Attribution for AI-mediated and cross-boundary writes: what was produced, when, by which agent (human or non-human), from what inputs, and -- for generated text -- which verbatim source passage each statement rests on. Used standalone against a produced resource, and contained in an AI-suggestion review Task against the suggestion contained alongside it.
Elements / Details
Every element this profile touches, with its full definition. Element names in the tables link here.
A record of what produced or changed a resource: the activity, when it was recorded, the
agents responsible (which may include a non-human agent), and the inputs it derived from.
Primary intended use is AI-action attribution -- keeping an AI-produced resource traceable to
the agent that produced it, the human who accepted it, and the record context it was generated
from. Where the product is generated text, the attributedFact extension carries the trail at
statement grain: each statement with the verbatim quote(s) backing it.
Comments
Some parties may be duplicated between the target resource and its provenance. For instance, the prescriber is usually (but not always) the author of the prescription resource. This resource is defined with close consideration for W3C Provenance.
The metadata about the resource. This is content that is maintained by the infrastructure. Changes to the content might not always be associated with version changes to the resource.
May be used to represent additional information that is not part of the basic definition of the element. To make the use of extensions safe and manageable, there is a strict set of governance applied to the definition and use of extensions. Though any implementer can define an extension, there is a set of requirements that SHALL be met as part of the definition of the extension.
Comments
There can be no stigma associated with the use of extensions by any application, project, or standard - regardless of the institution or jurisdiction that uses or defines the extensions. The use of extensions is what allows the FHIR specification to retain a core level of simplicity for everyone.
The version specific identifier, as it appears in the version portion of the URL. This value changes when the resource is created, updated, or deleted.
Comments
The server assigns this value, and ignores what the client specifies, except in the case that the server is imposing version integrity on updates/deletes.
When the resource last changed - e.g. when the version changed.
Comments
This value is always populated except when the resource is first being created. The server / resource manager sets this value; what a client provides is irrelevant. This is equivalent to the HTTP Last-Modified and SHOULD have the same value on a read interaction.
Identifies EMR instance & pipeline version the resource came from
Definition
A URI that identifies the EMR pipeline and version from which this resource originated. This tells you which EMR instance (recommend using the instance identifier), and the version of the pipeline code/transformations.
Comments
In the provenance resource, this corresponds to Provenance.entity.what[x]. The exact use of the source (and the implied Provenance.entity.role) is left to implementer discretion. Only one nominated source is allowed; for additional provenance details, a full Provenance resource should be used.
This element can be used to indicate where the current master source of a resource that has a canonical URL if the resource is no longer hosted at the canonical URL.
It is up to the server and/or other infrastructure of policy to determine whether/how these claims are verified and/or updated over time. The list of profile URLs is a set.
Security labels applied to this resource. These tags connect specific resources to the overall security policy and infrastructure.
Comments
The security labels can be updated without changing the stated version of the resource. The list of security labels is a set. Uniqueness is based the system/code, and version and display are ignored.
Tags applied to this resource. Tags are intended to be used to identify and relate resources to process and workflow, and applications are not required to consider the tags when interpreting the meaning of a resource.
Comments
The tags can be updated without changing the stated version of the resource. The list of tags is a set. Uniqueness is based the system/code, and version and display are ignored.
A set of rules under which this content was created
Definition
A reference to a set of rules that were followed when the resource was constructed, and which must be understood when processing the content. Often, this is a reference to an implementation guide that defines the special rules along with other profiles etc.
Comments
Asserting this rule set restricts the content to be only understood by a limited set of trading partners. This inherently limits the usefulness of the data in the long term. However, the existing health eco-system is highly fractured, and not yet ready to define, collect, and exchange data in a generally computable sense. Wherever possible, implementers and/or specification writers should avoid using this element. Often, when used, the URL is a reference to an implementation guide that defines these special rules as part of it's narrative along with other profiles, value sets, etc.
