"Used across thousands of hospitals" sounds like institutional adoption on an extraordinary scale. The mental picture is of procurement teams completing evaluations, executives authorising deployment and clinicians integrating a tool into routine care.
That picture may be doing more work than the statement itself.
OpenEvidence's July 2025 release described use across more than 10,000 hospitals and medical centres. The wording includes medical centres, not simply hospitals, and a site-reach claim does not by itself identify the contractual or operational relationship with each organisation.
This is a general problem in healthcare software reporting. Presence, access, approval, integration and adoption can all be commercially meaningful. They are also different facts.
A person can bring software into a building
Consider a hypothetical reference application used by one clinician in each of 1,000 workplaces. It has a geographically broad footprint. Its users may be enthusiastic and may introduce colleagues to the service. None of that establishes 1,000 organisational purchasing decisions.
Now imagine another application formally deployed in 20 hospitals, with several hundred active staff members at each. It has a smaller site count but potentially deeper use within those organisations. Which is more established depends on the question being asked.
For a founder, broad individual uptake might demonstrate distribution. For a hospital buyer, sustained departmental use might be more relevant. For a clinician, either product could be useful. The error is to give the first metric the organisational implications of the second without showing the intervening evidence.
A location in a user's professional profile can establish affiliation. It does not explain whether the employer has reviewed that person's use of the product.
The stages should remain visible
A practical vocabulary makes adoption reporting easier to interpret:
| Stage | What the evidence would need to show |
|---|---|
| Individual use | A person at the organisation performed a defined activity |
| Organisational availability | Relevant staff were given access or made eligible |
| Approved deployment | The organisation authorised a specified use under stated arrangements |
| Technical integration | The product was connected to a defined clinical system or workflow |
| Sustained adoption | Intended users continued using it for relevant tasks over time |
| Demonstrated benefit | An evaluation found an improvement in a specified outcome |
The stages are not necessarily a single compulsory sequence. A reference website may be useful without integration, and an organisation may approve a restricted pilot before a broader rollout. What matters is describing the relationship that actually exists.
A vendor can report several stages at once. It need not compress them into one oversized number.
A real enterprise announcement contains more information
The distinction is visible in Cedars-Sinai's May 2026 announcement. It describes enterprise access to OpenEvidence, connection with patient information from the electronic health record, and organisational review before AI deployment.
That is a more specific proposition than a count of sites where individual users happen to work. It identifies an operational relationship and a clinical context. It also supplies questions that a comparable institution can investigate: which workflows, which staff, which information and what oversight?
Specificity should be rewarded. A short description of a defined deployment may help a buyer more than a much larger count with unclear inclusion rules.
However, even an enterprise announcement does not reveal whether every eligible staff member has started using the service. Access is the opportunity to adopt. Continued activity remains a separate observation.
One health system can contain many sites
Site counting becomes complicated when a single organisation operates several hospitals, clinics and outpatient facilities. The same agreement could be described as one health-system partnership or many care locations. Both descriptions can be correct, provided the unit is explicit.
A clinician can also work across several sites. Counting that clinician once per location may be appropriate for a map of geographical presence. It would be inappropriate for a count of unique people. The mapping from people to sites is not automatically one-to-one.
The same principle applies to institutional names. A university, its medical school, an affiliated hospital and a wider health system can be related entities. A supplier should explain whether a partnership list contains distinct contracts, distinct operating organisations or several names attached to one collaboration.
This is not administrative trivia. It determines what readers can infer about distribution, procurement experience and implementation depth.
A rollout date is not a retention curve
The day a service becomes available is an important milestone. It can reflect extensive preparatory work and meaningful organisational commitment. It remains the beginning of an observation period.
Suppose an imaginary hospital enables 2,000 staff accounts. During the first month, 900 people sign in, 500 complete a relevant task and 240 use the tool during three separate weeks. Each figure tells a different part of the implementation story.
Reporting only the 2,000 enabled accounts describes availability. Reporting only the 900 logins describes initial activation. Reporting the returning group begins to reveal whether the product has found a place in work. None of these figures alone establishes whether time was saved or decisions improved.
The useful follow-up is not another launch photograph. It is a report of what happened after access was granted, with enough context to understand the jobs staff were trying to do.
Historical claims must not rewrite later relationships
A company can grow from individual adoption into enterprise deployments. That is an ordinary and potentially successful development. Earlier reach claims should be interpreted in their original period, while later partnerships should receive credit for what they add.
The October 2026 Johns Hopkins collaboration announcement concerns joint work on medical education and clinical care. It should be read as that particular collaboration, rather than used to retrospectively define every institution included in a previous reach total.
The reverse mistake is also possible. An observer should not dismiss a current enterprise integration because an earlier version of the product was used independently. Dates belong beside claims because the product and its relationships change.
What a clinician should ask before relying on a local adoption claim
For an individual clinician, the immediate question is what their organisation has actually authorised and made available. A product's presence elsewhere does not answer that local question.
Identify the intended task, the version or service offered, and whether the available arrangement is an individual account, a managed organisational service or an integrated workflow. Those distinctions can affect what information is appropriate to use and where support is available.
For procurement teams, request adoption information at the same level as the proposed decision. A department buying a specialised workflow needs departmental experience. An organisation planning broad access needs evidence about activation, continued use and support demands across relevant staff groups.
The active-clinician reporting framework helps distinguish eligibility from behaviour. The guide to evaluating medical AI claims extends that distinction to evidence of performance.
Reach deserves an honest label
Individual adoption can be a powerful route into healthcare. It deserves recognition on its own terms. Formal deployment involves a different relationship, and sustained organisational use requires further work again.
The industry would become easier to assess if companies reported their footprint as a small set of defined measures: organisations under agreement, enabled sites, active sites, unique active clinicians and relevant completed tasks. Buyers could then ask about the part that matters to them.
For iatroX, readers can inspect how the platform works and assess its features for their intended use. The same rule applies throughout the market: a large place-name collection cannot stand in for a precise description of clinical adoption.
Frequently asked questions
Does use at a hospital mean the hospital has approved the software?
Not necessarily. An individual member of staff may use a service independently. Organisational approval needs its own evidence and a clear description of the authorised activity.
Are site counts misleading by definition?
No. They can describe geographical or organisational reach usefully. They become difficult to interpret when the reporting does not define a site, the inclusion threshold or its relationship with the supplier.
What is the best follow-up to an enterprise rollout announcement?
Look for relevant activation and retention data, the workflows being used, implementation experience and evaluated benefits. Eligible accounts and signed agreements explain access; they do not establish the depth or quality of subsequent use.