Stream: social
Topic: FHIR Usage
Grahame Grieve (Sep 03 2020 at 20:37):
@James Agnew just generated a new map of hapi.fhir.org usage
Grahame Grieve (Sep 03 2020 at 20:37):
Grahame Grieve (Sep 03 2020 at 20:38):
encouraging to see use across Africa and Asia growing. But what is going on with Chicago?
Grahame Grieve (Sep 03 2020 at 20:38):
Usage continues to grow pretty much linearly
Grahame Grieve (Sep 03 2020 at 20:38):
Dave deBronkart (Sep 08 2020 at 18:10):
I just ran across this. Chicago? That's where Leavitt Partners and thus perhaps CARIN Alliance lives; might they be pounding the daylights out of HAPI? @Ryan Howells ?
Yunwei Wang (Sep 09 2020 at 14:55):
Is there any link to get this map?
Grahame Grieve (Sep 09 2020 at 17:23):
http://www.healthintersections.com.au/?p=3046
Melva Peters (Sep 09 2020 at 18:30):
Any chance it will get more granular so we could see which resources are of most interest? Might be a way for WGs to get a sense of the maturity of their resources.
Grahame Grieve (Sep 09 2020 at 18:35):
this is generated on the HAPI API as a whole, so no, I don't think it can
Jens Villadsen (Sep 09 2020 at 18:37):
nop - that wouldn't be possible. If you would like to chase which resources that gains most interest, I would suggest to track all traffic that hits the documentation (on all versions as well).
Jens Villadsen (Sep 09 2020 at 18:38):
and spice it up with all public available IG's and cross it with what resources that gets the most profiling
Karel Groot (Sep 09 2020 at 18:44):
I build an intake app in android that connects via HTTPS to a hosting server that runs in linux. Simple html/php admin panel stored in mySQL. I need to be able to handle hl7 documents. This is a simple questionnaire app that I developed free of charge for a practice here in town and would love to help him exchange data with the other practice that use the FHIR HL7 structure. Any help is appreciated!
Jenni Syed (Sep 09 2020 at 19:53):
For our Cerner US prod cloud, here's an overall ranking of resource popularity DSTU2_Overall.png
Jenni Syed (Sep 09 2020 at 19:54):
And R4 (we haven't produced full like-for-like yet, so apps are slowly moving over or using if it has what they need... some split calls between the two versions): R4_Overall.png
Jenni Syed (Sep 09 2020 at 19:55):
For R4, "Basic" is a FinancialTransaction resource
Jenni Syed (Sep 09 2020 at 19:57):
This is last 30 days only, 90M+ calls
Josh Mandel (Sep 09 2020 at 20:07):
Awesome! (any way to share relative usage -- like is Observation 10x usage from the others?)
Jenni Syed (Sep 09 2020 at 20:08):
Interestingly, it varies quite a bit by persona once you start digging in :)
Jenni Syed (Sep 09 2020 at 20:09):
For the overall DSTU 2, Observations is about 3x the one below it. Then they stay very close/fall off slowly until about Person, then a dramatic fall off at CarePlan
Jenni Syed (Sep 09 2020 at 20:09):
StructureDefinition is only in the 10s of calls... far below others. Not that unexpected though
Jens Villadsen (Sep 09 2020 at 20:19):
@Jenni Syed is your ranking based on page hits on your documentation?
Josh Mandel (Sep 09 2020 at 20:20):
API!
Jens Villadsen (Sep 09 2020 at 20:20):
right
Elliot Silver (Sep 09 2020 at 20:22):
@Jenni Syed, is that ranked by number of instances, access per resource type, or access per resource instance, or...?
Michael Lawley (Sep 09 2020 at 20:22):
It's "interesting" that there are no terminology resources there -- is this popularity by #instances or usage?