Cardinality
0..1
Type
uri
Modifier
yes — This element is labeled as a modifier because the implicit rules may provide additional knowledge about the resource that modifies it's meaning or interpretation
The base language in which the resource is written.
Comments
Language is provided to support indexing and accessibility (typically, services such as text to speech use the language tag). The html language tag in the narrative applies to the narrative. The language tag on the resource may be used to specify the language of other presentations generated from the data in the resource. Not all the content has to be in the base language. The Resource.language should not be assumed to apply to the narrative automatically. If a language is specified, it should it also be specified on the div element in the html (see rules in HTML5 for information about the relationship between xml:lang and the html lang attribute).
A human-readable narrative summary of the NexusEmrCoreProvenance resource for human interpretation.
Comments
Contained resources do not have narrative. Resources that are not contained SHOULD have a narrative. In some cases, a resource may only have text with little or no additional discrete data (as long as all minOccurs=1 elements are satisfied). This may be necessary for data from legacy systems where information is captured as a "text blob" or where text is additionally entered raw or narrated and encoded information is added later.
These resources do not have an independent existence apart from the resource that contains them - they cannot be identified independently, and nor can they have their own independent transaction scope.
Comments
This should never be done when the content can be identified properly, as once identification is lost, it is extremely difficult (and context dependent) to restore it again. Contained resources may have profiles and tags In their meta elements, but SHALL NOT have security labels.
May be used to represent additional information that is not part of the basic definition of the resource. To make the use of extensions safe and manageable, there is a strict set of governance applied to the definition and use of extensions. Though any implementer can define an extension, there is a set of requirements that SHALL be met as part of the definition of the extension.
Comments
There can be no stigma associated with the use of extensions by any application, project, or standard - regardless of the institution or jurisdiction that uses or defines the extensions. The use of extensions is what allows the FHIR specification to retain a core level of simplicity for everyone.
Per-fact citations for AI-generated text (statement + verbatim quotes)
Definition
May be used to represent additional information that is not part of the basic definition of the resource. To make the use of extensions safe and manageable, there is a strict set of governance applied to the definition and use of extensions. Though any implementer can define an extension, there is a set of requirements that SHALL be met as part of the definition of the extension.
Comments
Present when the target is AI-generated text produced under the cite-every-fact contract (summaries, chiefly). Repeats in fact order; order is significant. A generated summary need not be persisted at all -- but when it or its provenance is, this is where the citation trail lives, so the trail survives the write-back at full fidelity.
May be used to represent additional information that is not part of the basic definition of the resource and that modifies the understanding of the element that contains it and/or the understanding of the containing element's descendants. Usually modifier elements provide negation or qualification. To make the use of extensions safe and manageable, there is a strict set of governance applied to the definition and use of extensions. Though any implementer is allowed to define an extension, there is a set of requirements that SHALL be met as part of the definition of the extension. Applications processing a resource are required to check for modifier extensions.
Modifier extensions SHALL NOT change the meaning of any elements on Resource or DomainResource (including cannot change the meaning of modifierExtension itself).
Requirements
Modifier extensions allow for extensions that cannot be safely ignored to be clearly distinguished from the vast majority of extensions which can be safely ignored. This promotes interoperability by eliminating the need for implementers to prohibit the presence of extensions. For further information, see the definition of modifier extensions.
Comments
There can be no stigma associated with the use of extensions by any application, project, or standard - regardless of the institution or jurisdiction that uses or defines the extensions. The use of extensions is what allows the FHIR specification to retain a core level of simplicity for everyone.
Cardinality
0..*
Type
Extension
Modifier
yes — Modifier extensions are expected to modify the meaning or interpretation of the resource that contains them
The resources produced or changed by the activity. For an AI write-back this is the resource
the AI produced -- the drafted DocumentReference, the suggested MedicationRequest, the
Task.
Multiple targets are permitted and are the normal case for a single AI action that writes a
bundle: one Provenance covering the set is preferable to one per resource, because the action
was one decision.
