Stream: implementers
Topic: Search for deleted resources
Sean McIlvenna (Apr 11 2019 at 05:07):
Is there a way to search for deleted resources? For example, to return a list of all StructureDefinition resources that have been deleted?
Lloyd McKenzie (Apr 11 2019 at 05:39):
No
John Moehrke (Apr 11 2019 at 12:25):
if there is an audit log, then one could find the delete events. - AuditEvent
John Moehrke (Apr 11 2019 at 12:26):
if there is a Provenance following the recommendation for Provenance-of-Delete, then one could find them -- http://build.fhir.org/provenance.html#removal
John Moehrke (Apr 11 2019 at 12:26):
so the answer depends on the policy of the server in question
nicola (RIO/SS) (Apr 20 2019 at 08:45):
In aidbox we keep all deleted resources in history table
Sean McIlvenna (Sep 10 2021 at 21:23):
Is it possible to search for all deleted resources? I know I can see the history of a deleted resource using /fhir/<resourceType>/<id>/_history, but is it possible to see a list of resources that have been deleted? i.e. /fhir/<resourceType>?_deleted=true or /fhir?_deleted=true
Craig McClendon (Sep 10 2021 at 21:49):
I don't believe such an option exists. It would be defined here if it did: http://hl7.org/fhir/search.html
Sean McIlvenna (Sep 13 2021 at 13:47):
Yah I have no doubt _deleted isn't a real thing.. But, wondering if there is something to search for deleted resources that is a real thing :)
Daniel Venton (Sep 13 2021 at 13:58):
From my experience, nothing is ever deleted. In the case of clinical resources the status would be marked with "entered-in-error" or perhaps corrected.
Sean McIlvenna (Sep 13 2021 at 14:06):
soft deleted though
Daniel Venton (Sep 13 2021 at 14:13):
In my dictionary "entered-in-error" status == soft deleted. Everything real should ignore it, unless you are specifically looking for it.
Sean McIlvenna (Sep 13 2021 at 14:21):
in my case, soft deletes are things that have been DELETE'd... you can see the history of that deleted item with /fhir/<type>/<id>/_history, but I need a way of seeing all of the resources that have been DELETE'd
Daniel Venton (Sep 13 2021 at 14:24):
I don't think things in the history are considered deleted, they are old versions. Under that definition of deleted, that would mean you could do a _history search on every item and every version that wasn't the current, is your deleted list.
Notification Bot (Sep 13 2021 at 14:43):
Sean McIlvenna has marked this topic as resolved.
Notification Bot (Sep 13 2021 at 14:43):
Sean McIlvenna has marked this topic as unresolved.
Sean McIlvenna (Sep 13 2021 at 14:43):
ooops... didn't mean to click "resolved"
Sean McIlvenna (Sep 13 2021 at 14:44):
@Daniel Venton, bottom line for me, is FHIR servers commonly allow for a resource to be deleted (not show up in searches) using the DELETE REST method, and I need a way of getting a list of resources that have been deleted
Sean McIlvenna (Sep 13 2021 at 14:45):
regardless of business logic on soft deletes
Lloyd McKenzie (Sep 13 2021 at 16:54):
Right now, there's no mechanism for searching history. Searching deleted resources is essentially searching history.
John Silva (Sep 14 2021 at 01:20):
I suppose the 'backend' (DB or otherwise) knows about this but there isn't a FHIR standard way of asking for it. Not sure but it seems like HAPI has a way of dealing with these DELETE'd resources to "fully expunge' them; not sure if they also have a way of "seeing "these DELETE'd resources. Aren't there certain countries (in EU) that have patient privacy requirements that require knowing about and "fully deleting" patient records if the patient requests it? How would this be accomplished? (of course right now it would probably have to be done in a proprietary way.)
Lloyd McKenzie (Sep 14 2021 at 04:15):
HAPI has a 'purge' operation for full deletion to satisfy EU needs to make something totally go away. It's regular 'delete' mechanism still preserves history.
Grahame Grieve (Sep 14 2021 at 05:51):
yes we explicitly decided to not support searching deleted resources, or historical versions
Sean McIlvenna (Sep 14 2021 at 14:44):
so, any plans do this in the future? seems pretty important to me...
Sean McIlvenna (Sep 14 2021 at 14:44):
R5?
Grahame Grieve (Sep 14 2021 at 17:08):
no plans, no. Why is it important? Deleted data is no longer current
Lloyd McKenzie (Sep 14 2021 at 22:13):
Deleting effectively says "Don't allow search for this anymore"...
Sean McIlvenna (Sep 15 2021 at 14:21):
It's important for a use-case I'm dealing with... Maybe not a common use-case. Surprised this hasn't come up more, though.
Sean McIlvenna (Sep 15 2021 at 14:22):
I'll plan to modify a fork of HAPI to use AuditEvent to track deletions and search the audit events
John Moehrke (Sep 15 2021 at 20:49):
Sean McIlvenna said:
I'll plan to modify a fork of HAPI to use AuditEvent to track deletions and search the audit events
presuming there are AuditEvents being recorded... would love to hear that there are.
John Moehrke (Sep 15 2021 at 20:50):
I would be even happier to know the AuditEvents conform to the IG I am working on, and if not to help me improve my IG. -- http://build.fhir.org/ig/JohnMoehrke/BasicAudit/branches/main/index.html
John Moehrke (Sep 15 2021 at 20:50):
but I have generally been ignored by all the platforms when I ask (except for Grahame)
John Moehrke (Sep 15 2021 at 20:52):
I do have firely audit spec, but it is very different.
John Moehrke (Sep 15 2021 at 20:59):
Theory is that a Provenance of the delete could be recorded too.. but I have never actually made one of those.
Sean McIlvenna (Sep 15 2021 at 21:53):
@John Moehrke , there aren't currently. But I'm planning to create an interceptor on CREATE/UPDATE/DELETE that automatically creates AuditEvent records for non-AuditEvent CUD requests.
Sean McIlvenna (Sep 15 2021 at 21:56):
Taking a look at your IG wrt AuditEvent creation, the biggest problem I see is the requirement of a Patient being involved. In my case, it's organization management... nothing to do with a patient
John Moehrke (Sep 15 2021 at 22:40):
if there is no patient involved, then there is no patient
John Moehrke (Sep 15 2021 at 22:40):
I said that in the narrative that the IG only covers cases where patient is involved.
John Moehrke (Sep 15 2021 at 22:41):
but the profiles could be easy to adjust to remove the patient and use.. let me know how well it goes. Glad to adjust my spec.. also glad to point at an open-source implementation.
Last updated: Apr 12 2022 at 19:14 UTC
