1. A Summary Care Record Architecture based on HL7 CDA and SOA
Mr. Peter Given
Computing and Mathematics Department
Institute of Technology, Tralee

2. Introduction This paper shows how
Service Oriented Architectures can be combined with
the HL7 Clinical Document Architecture to facilitate
the sharing of Summary Care Records between health care information systems

3. Goals The solution demonstrates that
A federated architecture based on SOA and coding standards is a viable alternative
to a centralized record architecture for the sharing of summary care records

4. System Architecture
The design describes a system that facilitates the sharing of documents from the electronic medical record between GPs and other healthcare providers (Out of hours, emergency etc)
The architecture facilitates storing healthcare data locally and querying the data via an indexing mechanism.

6. Data Layer
The Data layer involves the capture and storage of patient information in a GP Information System
An Open source GP Information System was chosen to represent legacy open source GP Systems (openEMR)

7. OpenEMR
A free open source medical practice management and electronic medical records application (
Licensed under GNU GPL
Operates in Linux, FreeBSD, MacOS X and Windows
A client/server architecture written in PHP with a mySQL database – downloadable as a WAMP application

8. OpenEMR
The system is being used successfully at a number of sites [86,87,89].
A number of commercial companies are also supporting the software [88,89,90].
Not based on any EHR Standards but has some support for coding standards

9. Functionality of OpenEMR
The application stores
patient demographic data
patient encounter information
vital signs
laboratory results
prescriptions and allergy information.
The system also allows the management of appointments and the generation of reports.

10. OpenEMR

14. Interfacing to OpenEMR
The project entailed building an Interface Engine was developed in order to extract:
A Summary Care Record (SCR).
This will hold patient demographic data along with known allergies and any medications prescribed in the last 6 months.
A Prescription Record (PR).
This will contain demographic details and prescription information that can be forwarded to a pharmacy system.
A Patient Index for the SCR

15. Interfacing to OpenEMR
Index
SCR

16. Interfacing to OpenEMR
The HL7 Clinical Document Architecture was chosen as the document architecture standard.
The Interface Engine retrieves data from openEMRs databases and construct CDA conformant documents using the CDA schema

17. HL7 CDA
“The HL7 Clinical Document Architecture (CDA) is a document markup standard that specifies the structure and semantics of "clinical documents" for the purpose of exchange” (HL7 website)
A CDA document can include
Text
Images
Sounds
other multimedia content.
Document-level, section-level and entry-level templates available
The content of CDAs defined in the HL7 Reference Information Model and HL7 Data Types

18. CDA Structure <ClinicalDocument> ... CDA Header ...
<StructuredBody>
<section>
<text>...</text>
<Observation>...</Observation>
<Observation>
<reference>
<ExternalObservation>...
</ExternalObservation>
</reference>
</Observation> <
/section>
<section>...</section>
</section>
</StructuredBody>
</ClinicalDocument>

19. Summary Care Record as a CDA document
The Summary Care Record
Header
Title, Time, Confidentiality Code, Version Number, Record Target (PatientRole, Patient Name, Gender, Birth Date), Provider Organization, Author, Custodian, Legal Authenticator, Encounter Information
Section
Allergies and Adverse Reactions
Medications

20. Prescription Record as a CDA document
The Prescription Record
Header
Title, Time, Confidentiality Code, Version Number, Record Target, Prescription Provider Organization, Author, Custodian, Legal Authenticator, Encounter,
Section
Prescription Information
drug name, quantity, dosage, refills and notes

21. Interfacing to OpenEMR
Index
SCR

22. Implementing the interface
The Interface Engine was coded using PHP/DOM and using a CDA document/section template schema
The mapping from the openEMR database to the CDA schemas was incomplete as the database did not hold all the data specified in the CDA document schema.
For example openEMR currently does not store clinical codes with separate parts of the medical record so the code fields in the CDA documents were left empty

23. Mapping inconsistencies
Also the openEMR system stores dosing intervals as
qid, bid, q3h
while the CDA document needs these intervals to be represented as time periods and unit (e.g. 3 hours).

