Regard

The diagnostic engine is the product, not a feature attached to a platform the company already sold. Regard describes a proprietary diagnostic intelligence layer that ingests and maps millions of chart data points to clinical concepts and recommends diagnoses with supporting evidence, and every one of its four modules runs off that same engine. Nothing here is an EHR bundle or a workflow wrapper carrying someone else's intelligence.

The human gate is structural rather than asserted. Regard recommends and the clinician accepts or declines, and the company's own headline metric is the count of recommendations accepted by clinicians, which makes acceptance the unit of account. Recommendations arrive with the supporting chart evidence attached, and a customer quote on the HCC module describes knowing where a condition can be found in the chart and how it is supported.

That is verification made checkable at the point of decision, the same property Abridge earns credit for with Linked Evidence, applied to a diagnosis rather than to note text. Held at B on the published denominator standard: 12,993,284 accepted is a numerator with no denominator, so the acceptance rate cannot be derived, and no confidence threshold, routing rule or abstention behaviour is published. Ask what fraction of recommendations clinicians decline, and what the product does when the chart evidence is weak.

The algorithms are described only as proprietary and clinically validated. No model stack, no accuracy figure, no model card, no validation publication and no error taxonomy was located. For the category's defining question this is the gap that matters: a product that reads the whole chart is judged on what it fails to surface, and no omission rate or false negative measure is published. The company does state that recommendations carry the chart evidence that supports them, which is a meaningful output level disclosure and is graded on the autonomy axis rather than here.

Nothing identifies any party in the chain: no model or model family, no foundation model provider, no hosting arrangement and no sub processor list was located in two passes. The public artifact that would ordinarily carry some of this is unedited template boilerplate, and the detail is worth stating because it is unusually stark: every section is prefixed with the words suggested text, and the content concerns blog comments, avatar services, cookies and image location data.

The placeholder instructions were never removed. It contains no provision on protected health information, retention of ingested chart data, training use or de identification, for a product that ingests the entire medical record. A fairness point belongs alongside and should not be skipped.

Enterprise health system agreements govern patient data through the business associate agreement and the customer's own security review rather than through a website policy, and this product is live at named academic and multi hospital systems that run those reviews, so nothing here establishes mishandling and it should not be read that way. What it establishes is that the public record answers none of these questions and is unmaintained.

Obtain three things in writing: how long ingested chart data is retained, whether customer data trains or improves models, and the de identification posture for anything used beyond the treating encounter.

Well above the tail and short of independent. Multiple named health systems with named clinical executives on record and quantified site level deltas: Sentara Health at a 17 percent increase in CC and MCC capture with a 4x return per user, an appeals win rate reported moving from 20 to 80 percent, 9.3 million dollars in denials prevented at a North Carolina system, a 20 percent query reduction at a Pennsylvania system, and an earlier Torrance Memorial case study reporting 2 million dollars in annual revenue and 20 percent documentation time saved.

That is measured operational outcome at named sites, not deployment volume standing in for evidence. Held at B because it is entirely vendor run and vendor published with no peer reviewed study located, and because the seller chose outcomes that are financial and throughput rather than diagnostic. Nothing published establishes the correctness of the recommendations themselves.

The grade describes the public disclosure surface, not conduct. Regard ingests the entire medical record, and the privacy policy published at regard.com is unedited template boilerplate: every section is prefixed with the words suggested text and the content concerns blog comments, Gravatar, cookies and image location data. It contains no provision on protected health information, retention of ingested chart data, training use or de identification.

Nothing here establishes mishandling, and it should not be read that way. Enterprise health system agreements govern PHI through the business associate agreement and the security review, not through a website policy, and Regard is live at named academic and multi hospital systems that run those reviews. What it does establish is that the public artifact answers none of these questions and is unmaintained.

Three items to obtain in writing: how long ingested chart data is retained, whether customer data is used to train or improve models, and the de identification posture for anything used beyond the treating encounter.

A second pass again located no health privacy compliance statement and no business associate agreement terms on any retrieved surface.

Agreements plainly exist. The product is deployed across named enterprise health systems including academic medical centres and multi hospital systems, and no organisation of that kind onboards a vendor reading its entire record estate without one. What the grade records is that nothing is published, so a prospective buyer cannot establish terms, scope or subprocessor position before entering procurement.

The distinction the earlier assessment drew is the one to keep, and it matters more than usual here because this vendor does hold a real credential that a reader may mistake for this one. Health information technology certification examines a product against defined criteria including privacy and security capabilities. It is not an attestation of compliance with the privacy rule, it says nothing about contractual allocation of responsibility for protected health information, and it is granted under a different framework by a different body. Do not let one stand in for the other.

Scope is the substantive question when the agreement is produced, and the second pass shows why it is larger than a documentation vendor's. The product ingests the complete chart rather than a single encounter, describing millions of data points mapped to clinical concepts, and it operates across a census to identify patients meeting criteria. So it holds protected health information about patients no clinician has used it to document. Its outputs then flow onward into revenue cycle and appeals workflows.

Establish which populations the agreement covers, and what is permitted for census wide analysis as distinct from documenting a patient in front of a clinician.

