Credo Health

Model work sits at the center: inferring that a record should exist from signals in the records you already hold is a real inference task, and agent assisted workflows drive the recovery.

But the company describes its offering as combining digital and manual retrieval with AI driven analysis and human expertise, and the value a customer buys includes an operational capability, chasing charts by telephone, portal and fax, that is not reducible to a model. That operational reach is the differentiator as much as the inference is.

A completeness signal and source citation are described, and the inference behind them is unmeasured. Earlier product material describes Care Map telling the user what is still missing rather than silently delivering an incomplete picture, and every element of the assembled history carries its source; suspected conditions are surfaced for a physician to review, and the terms leave coding decisions with the customer. A clinician can see where each fact came from and, on that description, where the holes are.

The home page promises "near-perfect patient histories" and "near-complete longitudinal records", bounded claims, and says the platform helps physicians "perfectly diagnose and treat every patient"; the full sentence makes the physician the one who diagnoses, so it does not grade here, but it tells a clinician the history is enough, which is the reliance a completeness signal exists to check. No false positive rate is published for records the system believes should exist, which sends staff chasing phantoms, and no false negative rate for gaps it fails to notice.

Output is source cited, which matters a great deal because the history is stitched from many organizations and a clinician needs to know which one each element came from. Beyond that nothing is disclosed: no model or model family named, no accuracy figure for encounter footprint inference or gap detection, no evaluation methodology. The figure the company does publish, that 30 to 50 percent of high value encounters are not returned digitally, is a claim about the national networks rather than about its own performance.

The retrieval chain is named, and the model layer is not. The master terms name the exchange frameworks Credo retrieves through (the eHealth Exchange data use agreement, Carequality, CommonWell and TEFCA), and the business associate addendum names an affiliate that handles records, Medical Records Exchange LLC, doing business as Chartfast, alongside Credo's own entities; the addendum binds any subcontractor to the same restrictions. The home page describes agents for record acquisition and analysis and point of care copilots that use generative AI inside the clinician's record system.

The model provider behind the generative copilots and the analysis agent is not named on the home page, terms, addendum or privacy policy, and no hosting provider or subprocessor list is published. Records also move through a release of information partner and through telephone and fax recovery, channels with no access control at the receiving end.

Named customers and vendor reported figures, none of it outcome data. The home page shows ChenMed, Clover Health, Imperial Health, P3, Longevity, Bluezones, CommonGround and other value based care organizations as customers, and reports a 50 percent or greater reduction in chart preparation time, lower chart acquisition costs and gains through improved risk and quality scores. The company also reports survey findings from more than 500 physicians and care team members, a point of care beta with more than 150 clinicians, and pilot work in Medicare populations; record increase figures are described as estimates based on partner reported data.

The figures carry no method, baseline or period, no named customer is tied to a result, and nothing measures whether more complete records changed a clinical decision or an outcome.

Training use and de identification are stated in the contract, and no retention period is. The master terms (last updated 29 August 2024) license customer data to Credo for improving and developing its services, including training its artificial intelligence and machine learning systems and creating de identified data sets, and make that license perpetual for any customer data Credo retains after termination. The business associate addendum authorizes de identification under the HIPAA standard, limited data sets and data aggregation, and requires records to be returned or destroyed at termination unless that is infeasible. Those are categorical positions Credo can be held to, and they run toward Credo's use rather than the customer's protection.

No retention period is stated for retrieved records or for the assembled history, which Credo alone holds, and the terms leave room to keep data where return is infeasible. The website privacy policy, last updated in 2022, speaks of minimum necessary use of PHI and nothing more. No safety engineering is published beyond source citation.

The business associate agreement is published, names the tier it applies to, and binds subcontractors. Credo publishes its Business Associate Addendum for enterprise customers (last updated 29 August 2024), incorporated into its master terms and covering Credo Health Solutions Holdings and its affiliates, including Medical Records Exchange LLC, doing business as Chartfast. It applies whenever Credo acts as a business associate or subcontractor business associate, requires every subcontractor that handles protected health information to agree in writing to the same restrictions, commits to report breaches and security incidents within ten business days, reimburses the customer's notification costs and a year of credit monitoring for a breach Credo causes, and provides for return or destruction of records at termination. The master terms add a verification step: the customer must show it is a HIPAA regulated entity before any record is retrieved.

The addendum also lets Credo de identify records, create limited data sets and provide data aggregation, and Credo may amend it on 90 days' notice, the customer's remedy being to stop using the service. Who is the requester of record when staff chase a chart by telephone or fax is not stated.

Controls are asserted and nothing independent stands behind them. The website privacy policy (last updated 1 March 2022) says the company uses reasonable organizational, technical and administrative measures, and the business associate addendum commits Credo to the HIPAA security rule's safeguards for electronic records. No SOC 2, HITRUST or ISO 27001 attestation, penetration testing statement or trust center appears on the home page, terms, addendum or privacy policy, and the footer's "HIPAA Security Policy" item links nowhere.

The exposure is specific. The business assembles complete clinical histories from many holders for patients who never chose the company, adds an affiliated retrieval business and a release of information partner, and recovers records by telephone, portal and fax, so the surface spans systems it runs and channels it does not. Participation in national exchange frameworks is a condition of connecting, not an audit of how Credo operates security.

