Remote Decision Support with CDS Hooks - FHIR DevDays 2017

CDS Hooks is a new emerging standard that leverages both FHIR and SMART to provide open and interoperable decision support in the EHR. Using CDS Hooks, you can provide guidance to the clinician such as a better medication to order, critical information about the patient, or push a SMART app that should be run. This presentation will provide you with an overview of CDS Hooks, how it aligns with existing HL7 efforts and standards, and what the future holds.

This deck was presented at FHIR DevDays 2017 (https://www.fhirdevdays.com) on November 16, 2017.

Transcript

1. Remote Decision Support with Kevin Shekleton FHIR DevDays 2017 November 16, 2017
3. The problem with SMART apps Bilirubin Chart Newborns < 120 hours old Bilirubin results that are either: Not documented Outside of the accepted range The user needs to know the app is available The user needs to know the app is relevant The user has to find the app and launch it
5. CDS Hooks A vendor agnostic remote decision support specification Created by the team behind SMART Open source (Apache) Emerging standard with active development and widespread participation from stakeholders
6. CDS Service A service that is: invoked by the EHR via a hook, evaluates its own logic using FHIR data, returns decision support via cards
7. Example Hooks patient-view When a patient’s chart is opened medication-prescribe When a medication is selected for prescription order-review Viewing pending orders for signing
8. CDS Services EHR Med Order Toprol XL 50 mg daily
   EHR triggers a CDS hook and invokes a remote service 1 2 CDS Service executes its own rules, leveraging FHIR data as needed EHR FHIR Server 3 Returns CDS cards (rendered and displayed by EHR) $200 per month (patient pays $30) Try HCTZ as first-line Switch to HCTZ Managing hypertension? Launch JNC 8 Rx Pro information card suggestion card smart app link card
9. POST https://example.com/cds-services/example-service EHR CDS Service { "hook": "patient-view”, "fhirServer": "https://fhir.example.com", "user": "Practitioner/789", "patient": "123", "encounter": "456", ... }
10. EHR CDS Service POST https://example.com/cds-services/example-service CDS Service Logic
11. POST https://example.com/cds-services/example-service EHR CDS Service HTTP 1.1/ 200 OK { "cards": [ { "summary": "Example card", "indicator": "info", "source": { "name": "Demo CDS Service” }}]}
12. Cards • A CDS Service can return any number of cards • The EHR renders each card as it sees fit • Each card must have: • A concise summary (140 characters) • An indicator noting the importance of the card • Information on the organization or data set that is the source of the card’s data
13. CDS Service Response JSON { "cards": [ { "summary": "Example card", "indicator": "info", "source": { "name": "Demo CDS Service” } } ]}
14. Card Indicator Informational Hard Stop Warning
15. Common Card Examples Suggestions Proposed actions encoded as FHIR resources App Links Proposed SMART app that should be used Information Only Textual information for the provider
16. Cards can contain any combination of information, suggestions, and links
17. Capturing user behavior for analytics • Since CDS Services return purely JSON, they don’t know how users interact with their suggestion • Each suggestion may contain a UUID • The UUID allows the EHR to notify the CDS Service that the user interacted with their suggestion POST {cds-service}/analytics/{uuid}
18. CDS Services CDS Service Provider Base https://example.com Discovery /cds-services Service /cds-services/{id}
19. Discovery { "services": [ { "hook": "patient-view", "name": "CDS Service Example", "description": "An example CDS service", "id": "example-service", "prefetch": { "patientToGreet": "Patient/{{Patient.id}}" } }, {...} } GET https://example.com/cds-services
20. CDS Service Invocation { "services": [ { "hook": "patient-view", "name": "CDS Service Example", "description": "An example CDS service", "id": "example-service", "prefetch": { "patientToGreet": "Patient/{{Patient.id}}" } }, {...} } POST https://example.com/cds-services/example-service
21. Security EHR provides FHIR access token to CDS Service CDS Service trusts the EHR via signed JWT EHR trusts the CDS Service via its TLS endpoint
22. Security > JWT { "iss": "https://ehr.example.com/", "aud": "https://example.org/cds-services/my-service", "exp": 1422568860, "iat": 1311280970, "jti": "ee22b021-e1b7-4611-ba5b-8eec6a33ac1e” } eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpc3MiOiJodHRwczovL2ZoaXItZWhyLmV4YW1wbGUuY29tLyIsI mF1ZCI6Imh0dHBzOi8vY2RzLmV4YW1wbGUub3JnL2Nkcy1zZXJ2aWNlcy9zb21lLXNlcnZpY2UiLCJleHAiOjE 0MjI1Njg4NjAsImlhdCI6MTMxMTI4MDk3MCwianRpIjoiZWUyMmIwMjEtZTFiNy00NjExLWJhNWItOGVlYzZ hMzNhYzFlIn0.Gwl3s301OMWpdEVAVj_T3JZV8bs7N5-V7QNNG7TQ33o