Comments
Per the Nexus EMR logical-reference model these MAY be identifier-only. A consumer resolving provenance MUST handle a reference with no literal reference -- this is the single most common cause of provenance appearing 'missing' when it is present.
Preferred over recorded as THE date. A provenance written asynchronously (a queued write-back) has a recorded well after the action; dating the action to the bookkeeping timestamp misplaces it in every timeline.
The instant of time at which the activity was recorded.
Comments
This can be a little different from the time stamp on the resource if there is a delay between recording the event and updating the provenance and target resource.
Policy or plan the activity was defined by. Typically, a single activity may have multiple applicable policy documents, such as patient consent, guarantor funding, etc.
Comments
For example: Where an OAuth token authorizes, the unique identifier from the OAuth token is placed into the policy element Where a policy engine (e.g. XACML) holds policy logic, the unique policy identifier is placed into the policy element.
An activity is something that occurs over a period of time and acts upon or with entities; it may include consuming, processing, transforming, modifying, relocating, using, or generating entities.
Comments
Deliberately unbound. R4's example binding (provenance-activity-type) does not anticipate AI-mediated authorship -- draft-generated, AI-suggested, human-accepted -- and asserting it would mis-describe the data. Treat the activity vocabulary as open and read activity.text, which is what the examples carry. A validator run against the base spec will warn that no code was supplied; that warning is expected here.
Who or what was responsible for the activity.
For an AI-mediated action there are normally TWO agents and both matter:
• the non-human agent that generated the content -- a Device, identifying the harness
and, where a consumer needs it, the model;
• the human who reviewed, accepted or attested to it -- a Practitioner.
Recording only the human makes AI-drafted content indistinguishable from human-authored
content, which is the specific outcome this profile exists to prevent. Recording only the AI
loses accountability.
Requirements
An agent can be a person, an organization, software, device, or other entities that may be ascribed responsibility.
Comments
Includes the AI agent and its human supervisor where content was AI-drafted.
May be used to represent additional information that is not part of the basic definition of the element. To make the use of extensions safe and manageable, there is a strict set of governance applied to the definition and use of extensions. Though any implementer can define an extension, there is a set of requirements that SHALL be met as part of the definition of the extension.
Comments
There can be no stigma associated with the use of extensions by any application, project, or standard - regardless of the institution or jurisdiction that uses or defines the extensions. The use of extensions is what allows the FHIR specification to retain a core level of simplicity for everyone.
Extensions that cannot be ignored even if unrecognized
Definition
May be used to represent additional information that is not part of the basic definition of the element and that modifies the understanding of the element in which it is contained and/or the understanding of the containing element's descendants. Usually modifier elements provide negation or qualification. To make the use of extensions safe and manageable, there is a strict set of governance applied to the definition and use of extensions. Though any implementer can define an extension, there is a set of requirements that SHALL be met as part of the definition of the extension. Applications processing a resource are required to check for modifier extensions.
Modifier extensions SHALL NOT change the meaning of any elements on Resource or DomainResource (including cannot change the meaning of modifierExtension itself).
Requirements
Modifier extensions allow for extensions that cannot be safely ignored to be clearly distinguished from the vast majority of extensions which can be safely ignored. This promotes interoperability by eliminating the need for implementers to prohibit the presence of extensions. For further information, see the definition of modifier extensions.
Comments
There can be no stigma associated with the use of extensions by any application, project, or standard - regardless of the institution or jurisdiction that uses or defines the extensions. The use of extensions is what allows the FHIR specification to retain a core level of simplicity for everyone.
Cardinality
0..*
Type
Extension
Modifier
yes — Modifier extensions are expected to modify the meaning or interpretation of the element that contains them
The agent's role (author, verifier, attester, ...)
Definition
The role this agent played in the activity -- for example author for the agent that generated
the content, and attester or verifier for the human who reviewed and accepted it.
