DynaMed

Artificial intelligence is a feature layer on a product whose value stands without it. The asset is a curated evidence corpus built over decades under a stated seven step evidence based methodology, maintained by a named clinical editorial team with daily literature surveillance. Remove Dyna AI and a complete, widely licensed reference product remains, which is what DynaMed was before July 2024.

Dyna AI is a real shipped generative product: retrieval augmented generation over the company's own content sets, extended in February 2026 into Dyna AI Mode, a dedicated interface a clinician can move in and out of, and in December 2025 to drug monograph and dosing summaries from Micromedex data in DynaMedex. It is sold as an add on to a content subscription, and the company's own positioning puts the content, the editorial process and the expert review ahead of the model.

Several real oversight affordances, one part missing, and one claim that needs resolving. Generative output is labeled as such under a published principle, answers link to their source, the corpus is bounded to the company's own curated content, and Dyna AI Mode lets a clinician choose between AI synthesis and classic topic search and move between them rather than routing every query through the model. Dyna AI Mode also captures the question, response, references and a timestamp in copy ready form and lets an earlier query be rerun, so a clinician can preserve the evidence as it stood at the moment of care.

What is missing is what the system does when the evidence is thin, absent or in conflict. EBSCO promotional material has said that every response is validated by a clinician and that there is always an expert clinician in the loop. For a real time generative system, that most plausibly describes clinical curation and review of the content and the tool rather than review of each generated answer. Where the person sits is the most consequential fact about an oversight model and should not require interpretation. No absolute accuracy claim is made.

The approach is named rather than implied, and the model is not. The company states a retrieval augmented generation framework, bounds the corpus to its own DynaMed, DynaMedex and Dynamic Health content sets, limits generative use in user facing applications to information in that curated content, links each answer to its origin and labels AI output. It publishes the composition of the team behind the system, naming physicians, advanced practice providers, pharmacists, nurses and search engineers on a clinical prompt engineering team. Dyna AI was introduced in July 2024, extended to drug monograph and dosing summaries from Micromedex data in December 2025, and moved into a dedicated Dyna AI Mode in February 2026.

No model class or provider is named, no evaluation harness is described, no release notes say when the model changed, and the seven step evidence methodology behind the content is not mapped onto how the generative layer applies it. No benchmark figures are published.

EBSCO describes the architecture but names no model provider. Dyna AI is a retrieval augmented generation system over EBSCO's own content, so a clinician's question reaches a language model, and neither the model nor its provider is identified on the Dyna AI page, the DynaMed product pages or EBSCO's AI and trust pages. No subprocessor list is published, and no retention or training terms for data reaching a model provider are stated.

What is published about the chain is the infrastructure EBSCO runs itself: its ISO/IEC 27001:2022 certificate covers DynaMed, DynaMedex, Dynamic Health and DynaMed Decisions and lists EBSCO's own private cloud data centers in Alabama and New Hampshire, with a privacy management system aligned to ISO/IEC 27701 and 27018. The Dyna AI page states no data selling. DynaMed Decisions draws patient data from the record through SMART on FHIR, and which fields it takes and where they go are not published.

Peer reviewed evaluation of the reference product, not of its generative layer. Edrees, Seger, Amato, Syrowatka, Garabedian, Dulgarian, Schultz, Jackson and Bates, International Journal of Medical Informatics 2025;202:105965 (doi 10.1016/j.ijmedinf.2025.105965), from Brigham and Women's Hospital, Mass General Brigham and Harvard with coauthors from IBM Watson Health, the former owner of Micromedex: two pharmacists ran 600 searches for 300 real world questions in DynaMedex, the joint DynaMed and Micromedex product, using keyword search and the Watson Assistant. Keyword search produced an answer 86.3 percent of the time and the assistant 81.0 percent; 86.0 and 74.5 percent of answers were relevant; most were easy to find; and evidence quality was rated good, fair or poor in 62.7, 36.4 and 0.9 percent of keyword answers. The authors recommend standardizing where drug and disease content sits and citing the highest quality evidence where available.

Dyna AI, the generative layer, launched after that assistant and has no published evaluation of accuracy, error rate or failure modes; EBSCO's Dyna AI page claims it saves clinicians more than two minutes per search, with no method. The company describes qualitative and quantitative research during development without publishing it. DynaMed has been named best in KLAS for point of care disease reference in several years and DynaMedex top performing in 2026, which is a structured survey of customer experience rather than a measure of answer correctness.