23. Security > JWT Field Value Example iss The URL of the issuer of the JWT https://ehr.example.com aud The CDS Service endpoint being called https://example.org/cds- services/my-service exp Time this JWT expires 1422568860 iat Time at which this JWT was issued 1311280970 jti Nonce to protect against replay attacks ee22b021-e1b7-4611-ba5b- 8eec6a33ac1e
24. sandbox.cds-hooks.org
25. Separation of Concerns Concern Owner Workflow & hooks (triggers) EHR CDS logic CDS Service Provider Initial context data EHR Additional data CDS Service Provider (via FHIR or other sources) CDS presentation EHR
27. Argonauts ❤ CDS Hooks Pushing SMART apps via a CDS Service upon opening the patient’s chart Validate and implement the CDS Hooks security model
28. Roadmap Release a 1.0 specification Q1 2018 Launch production pilots 2018
29. What’s after 1.0? Performance: trigger guards, SLAs Event & time based hooks Specific use cases (ACR ordering guidelines, CDC opioid guidelines, etc)
30. Wait, doesn’t FHIR already define an approach for CDS?
31. Let’s compare! Clinical Reasoning CDS Hooks Management Rule definition  Local execution  Remote execution  
32. What the?! More  wins, right?
33. broad vs narrow
34. Local Decision Support Remote Decision Support Clinical Reasoning
35. Let me predict your questions… Q: Is CDS Hooks in production? A: Not yet :'-( Q: Will CDS Hooks support <insert feature>? A: Log a Github issue or let’s discuss over our Google Group or Zulip chat!
36. Let me predict your questions… Q: What is the planned support for <insert your favorite EHR, hospital, etc>? A: ¯\_()_/¯ Q: What is the business model for CDS Hooks? A: ¯\_()_/¯
37. References Specification & documentation http://cds-hooks.org Sandbox http://sandbox.cds-hooks.org Source code https://github.com/cds-hooks
38. Let’s code! Complete the tutorial and write your own CDS Service! github.com/cerner/cds-services-tutorial
39. Thanks!
40. Addendum Sample content from CDS vendors developing on CDS Hooks
42. PerfectChoice™ Helping physicians make the perfect antibiotic choices. STEVENSON, LAURA Age: 4 years CrCl: >100mL/min Allergies: No Known Allergies SCr: 0.9gm/dL Previous Admit: 01/18/13 NEW CULTURE INFORMATION 12/22/14 08:45 Specimen ID: 00564783 Status: Prelim Culture: Blood Result: E.Coli ID SUMMARY Microbiology 012/20/14 Urine Culture: E.coli Antibiotic Medications: 01/24/13 – 01/25/13 Amoxicillin-clavulanate 150mg PO q8h Flags: Community-acquired infection Isolation NHSN Documentation: None Change Orders Dismiss PerfectChoice – ID: New Organism Identified PerfectChoice OPTIONS Premier Memorial Hospital – Last 12 mos. E.Coli - Specimen Source Blood Antibiotic Medications (% Susceptible) Ampicillin-Sulbactam (93%)* Levofloxacin (88%)* Cefazolin (79%)* SMX-TMP (76%)* Gentamicin (73%)* IDENTIFY. DECIDE. ACT. Access the EMR to make the appropriate medication order decisions. STEVENSON, LAURA Age: 4 years CrCl: >100mL/min Allergies: No Known Allergies SCr: 0.9gm/dL Previous Admit: 01/18/13 NEW CULTURE INFORMATION 12/22/14 08:45 Specimen ID: 00564783 Status: Prelim Culture: Blood Result: E.Coli ID SUMMARY Microbiology 012/20/14 Urine Culture: E.coli Antibiotic Medications: 01/24/13 – 01/25/13 Amoxicillin-clavulanate 150mg PO q8h Flags: Community-acquired infection Isolation NHSN Documentation: None Change Orders Dismiss PerfectChoice – ID: New Organism Identified PerfectChoice OPTIONS Premier Memorial Hospital – Last 12 mos. E.Coli - Specimen Source Blood Antibiotic Medications (% Susceptible) Ampicillin-Sulbactam (93%)* Levofloxacin (88%)* Cefazolin (79%)* SMX-TMP (76%)* Gentamicin (73%)* PerfectChoice Notification: Based on new culture information and facility antibiogram, the following antinfectives have the highest likelihood (% susceptible) of effectively treating the infection. Source: Launch Premier TheraDoc for more details. A Premier Solution View the best antibiotic options based on facility antibiogram in context of the patient’s relevant clinical data to reach the right decision. Know the priority patients with new microbiology results with automated surveillance that notifies clinicians in their workflow.
43. Medication Management for Adherence (CDS Hooks) Real-time medication adherence insights delivered directly into workflow during patient visits Bi-directional communication to enable users to provide real-time feedback
44. Real time, workflow integrated, patient specific, evidence based Reduces low-value and unnecessary care Stanson Health’s CDS Hooks service
46. CDS Hooks Dose Calculator Prevent ADEs with system-calculated, safe, patient-specific doses • See common orders for this patient based on their age and the ordered drug • Smart logic knows this order should be dosed as trimethoprim • Select from safe rounded admin amounts Leverages FDB Cloud Connector web services Calculations happen in real-time using current knowledge base Med ordering workflow: pediatric patient, sulfamethoxazole/trimethoprim
47. (CDS Hooks)