Stronger than most of this index and short of what a health system security review asks for. Regard's ONC Health IT Module certification covers privacy and security criteria certified for the company itself by an ONC Authorized Certification Body, and the certification number 15.04.04.3192.Rega.01.00.0.240502 is publicly verifiable in the federal Certified Health IT Product List.

That is a materially better artifact than the supplier certification language this index tracks as a watchlist, because it names the criteria, names the certifying body and can be looked up by the buyer. It also lists ISO 9001 as its quality management standard.

Held at B because no SOC 2 Type II, no ISO 27001 and no trust centre was located, and because the multi factor authentication criterion is satisfied by relying on the EHR's own SMART on FHIR authentication rather than by a control Regard owns.

A real regulatory credential exists and it is the wrong framework for the risk the product carries. The ONC Health IT Module certification, dated 2 May 2024 against 2015 Cures Update criteria, covers electronic health information export, authentication and related privacy and security criteria. It is an interoperability and security certification, and its own mandatory disclosure states it does not represent an endorsement by the Department of Health and Human Services.

It is not FDA clearance. No device authorisation was located, and no published clinical decision support exemption analysis was located either, for a product whose core function is recommending diagnoses. Regard's design points toward the exemption criteria, since recommendations are evidence linked to the chart and a clinician can independently review the basis, but that analysis should be documented by the vendor rather than inferred by the buyer.

Worth noting without drawing a conclusion from it: the certified criteria list does not include the Decision Support Interventions criterion. Ask which framework the company believes governs the diagnostic recommendation and to see the reasoning.

The grade describes incentive structure and disclosure, not wrongdoing, and the legitimate argument belongs in the same paragraph. Undocumented comorbidities are a genuine patient safety problem, the conditions named as examples are malnutrition, sepsis and hypertension, the physician remains the decision maker and signs, and every recommendation carries the chart evidence supporting it so a clinician can check whether the condition is actually present.

The company positions itself against query driven capture with the line about earning revenue through better care rather than queries. Against that: the commercial pitch is dominated by revenue, the product ships a dedicated Mid Revenue Cycle module and an HCC Capture module, and the company's own impact calculator states that its recommended diagnoses lead to an upgraded MS DRG for 1.5 to 3 percent of encounters seen and that clients see reduced automated denials protecting a further 1 to 10 percent of charts with upgraded DRGs.

Risk adjustment and severity capture are among the most litigated areas of United States healthcare compliance, and this index has graded milder versions of the same gradient down. Separately, no fairness, subgroup or accent disclosure of any kind was located.

One output level mechanism is real and the measurement that matters for this product type is absent. The company states that recommendations carry the chart evidence that supports them, so a clinician receiving a suggested diagnosis can follow it back to the findings that produced it and reject it on the evidence rather than on instinct.

For a product that reads the whole chart and proposes conditions the clinician may not have considered, evidence carrying output is the right control and it makes an unsupported suggestion visible at the point of review. Held at C because the algorithms are described only as proprietary and clinically validated, with no model stack, accuracy figure, model card, validation publication or error taxonomy located, and no warranty, indemnity or remediation commitment.

The specific gap is the one this category is judged on: a product that reads the whole chart is judged on what it fails to surface, and no omission rate or false negative measure is published anywhere. A missed condition produces no suggestion, no evidence panel and no artefact, so a hospital running this for a year sees the diagnoses it helped capture and never the ones it passed over.

Clinically validated without a method is also the claim shape this index records as weaker than no claim, because it asserts the thing a buyer would otherwise know to ask about. Ask what the validation consisted of, and for the omission rate against a reviewed chart sample.

The strongest interoperability evidence in this category so far, because it is certified rather than claimed. Regard is a certified ONC Health IT Module implementing FHIR single patient and bulk electronic health information export, and it embeds in the EHR through SMART on FHIR, which is also what supplies its authentication. It describes deep data ingestion mapping millions of data points to clinical concepts, and it runs inside the chart at multiple large multi hospital systems.

One gap named for the buyer: the company does not enumerate which EHR vendors are supported, so an organisation outside the major install bases should confirm its own system directly rather than infer it from the customer list.

The ONC mandatory disclosure names the certified product as Regard SaaS Solution v.1, which establishes a single hosted delivery model and no on premise option. Beyond that nothing is published: no hosting provider, no region, no data residency commitment and no customer choice over where ingested chart data is processed. For a product that ingests the entire record this is the open question, and it is the axis where this category will separate the same way the ambient lane did.

No public pricing of any kind. Every route ends at a demo request.

Worth distinguishing from pricing: Regard does publish a detailed return on investment calculator with its assumptions stated, including per query and per note time inputs and national coding benchmarks. Publishing the assumptions rather than only the conclusion is better practice than most of this category manages and it is credited here. It is still transparency about claimed value rather than about what the product costs, and the two should not be confused when comparing vendors on this axis.

Inpatient first and broadening. The product began as a hospitalist tool and the outcome measures it leads with, CC and MCC capture and MS DRG movement, are acute inpatient instruments. Coverage has since been described as system wide across all specialties, and the HCC Capture and Screening modules reach into population and ambulatory work.

Graded B rather than A because the all specialties claim is asserted rather than demonstrated through specialty specific behaviour: no equivalent of the instrument level depth this index credits elsewhere, such as body mapped lesion documentation or paediatric growth charting, was located.