Categorical commitments are published, without a retention schedule. The Dyna AI page states no ads and no data selling, ever; a published principle commits to protecting data in accordance with HIPAA standards and to designing and monitoring systems under established AI safety principles; and EBSCO's ISO/IEC 27001:2022 certificate covers DynaMed, DynaMedex, Dynamic Health and DynaMed Decisions within an integrated security and privacy management system aligned to ISO/IEC 27701 and 27018, the standards for personal information management and for personal data in cloud services.

Two products answer the stewardship question differently. Dyna AI reads a curated content corpus rather than a patient record, so the generative layer needs no patient data. DynaMed Decisions takes patient data through a SMART on FHIR integration to pre populate decision aids and calculators, and that module is where retention and use questions land.

Nothing published says how long queries and captured Dyna AI conversations are kept, whether they are used to train or tune any model, or what deidentification applies. No safety engineering is published beyond the corpus bound and output labeling: no guardrails policy and no process for handling a safety event.

Compliance is asserted repeatedly and nothing underneath it can be inspected. DynaMed Decisions is described as a HIPAA compliant clinical decision support solution with a HIPAA compliant EHR integration, and a published principle commits to protecting data in accordance with HIPAA standards. EBSCO's ISO/IEC 27001:2022 certificate, which covers DynaMed Decisions and aligns to ISO/IEC 27701 for personal information, is real assurance about how data is managed.

No business associate status, business associate agreement, terms, execution path or tier is published, and nothing says whether a BAA is available for the reference product and Dyna AI or only for the EHR integrated DynaMed Decisions module. Dyna AI itself reads a content corpus rather than a patient record.

The certificate itself is public, names its standard and version, and covers DynaMed by name. EBSCO's Trust and Security page links a downloadable ISO/IEC 27001:2022 certificate of registration issued by Schellman Compliance to EBSCO Information Services (EBSCO Industries, Inc.), certificate number 1737791-6, version 6, issued 14 November 2024, originally registered 8 December 2020 and expiring 6 December 2026, subject to annual surveillance audits. The services listed in scope include Health Knowledge Products (DynaMed, DynaMedex, Dynamic Health) and Clinical Decision Support Services (DynaMed Decisions, myHealthDecisions), and the scope covers an integrated security and privacy management system aligned to ISO/IEC 27701:2019, with the statement of applicability (version 1.10, 14 July 2024) aligned to ISO/IEC 27017 and 27018. The in scope locations include EBSCO's Ipswich, Massachusetts main campus and private cloud data centers in Alabama and New Hampshire. The page also states ISO 27017, 27018 and 27701 certifications.

No SOC 2 report or penetration testing cadence is published, and the certificate's own conditions note that an ISO audit is not a guarantee against breach. The current certificate runs only to December 2026.

Alongside a sibling module that takes patient data, the company makes no device claim and publishes no regulatory position. For DynaMed and Dyna AI, the main product, the basis of each answer is reviewable: the corpus is bounded and named, output is labeled as generated, each answer links to its source, and Dyna AI Mode keeps a timestamped copy of question, answer and references. That is what the clinical decision support exclusion in the 21st Century Cures Act turns on. The company's own principles call applying generative AI to clinical diagnosis and treatment experimental in nature.

DynaMed Decisions, sold under the same brand, uses a patient's data to calculate an individual risk profile and surface tailored guideline recommendations, a different act from displaying what the evidence says; its recommendations trace to named guidelines, so a basis is shown. No analysis of where either product sits is published, and Dyna AI's withdrawal from the European Union is not explained in regulatory terms.

A governance framework with named process behind it, and no measurement. EBSCO Clinical Decisions publishes five principles for the responsible use of AI: quality, limiting generative use in user facing applications to curated content; security and patient privacy, referencing HIPAA standards; transparency, requiring generative output to be labeled and sources shown; governance, placing clinical experts over development, validation and continuous monitoring; and equity. It commits to ongoing review for bias, quality, safety, ethics, regulatory considerations and scientific rigor, and states that applying generative AI to clinical diagnosis and treatment is experimental in nature. An external expert advisory council includes clinicians, legal counsel and data security and privacy experts, and EBSCO Clinical Decisions has been a member of the Coalition for Health AI (CHAI) since September 2024, which the Dyna AI page repeats. EBSCO also publishes company wide AI tenets.

