete, current, or appropriate for their downstream use.
This is the gap that matters most for healthcare AI, clinical decision support, and population health programs. You cannot build reliable models on unreliable data, no matter how many networks that data traverses.
How CommonWell differs from Carequality
CommonWell Health Alliance is a membership-based nonprofit founded in 2013 by five EHR vendors: athenahealth, Cerner (now Oracle Health), Greenway Health, AllScripts (now Veradigm), and MEDITECH. It operates a centralized patient-linking service that enables record location and document retrieval across member organizations. CommonWell charges membership fees and focuses on ambulatory and post-acute care settings.
Carequality is not a health information exchange. It is a framework. Specifically, it is a set of legal agreements, technical specifications, and governance rules that allow different health data networks to interoperate with each other. The Sequoia Project manages Carequality, and its participants include CommonWell itself, Epic's Care Everywhere network, Surescripts, and dozens of regional HIEs.
The distinction matters. CommonWell is a network. Carequality is a set of rules that lets networks talk to each other. CommonWell connects to Carequality so that its members can exchange data with Epic-based organizations and other Carequality participants.
But neither CommonWell nor Carequality defines what "good data" looks like. Both focus on transport and legal permissions. Neither scores the data that moves through their rails.
What are some common problems with healthcare data?
The problems with healthcare data are structural, not accidental. They persist because no layer in the current infrastructure is designed to catch them.
First, provenance is unknown. A CCD document arriving through Carequality does not carry metadata about how it was created, who entered the data, whether it was manually typed or auto-populated, or how many times it has been transformed. The receiving system gets a document. It does not get a trust score.
Second, duplicates are rampant. CommonWell's patient-linking service attempts to match patients across organizations, but matching rates vary. Studies have shown that patient matching error rates across organizations range from 5% to 20%, depending on the quality of demographic data. A false match means two patients' records merge. A missed match means a fragmented longitudinal record.
Third, data staleness is invisible. A medication list exchanged through these networks might reflect a reconciliation performed three years ago. Nothing in the document header tells the receiving clinician whether the list was reviewed last week or last decade. Recency, which accounts for 15% of the DTI score, is simply absent from exchange metadata.
Fourth, coded data is inconsistent. The same clinical concept can be represented using ICD-10, SNOMED CT, or local proprietary codes depending on the sending system. Terminology mapping failures mean that a diagnosis of "type 2 diabetes mellitus" in one system might not match the same condition in another. We have written extensively about how SNOMED CT mapping failures become a liability and how ICD-10 coding accuracy breaks downstream AI.
Fifth, consent governance is a floor, not a framework. The networks verify that HIPAA treatment-purpose exchange is permitted. They do not track whether a patient has consented to secondary use, AI model training, research inclusion, or data sharing with third-party analytics vendors.
Key statistics
CommonWell and Carequality have achieved impressive connectivity numbers, but the gap between connectivity and trust is quantifiable.
Should I opt in to Carequality?
This question appears frequently in patient-facing forums and provider decision-making conversations. The answer depends on what you expect Carequality participation to accomplish.
For healthcare organizations, Carequality participation enables document exchange with a broad set of national networks. If your EHR vendor supports Carequality (Epic does natively through Care Everywhere; others connect through intermediaries), opting in expands the number of organizations you can query. For providers who treat patients with records scattered across multiple health systems, this is valuable.
For patients, Carequality participation by your providers means your records are more likely to be available when you receive care at a new facility. The 21st Century Cures Act and information blocking rules make it difficult for providers to refuse to participate without a legitimate reason. We covered this regulatory landscape in our post on information blocking rules and data trust.
But opting in to Carequality does not solve the trust problem. Your data will flow. It will not be scored. There is no mechanism within Carequality to verify that the data arriving at the other end is accurate, current, or complete. The framework guarantees transport. It does not guarantee truth.
Is Carequality a health information exchange?
No. Carequality is not an HIE. It is a query-based exchange framework that establishes rules for how HIEs, EHR networks, and other health data networks connect to each other.
Traditional HIEs like those operated by states or regions (examples include Healthix in New York, CRISP in Maryland, or the Indiana Health Information Exchange) are actual repositories or intermediaries that collect, store, and make available patient records. They operate infrastructure.
Carequality operates governance. It publishes the Carequality Connected Agreement, which defines the legal, privacy, and technical terms under which networks agree to exchange data. When an Epic hospital queries a CommonWell-connected ambulatory practice, the Carequality framework is what makes that cross-network query possible.
This distinction is critical because it explains why Carequality cannot solve the trust problem. Carequality governs the handshake between networks. It does not inspect the contents of the envelope.
The CCD document quality problem in network exchange
The primary unit of data exchange across CommonWell and Carequality is the Consolidated Clinical Document Architecture (C-CDA), often in the form of a Continuity of Care Document (CCD). These XML-based documents contain patient demographics, problem lists, medication lists, allergies, procedures, and other clinical summaries.
The problem is that CCD quality is abysmal.
Studying this in practice reveals consistent patterns. Problem lists contain diagnoses from years ago that were never removed. Medication lists include drugs the patient stopped taking. Allergy sections are marked "No Known Allergies" not because the patient has no allergies, but because no one entered them. Procedure histories are incomplete because they only reflect what happened within the sending organization.