Grahame Grieve (Sep 09 2020 at 20:24):
Cerner don't provider terminology services over fhir
Jenni Syed (Sep 09 2020 at 20:28):
This is ranked by call to the resource (both read and write grouped), not instances. EG: a search counts as 1, each read is 1... etc
Jenni Syed (Sep 09 2020 at 20:29):
If you want to compare that list with what we expose on R4 and DSTU2, you can see a chart at the bottom of the page here: http://fhir.cerner.com/millennium/overview/#supported-resources-between-versions
Michael Lawley (Sep 09 2020 at 20:48):
How then do SMART Apps access terminology capability for data entry, for example?
Jenni Syed (Sep 09 2020 at 20:58):
Usually when they first work with the client they determine what makes sense. However, we spoke I think at the connectathon about trying to figure out a better way for someone to call and get "This is what is available in my system (vs all loinc, for EG)" AND make it more user friendly/some challenges with existing terminology capabilities
Jenni Syed (Sep 09 2020 at 20:58):
It is something we'll eventually expose
Paul Church (Sep 09 2020 at 21:00):
Google Cloud Healthcare, by resources in storage, starting from the highest: Observation, Practitioner, Organization, ServiceRequest, Encounter, MedicationRequest, Condition, Patient, Claim, Provenance, MessageHeader, MedicationAdministration, ExplanationOfBenefit, DiagnosticReport, Procedure.
Observation is in the lead by a lot. I don't have stats handy for resources by API calls.
Lloyd McKenzie (Sep 09 2020 at 21:09):
@Karel Groot Suggest raising your question on the #implementers stream - and be sure to set the topic name to be something specific to your question rather than adding it into a current discussion
James Agnew (Sep 09 2020 at 23:37):
HAPI's resource counts look pretty aligned with the Cerner and GCP numbers
Doug Seven (Sep 10 2020 at 05:11):
Microsoft Azure API for FHIR is similar to the others - it varies by persona and scenario, but at scale Observation is in the lead by around 3x. The next grouping of similar scale is Patient, ValueSet, MedicatinStatement and Coverage (all around 1/3 of Observation scale, with Coverage being a little lower). This is measured by API call (all types), not what is in storage, and is a snapshot over a few months. As I said, it varies by persona and scenario.
Martijn Harthoorn (Sep 10 2020 at 08:02):
Some stats from Simplifier projects:
https://simplifier.net/resources
Martijn Harthoorn (Sep 10 2020 at 08:05):
And from the content of files in packages:
image.png
Grahame Grieve (Sep 10 2020 at 11:11):
what are the profiles of?
Martijn Harthoorn (Sep 11 2020 at 08:46):
Dont know :) In the beginning we we didn't track the CoreBaseType, We might fix that later retroacively. So that's why StructureDefinition with null is a big group.
Yunwei Wang (Sep 11 2020 at 19:41):
@Jenni Syed Slot is #3 (for DSTU2) is totally surprise to me. I would think either Medication[xxxx] or Condition but Condition is not even in the top 10!
Yunwei Wang (Sep 11 2020 at 19:49):
I found that almost all the list has Encounter in the top while Encounter has maturity level 2
Jenni Syed (Sep 11 2020 at 20:20):
Scheduling is a very high traffic area for sure :) To be clear, I think observations gets a "bump" in popularity on our server because we have paging... and each page is also a call. And if queries aren't "well bound" there are a ton of vitals, social history results, and labs a patient will have over a lifetime. The other resources are generally less likely to get massive amounts of data behind the scenes, though meds can also get up there :)
Brendan Keeler (Sep 12 2020 at 15:28):
Yeah observation covers so much ground. Seems super logical to me
Ryan Howells (Sep 14 2020 at 22:05):
Dave deBronkart said:
I just ran across this. Chicago? That's where Leavitt Partners and thus perhaps CARIN Alliance lives; might they be pounding the daylights out of HAPI? Ryan Howells ?
It's not us. Maybe BCBSA? AHA? AMA? All have some presence in Chicago.
Last updated: Apr 12 2022 at 19:14 UTC