None of it produces a published result: no bias evaluation, subgroup analysis, error taxonomy or audit output, and equity is named as a principle with nothing published against it.

Mechanisms exist that let a clinician challenge an answer, and nothing stands behind the output. The generative layer is bounded to the company's own named content sets and never draws on external unvetted sources, so it cannot invent a guideline that does not exist, though it can misapply one that does. Each answer links to its origin, generated text is labeled as such, and Dyna AI Mode captures the question, response, references and a timestamp in copy ready form and lets a clinician rerun an earlier query, which preserves what the tool said and when.

No accuracy figure, error rate, warranty, indemnity or correction commitment attaching to an answer is published, and the seven step evidence methodology behind the content is not mapped onto how the generative layer applies it, so a buyer cannot tell whether a generated answer carries the evidence grading of its source or flattens it. The one peer reviewed evaluation of the family (International Journal of Medical Informatics, 2025) measured whether pharmacists could find answers in DynaMedex, not the error rate of Dyna AI.

Named systems with read access and standards support, without published evidence of two way exchange. DynaMed Decisions, the family's patient specific module, was listed in Epic's App Orchard marketplace in June 2021 (Epic has since renamed the marketplace, and nothing on the site shows a current listing), and the company names SMART on FHIR and HL7 as the mechanism, describing the integration connecting to a patient's record to gather the data a decision aid or calculator needs. The practicing clinicians page describes DynaMed Decisions tools that surface guideline recommendations based on each patient's data. Across the product line the named methods include embedded links, search box widgets, the HL7 Infobutton context retrieval API, a web services API, single sign on and a mobile app; Infobutton is the standard built for surfacing reference content in context from inside the record.

There is no published evidence of writing back to the record, or that a named integration runs in production today. Reading patient data into a calculator is one directional.

Options and residency are stated, with isolation and the generative processing path left open. The products are delivered on the web and a mobile app with single sign on, EHR embedded access and multi organization affiliation, with institutional implementation through EBSCO professional services. EBSCO's ISO/IEC 27001:2022 certificate says its management system governs client data that resides in its in scope facilities and lists them: the Ipswich, Massachusetts main campus, offices in Canada and North Carolina, and private cloud data centers in Birmingham, Alabama and Contoocook, New Hampshire. The company states the AI's geographic limits plainly: Dyna AI is currently unavailable in the European Union, with separate material for European audiences, and DynaMed with Dyna AI for individual subscribers is available only in the United States.

The site does not say where the generative layer's model calls are processed or by whom, and it publishes nothing on tenant isolation or on deployment inside a customer's own environment. The generative layer is treated as the regulated component and the content beneath it, still sold in the European Union, is not.

No price is shown, and the sales posture is easy to see. EBSCO's DynaMed pages offer a free trial and a newsletter sign up, individual subscriptions are sold through the DynaMed site, and EBSCO has announced Dyna AI for DynaMedex individual subscribers. Individual prices may sit behind the DynaMed site's subscribe and store pages; no figure, tier or unit of charge is published anywhere else. Dyna AI is sold as an add on to a content subscription.

Institutional pricing, the tier a health system buys and the one that governs Dyna AI, is by sales contact only, with implementation through EBSCO professional services and no rate card, per seat basis or implementation fee published. The company states two limits at the point of sale: Dyna AI is unavailable in the European Union, and DynaMed with Dyna AI for individual subscribers is available only in the United States.

Coverage is named, with validation behind part of it. The family covers distinct segments: DynaMed for disease and condition reference, Dynamic Health for nursing and allied health, and DynaMedex adding Micromedex drug content for care teams and pharmacy, with DynaMed Decisions for patient specific risk tools and shared decision making. Continuing education and maintenance of certification credit accrue from use, and a clinician can affiliate with several organizations.

The reference product's coverage and the AI's coverage differ, and the company says so: Dyna AI is unavailable in the European Union and, for individual subscribers, available only in the United States, so a buyer outside those markets licenses the content, not the generative layer. The one peer reviewed evaluation (International Journal of Medical Informatics, 2025) categorized 300 real world questions by information type and specialty for DynaMedex; no performance evidence is published for Dyna AI in any specialty or setting.