The transition from brokered to networked is essentially the same transition eHX did with IHE from p2p to hub, but networked FHIR also has some data/application value add opportunities. IHE hub is just relaying (but simplifying connectivity).
Think of this as "archiving your history for future use*. The next step requires programming -- e.g., consider raising an issue in an open source podcast app to import play history in this format. Or just use it to mark played episodes by hand during a one-time migration.
Thanks for sharing. This is a very interesting demo! However it is quite intimidating to a patient like me, and a lot more than I may be able to handle without dedicating a good amount of time learning Python and how to implement APIs. Plus, I wouldn't know how to get started with an Epic export file of this magnitude or even know where to go for a copy of my data. For now, I am stuck with web-based interfaces and their limitations. I would love to explore how this can be improved to empower patients like me.
Feedback with regards to AI is mentor that seems to be more of an issue with clinical training and clinical acumen, and imo that's downstream of poor training, poor clinical understanding, inherent uncertainty, and risk averse over investigation and management. I'm not sure it's a huge problem though as most patients can be treated in a pretty standard way and still end up totally fine. We're all inspired by doctors who are devoted, curious, diligent, humble, and have brilliant ideas. But we don't all need that all the time. And probably most doctors are not like that and yet we've done pretty well as a society
Hi Josh, there is a flaw in your way of working. It is very cyclic in nature. You start from structured data, where all expected info is present. You create free text from that, but it is not really free, because it contains all information from the structured file. Of course the AI is able to extract a new structured model, because all expected info is there, because the free text was based on a structured file in the first place. So the feat is less impressive than you make it look. Try this without the first step (the conversion from FHIR to free text) but based on real text from an EHR, and see whether the result is equally impressive.
Thanks for the comments. This exercise is indeed far from rigorous; it's meant to spark ideas and more controlled experimentation. To be clear though, I totally agree that clinical care processes need to collect data in a consistent way, following clinical standards for diagnosis and care. Asking the right questions, performing the right assessments, etc .The interesting question here is whether technical standards are a core part of that data collection loop. Today they aren't often involved during data collection (rather, they're applied on "the way out").
Hi Dr. Mandel, great video. I'm trying to make an application using CDS hooks that parses the written prescription instructions, but I don't believe that the CDS Hooks Sandbox supports written instructions, and I haven't been able to find an alternative sandbox that does. I contacted Epic, and they said that they only support CDS hooks testing with their Vendor Services plan, which is a little out of my college student budget range. I was wondering if you had any suggestions for other sandboxes, or perhaps a self-hosted EMR test server, where I could test order-select CDS hooks with written dosage instructions?
Dear, I really interested in your GPT-4 FHIR plugin ^^. Yesterday, OpenAI annouced about the funcion calling with ChatGPT or OpenAI API. Do you have a plan to build some re-built API for function calling?
This would be killer set of APIs for personal health record applications, and fills a missing gap for an open source PHR I've been developing. Hope to see this in the wild in the near future!
@@JoshMandelMD Will do! Currently studying for boards (DO student), but will flesh out a prototype with the repo mentioned in the video and give some feedback in August.
Thank you for your response on my comment. We are trying to use sushi to create our profiles so can you suggest a tool that can take output of sushi fsh-generated folder and be able to create a static html pages like the IG publisher does.
My best suggestion would be to join the discussion at chat.fhir.org/#narrow/stream/179252-IG-creation and ask the community. It would be important to describe specific examples of what you're trying to accomplish and what you have tried so far. See confluence.hl7.org/display/FHIR/FHIR+Rules+for+Asking+Questions for context.
I have been trying to create IGs within my organization and feels like the IG publisher tool makes a few if not many server calls to create the static web pages of the IG. Is there a plan or a way that this IG publisher can be used by other organizations like a simple web app build so that it could get easier?
The IG Publisher is a command line tool maintained by HL7. There are other tools out there maintained by other organizations that may provide a more friendly interface including simplifier.net/forge and trifolia.lantanagroup.com/
Since I will launch EPIC from hospital, I think I will use EHR launch. In theory, it should be easier than others for authentication. Great to understand little by little.