Required (1..1) in this profile, matching US Core. This is the element that distinguishes a
non-human generating agent from the human who took responsibility for its output; without it, a
Provenance naming two agents does not say which did what.
Comments
Bound extensible to provenance-agent-type by the base. author + attester is the expected pair for an AI-mediated write -- see the AI-drafted-note example.
The agent itself -- Practitioner, Device (AI/system), Organization, Patient
Definition
The individual, device or organization that participated in the event.
Comments
Practitioner is the primary provider identity in this IG. A non-human agent is a Device, referenced logically (by identifier) where the harness graph has no stored Device instance; set Reference.type to Device so a reader can tell what kind of agent it is without resolving. Whether model identity rides on that Device instance is open -- see the token/model accounting note at the top of this file.
The inputs the activity derived from -- for AI, the record context it was given
Definition
The resources the activity used as input. For an AI action this is the record context the model
was given: the resources drawn from the patient compartment that informed the output.
This is what makes an AI output auditable rather than merely attributable -- "this draft was
generated from these twelve resources" is a reviewable claim, where "this draft came from the
harness" is not.
Comments
Records WHICH RESOURCES the activity read, and stays at that grain deliberately: statement-level attribution -- which generated sentence rests on which passage -- rides the attributedFact extension instead, and a citation there MAY point back to the entity it was drawn from. For a chart-wide compartment load, listing every member may still be prohibitively large, and a summarizing convention (the compartment snapshot's own identifier rather than each member) is the expected direction for that case.
May be used to represent additional information that is not part of the basic definition of the element. To make the use of extensions safe and manageable, there is a strict set of governance applied to the definition and use of extensions. Though any implementer can define an extension, there is a set of requirements that SHALL be met as part of the definition of the extension.
Comments
There can be no stigma associated with the use of extensions by any application, project, or standard - regardless of the institution or jurisdiction that uses or defines the extensions. The use of extensions is what allows the FHIR specification to retain a core level of simplicity for everyone.
Extensions that cannot be ignored even if unrecognized
Definition
May be used to represent additional information that is not part of the basic definition of the element and that modifies the understanding of the element in which it is contained and/or the understanding of the containing element's descendants. Usually modifier elements provide negation or qualification. To make the use of extensions safe and manageable, there is a strict set of governance applied to the definition and use of extensions. Though any implementer can define an extension, there is a set of requirements that SHALL be met as part of the definition of the extension. Applications processing a resource are required to check for modifier extensions.
Modifier extensions SHALL NOT change the meaning of any elements on Resource or DomainResource (including cannot change the meaning of modifierExtension itself).
Requirements
Modifier extensions allow for extensions that cannot be safely ignored to be clearly distinguished from the vast majority of extensions which can be safely ignored. This promotes interoperability by eliminating the need for implementers to prohibit the presence of extensions. For further information, see the definition of modifier extensions.
Comments
There can be no stigma associated with the use of extensions by any application, project, or standard - regardless of the institution or jurisdiction that uses or defines the extensions. The use of extensions is what allows the FHIR specification to retain a core level of simplicity for everyone.
Cardinality
0..*
Type
Extension
Modifier
yes — Modifier extensions are expected to modify the meaning or interpretation of the element that contains them
The entity is attributed to an agent to express the agent's responsibility for that entity, possibly along with other agents. This description can be understood as shorthand for saying that the agent was responsible for the activity which generated the entity.
Comments
A usecase where one Provenance.entity.agent is used where the Entity that was used in the creation/updating of the Target, is not in the context of the same custodianship as the Target, and thus the meaning of Provenance.entity.agent is to say that the entity referenced is managed elsewhere and that this Agent provided access to it. This would be similar to where the Entity being referenced is managed outside FHIR, such as through HL7 v2, v3, or XDS. This might be where the Entity being referenced is managed in another FHIR resource server. Thus it explains the Provenance of that Entity's use in the context of this Provenance activity.