When a clinician receives one of these documents through a Carequality query, they have no way to distinguish between verified current information and stale auto-populated entries. The document format does not support data quality metadata. There is no field for "last clinically reviewed" or "confidence level" or "source verification status."
This is the CCD document quality problem, and it is arguably the most consequential hidden trust failure in health data exchange today.
What TEFCA adds and what it still misses
The Trusted Exchange Framework and Common Agreement (TEFCA), managed by the Sequoia Project under ONC's direction, is designed to create a single on-ramp for nationwide health data exchange. TEFCA establishes Qualified Health Information Networks (QHINs) that agree to exchange data under a unified legal framework.
TEFCA improves on Carequality in several ways. It standardizes the legal agreements more tightly. It defines specific Exchange Purposes (Treatment, Payment, Health Care Operations, Public Health, Government Benefits Determination, and Individual Access). It requires QHINs to support both query-based and document-based exchange.
But TEFCA still does not address data trust. The framework specifies how data should move and under what legal authority. It does not specify how to verify that the data being moved is accurate, complete, appropriately consented for its intended use, or recent enough to be clinically relevant.
We wrote about TEFCA's structural limitations in TEFCA and the interoperability imperative. The core argument holds: interoperability solves the plumbing problem. Trust requires a different layer entirely.
Why connectivity does not equal trust
The fundamental confusion in health IT policy over the past decade has been conflating data access with data trust. The assumption has been: if we can get the data to flow, the problems will resolve themselves.
They have not.
CommonWell, Carequality, and TEFCA have each made significant contributions to reducing information blocking and expanding data access. But the downstream consumers of this data, particularly AI models, clinical decision support tools, population health algorithms, and risk adjustment programs, need more than access. They need assurance.
Assurance means knowing where data came from (provenance). It means knowing whether the patient consented to this specific use (consent governance). It means knowing when the data was last verified (recency). It means knowing whether the same fact is represented consistently across sources (concordance). It means knowing whether the data has been validated against an authoritative source (validation).
None of these dimensions are measured by any existing health data network. Not CommonWell. Not Carequality. Not TEFCA. Not any state HIE.
This is why the Data Trust Index exists. The DTI scores every health data record from 0 to 100 across eight dimensions: Provenance (25%), Consent (20%), Recency (15%), Quality (10%), Concordance (10%), Validation (10%), Breadth (5%), and Stability (5%). It operates at the record level, not the network level, because trust is a property of data, not of pipes.
The competitive trust problem between health systems
There is another dimension to the health data network trust problem that the FAQ pages and technical documentation rarely address: competitive dynamics.
Health systems compete with each other. Epic-based academic medical centers compete with Oracle Health community hospitals. Both compete with retail health entrants like Amazon Clinic and vertical integrators like Optum and CVS/Aetna. When these organizations exchange data through Carequality, they are sharing clinical information with competitors.
This creates perverse incentives. Some organizations participate in exchange minimally, sharing only what is legally required and no more. Others flood the network with low-quality CCD documents that technically satisfy information blocking rules but provide little clinical value. The result is a network where participation is technically broad but functionally shallow.
We explored the competitive trust dynamics in detail in Optum data assets and the trust question for competing health systems and Epic vs Oracle Health and what the EHR duopoly means for health data trust.
What the trust layer needs to do
A trust layer for health data exchange must do five things that no current network does.
First, it must score data at the point of ingestion. Before a record enters an AI training pipeline, a clinical decision support system, or a population health model, it needs a trust score. Not after. Before.
Second, it must track provenance across transformations. When a lab result moves from a reference lab to an EHR to a CCD document to a Carequality query response to a receiving EHR, each transformation introduces potential error. The trust layer must maintain a chain of custody record for every data element.
Third, it must enforce consent governance beyond HIPAA. HIPAA permits treatment-purpose exchange. It says nothing about whether a patient consented to their data being used for AI model training, commercial analytics, or secondary research. The trust layer must track five tiers of consent, from clinical care through commercial use.
Fourth, it must measure recency. A blood pressure reading from 2019 and a blood pressure reading from yesterday are not equivalent inputs for a clinical algorithm. The trust layer must time-stamp not just when data was created, but when it was last clinically verified.
Fifth, it must operate without requiring data movement. The zero-copy architecture principle means that trust scoring happens where the data lives, not by pulling records into a centralized repository. We built IntegrityNet specifically for this purpose.
The bottom line for health system and payer leaders
CommonWell and Carequality solved the connectivity problem. They did not solve the trust problem. TEFCA is extending the connectivity solution with better governance, but it too stops short of data trust.
If your organization is deploying AI models that consume data from exchange networks, you are building on an unscored foundation. The data flowing through these networks has no provenance chain, no quality score, no consent verification beyond the HIPAA minimum, and no recency indicator.
The question is not whether your data can flow. The question is whether your data can be trusted.
The DTI Engine scores every health data record 0 to 100 across 8 trust dimensions before your AI model sees it. If your team is evaluating data for training, compliance, or clinical use, and especially if that data arrives through CommonWell, Carequality, or TEFCA exchange, contact Louis Simeonidis 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
The FICO score for health data.
8 dimensions. 0–100. Travels with every record permanently.