No device claim is made, and the marketing sits in tension with the terms. The home page says the platform helps physicians, MSOs, ACOs and health plans "perfectly diagnose and treat every patient", and its Inspect agent analyses clinical data to deliver actionable intelligence that informs next steps, improves coding and documentation and raises quality scores; the company's announcements describe suspected risk adjustment conditions surfaced for physicians to review. The master terms make any diagnosis code assigned for risk adjustment from Credo's data the customer's sole responsibility. The marketing claims help for the physician's diagnosis rather than a diagnosis by the product, and suspected conditions come with source cited records a clinician can check, so this is a tension, not a contradiction.

The rules that bite are information exchange and release of information law, and the risk adjustment audit regime, where suspected conditions drawn from retrieved records meet active enforcement.

The commercial incentive is stated plainly, and no fairness or performance by group is published. The home page sells the platform to value based care organizations on risk and quality scores: the Inspect agent improves coding and documentation and raises quality scores, the Optimize dashboards show how that work translates into risk adjustment factor scores, and the home page lists more diagnoses among the results. The mechanism is retrieval rather than aggressive coding: conditions that are documented at another organization are brought into view, which also improves care independent of payment. Retrieved documentation feeds retrospective risk adjustment, though, which is where audit scrutiny concentrates, and no governance framework or monitoring for that gradient is published.

The equity point is specific. Patients whose care is most fragmented across organizations, and whose records sit in systems that exchange poorly, are disproportionately those with unstable housing, frequent moves or coverage churn. If retrieval completeness varies by population, so does everything downstream, and nothing measures it.

Every element is cited to its source, and the contract stands behind none of it. Credo presents the assembled history with source citation, so a clinician can see which organization each fact came from and weigh it; a medication list from the prescribing practice and one transcribed from a fax carry different weight. The citation is how a clinician can challenge what the history says.

The master terms (last updated 29 August 2024) provide the services and data as is, disclaim any warranty that they are accurate, complete, free from false matches or error free, state that matching a patient across record sources is inherently prone to mismatching, and make risk adjustment coding from Credo's data the customer's sole responsibility. Liability is capped at the prior 12 months' fees, apart from breach notification costs under the business associate addendum, capped at $1 million. No accuracy or recall figure for finding the records that should exist is published. A history presented as complete that quietly omits an encounter, or merges another patient's record, is the failure a citation cannot show.

Named record systems and national networks, with the depth of each integration left open. The home page says Credo integrates with leading record systems and shows athenahealth, eClinicalWorks, Elation Health, Greenway Health, NextGen, MDLand, Office Ally and Practice Fusion, with point of care copilots working inside the clinician's record system; the master terms name the exchange frameworks Credo retrieves through, the eHealth Exchange data use agreement, Carequality, CommonWell and TEFCA, alongside regional exchanges. Where the networks come back empty, the company recovers records from source providers by telephone, secure portal and fax, which reaches holders no network does, and stitches everything into a source cited history.

The record systems are shown as logos without saying which are read, which receive results and through what interface, no marketplace listing or integration documentation is published, and earlier material placed point of care delivery through a middleware partnership rather than native integration.

A single hosted service with United States servers stated for personal information and nothing committed for records. The website privacy policy (last updated 1 March 2022) says personal information is transferred to the company's facilities and servers in the United States, and the terms describe a hosted platform reached through the customer's record system. Records enter from many directions: national exchange networks, an affiliated retrieval business, a release of information partner, and telephone, secure portal and fax recovery where a holder supports nothing better.

The privacy policy's scope is website and app users, not the clinical records retrieved for customers, and no cloud provider, region, tenancy model or residency commitment for assembled histories is published on the home page, terms, addendum or privacy policy. A received fax lands somewhere before it is digitized, and where that is, and what protects it, is a deployment question. The accumulated store of assembled histories, and whether one customer's retrievals are separated from another's, is the other.

A pricing basis is published without figures. An article republished on the company's own blog, from its 2022 launch, says Credo charges a flat fee for each successful placement of data and then a subscription fee for keeping those records current, focused on Medicare Advantage and other value based care providers. The master terms describe enterprise contracting through order forms, after a verification process that confirms the customer is a HIPAA regulated entity. No fee amounts or tiers are published.

The basis dates from 2022, before the current agent products, and nothing says whether it still applies or whether manual retrieval by telephone and fax, which the vendor pays for per record chased, is priced separately from exchange retrieval, which is close to free at the margin. A patient with care spread across a dozen organizations is the one the product is most valuable for and most expensive to serve.

Coverage is defined by buyer and by population rather than by specialty, and the population argument is a sharp one. Three buyer types are addressed, value based care providers, management services organizations and health plans, and the company singles out Medicare populations as the setting where care is most distributed across organizations and where retrieval therefore matters most.

That is a coherent targeting argument rather than a generic claim, since the value of finding records elsewhere scales directly with how fragmented a patient's care has been. No clinical specialty depth or instrument level behavior is described, and the workflow addressed is pre encounter preparation rather than a range of care settings.