FHIR Chat · Hooks as "activity types" · cds hooks

Stream: cds hooks

Topic: Hooks as "activity types"


view this post on Zulip Josh Mandel (Jun 06 2019 at 20:52):

We've been trying to figure out the relationship between CDS Hooks (like order-select or patient-view and something that the SMART Web Messaging spec calls an activityType. In Web Messaging, the idea is that an app can request a navigation event in the host EHR, e.g., to navigate to a particular screen or activity such as "problem list" or "orders" or "patient face sheet".

view this post on Zulip Josh Mandel (Jun 06 2019 at 20:53):

I'd love to tell a simple story like "for every activity type we care about, there's a CDS Hook to represent that point in the navigation". This would avoid trying to model the same basic workflows in two competing ways.

view this post on Zulip Josh Mandel (Jun 06 2019 at 20:54):

In practice, this might mean leaning on existing hooks, and defining new ones as the need arises, like allergy-list-view or problem-list-view. Does this strike anyone as a fundamentally bad/backwards idea?

view this post on Zulip Lloyd McKenzie (Jun 06 2019 at 22:24):

I think there's a many::1 relationship between screens and CDS Hooks.

view this post on Zulip Lloyd McKenzie (Jun 06 2019 at 22:25):

order-select could easily be triggered from a pharmacy screen, a lab screen, a general order screen (referrals, etc.) and maybe one or two others.

view this post on Zulip Lloyd McKenzie (Jun 06 2019 at 22:25):

Similarly the patient-admit could show up a few different places too (e.g. inpatient vs. outpatient workflows)

view this post on Zulip Lloyd McKenzie (Jun 06 2019 at 22:26):

One of the points of the cds hook definitions was to be independent of the particular "screens" and focus more on what type of workflow was happening. If you're going to do navigation, you're going to need specific screens.

view this post on Zulip Josh Mandel (Jun 06 2019 at 23:33):

But navigation is also not about screens per se -- it's about functionality.

view this post on Zulip Josh Mandel (Jun 06 2019 at 23:33):

It's about navigating to activities the same way CDS Hooks is about notifying when activities are occurring.

view this post on Zulip Josh Mandel (Jun 06 2019 at 23:34):

In practice, responding to a navigation request will mean picking a particular screen -- but that could be considered an EHR-internal detail, or might even be configurable.

view this post on Zulip Josh Mandel (Jun 06 2019 at 23:35):

I should also say that I don't think every CDS Hook would need to be associated with a navigatable activity; just the other way around (for each navigatable activity, we'd need to ensure a hook existed).

view this post on Zulip Lloyd McKenzie (Jun 07 2019 at 02:35):

Ok, but if order-select corresponds to 3-4 different functional areas of an EHR, as does patient-view or encounter-start, what would it mean for an activityType to say "transition to order-select" or "patient-view"?

view this post on Zulip Isaac Vetter (Jun 07 2019 at 03:27):

The narrow, accurate path is to only align view-type hooks with standardized activity names. Josh, note that you did this, implicitly, above. However, view hooks tend to be less actionable than action hooks (select, sign, create) and therefore less valuable. This is demonstrated by our list of current, proposed, and defunct hooks: order-select, order-review, order-sign, med-prescribe, appt-book, encounter-start, encounter-discharge, med-refill, allergy-create, problem-create and, of course, patient-view. Note that all but one of these hooks is a user action, not an activity view.

view this post on Zulip Isaac Vetter (Jun 07 2019 at 03:28):

By limiting SMART messaging to use CDS Hooks-aligned activityType, we'd simply encourage the creation of fluff in CDS Hooks.

view this post on Zulip Isaac Vetter (Jun 07 2019 at 03:28):

(Note that we're attempting to standardize clinical workflow -- this is the realm of BPMN and such and can get really complicated really quick).

view this post on Zulip Josh Mandel (Jun 07 2019 at 15:23):

OK - so I'm hearing "no, go build a separate catalog of activities that will overlap with CDS Hooks *-view hooks but will be maintained independently". Feels like a unnecessary duplication to me, given that the CDS Hooks catalog is open-ended (i.e., one person's "fluff" is another's substance).

view this post on Zulip Josh Mandel (Jun 07 2019 at 15:23):

But I won't push it if there's no support. So: @Isaac Vetter are you willing to help me iterate on an navigation-relevant activity list that makes sense in Epic? Basically, a list of view-type activities that should serve as the target of navigation requests?

view this post on Zulip Isaac Vetter (Jun 08 2019 at 04:47):

Hey Josh!

view this post on Zulip Isaac Vetter (Jun 08 2019 at 04:48):

Two points:

view this post on Zulip Isaac Vetter (Jun 08 2019 at 04:48):

1) There are probably thousands of "activities" in an EHR. Frankly, I don't think that we know (or at least I don't) which of these would be valuable as the target of a navigation request. I imagine that it will be highly dependent upon the goals and relevant workflow of the particular app. I'd suggest that the best source of knowledge for workflow requirements is provider organizations, not app developers, not EHR developers. (E.g. Epic uses developer immersion (sorry, couldn't find better public explanation) in health systems to understand provider workflow). I'd suggest starting with a sophisticated app developer and committed health system for a popular use-case.

view this post on Zulip Isaac Vetter (Jun 08 2019 at 04:48):

2)

feels like unnecessary duplication to me, given that the CDS Hooks catalog is open-ended (i.e., one person's "fluff" is another's substance).

view this post on Zulip Isaac Vetter (Jun 08 2019 at 04:48):

I hate unnecessary duplication too!, and definitely do not want to block any person’s substance. I don’t understand what substantive use-cases are being proposed. Are you sure that navigation to view-type activities is what’s needed? Are you certain that navigation to “action”-type activities (aligning perfectly with our hooks) wouldn’t better satisfy clinical workflow?

view this post on Zulip Isaac Vetter (Jun 08 2019 at 04:48):

:idea: As a middle ground — our hooks are currently constructed as noun-action. Could we repurpose noun as activity?

view this post on Zulip Isaac Vetter (Jun 08 2019 at 04:48):

p.s. So, HL7 has an EHR workgroup, created by the ONC in the oughts. I suspect that our notion of “activity” should be part of this WG’s researched and published “functional model”. I got lost in pdfs upon downloading and reading the HL7 Electronic Health Record-System (EHR-S) Functional Model (FM), Release 1 and so don’t have anything useful to say here … :(

view this post on Zulip Josh Mandel (Jun 08 2019 at 14:43):

"lost in PDFs" is precisely where I don't want to be :)

view this post on Zulip Josh Mandel (Jun 08 2019 at 14:48):

"view" is one kind of action, and clearly a relevant one from the navigation perspective. Ordering actions like "order-select" are also pretty natural targets. I don't want to start a project to model all EHR activities; I just want a clear place for Web Messaging navigation events to plug into the ecosystem.

view this post on Zulip Josh Mandel (Jun 08 2019 at 14:48):

My current plan to include a navigtion-targets.md file in the Web Messaging spec and list targets ad-hoc, but I'll follow the CDS Hooks naming convention and probably include values that already are defined in CDS Hooks, just to preserve the possibility that we can reconcile.


Last updated: Apr 12 2022 at 19:14 UTC