25. Designing the Communication Layer
The Communication layer was designed in two parts.
MPI Subsystem :The first part is the design of a distributed Master Patient Index
Record Retrieval Subsystem : The second part of the design allows the user to retrieve the Summary Care Record or prescription record using a Service Oriented Architecture.

26. Master Patient Index Subsystem
The Master Patient Index was designed on top of a peer-to-peer network called the GP Peer Group.
A GP with the correct credentials can join and leave the peer network.
The GP can then publish patient indices to the network.
The index advertises the existence of a Summary Care Record for a patient and the endpoint address for that record

28. Patient Index
The patient index is designed to hold the following information:
document type etc
The patients’ surname and first name
The patient’s date of birth
The patient’s PPS number
The organisations’ id and name
The endpoint where the full Summary Care Record can be found.
The information to build the index was extracted from the openEMR system by the Interface Engine.

29. Generating Master Patient Index
SCR

30. Generating a patient index from OpenEMR
The index information is initially retrieved from the openEMR Clinical Information System using the data layer interface as described above
The index information is wrapped in an XML message. A typical message is shown below:

31. Example Mater Patient Index
<?xml version="1.0" ?>
<SCR_Advert>
<Document_CDA_Type> </Document_CDA_Type>
<Document_Name>Summary Care Record</Document_Name>
<Document_Description>Clinical Document</Document_Description>
<Patient_Surname>Bloggs</Patient_Surname>
<Patient_Firstname>Joe</Patient_Firstname>
<Patient_DOB> </Patient_DOB>
<Patient_PPS>1</Patient_PPS>
<Organisation_ID>6</Organisation_ID>
<Organisation_Name>givens sugery</Organisation_Name>
<Endpoint>
</SCR_Advert>

32. Peer Advertisement
The peer-to-peer software uses this information to generate a peer advertisement which appends the following header
<?xml version="1.0" ?>
<!DOCTYPE SCRAdvertisement>
<SCRAdvertisement xmlns:jxta=
<Advert_ID>urn:jxta:uuid E C475C1C </Avert_ID>
MPI Infromation
</SCRAdvertisement>

33. Peer to peer network
A node can both publish to and search this network for indices.
To search for a particular patients SCR a discovery message is propagated to all peers in the network.

34. Searching for a Patient Index
The metadata contained within the index will improve the search results from the discovery process.
Once a peer has retrieved the index for a patient the endpoint information is retrieved from the index.
The second part of the communication layer will now use this endpoint to retrieve the actual Summary Care Record from the system in which it is stored.

35. Implementing MPI Application
The MPI application itself consists of Java Swing components that implement a Graphical User Interface, along with calls to the JXTA API to join the GP Peer Group network and to publish/discover advertisements on the network.

37. Searching for a patient summary care records
Once the user provides values the peer sends a discovery message to the other peers on the sub-network.
Any Rendezvous peers on the local sub-network will further propagate the discovery message to other sub-networks they know about.
Any advertisements that match the surname/PPS number are returned to the requesting peer.
Clicking on a particular patient from this list will show the endpoint where the full Summary Care Record can be found.
This endpoint information is now presented to the Web Service client to retrieve the patient’s full Summary Care Record

38. Security in the Peer Network
The JXTA configuration requires a username and password. A certificate may also be required.
Once the peer starts up it can issue discovery messages to locate other peer groups it may wish to join.
The peer groups enforce a second level of security. The user must supply a login name and password to join the GP Peer Group.

39. Record Retrieval Subsystem
SOA is used to deliver the Summary Care Record
A client program that knows the endpoint address can send a message to a Web Service on the GP site to retrieve the Summary Care Record as a CDA document

41. SOA Solution
A Web Services Server called Cape Clear was used to build the SOA
The Web Service was defined in WSDL using Cape Clear Studio.
The Web Service was implemented in Java and uses the SOAP protocol at the messaging layer.
A Web Services client was built in Java
It generates SOAP requests which are sent to the sever and the response contains either a Summary Care Record or a Prescription Record

42. Retrieving the Summary Care Record
The Web Service had two services; getSumaryCareRecord and getPrescriptionRecord.
Both services require the patient’s PPS number (a java.String) and the patient’s surname (java.String).
The service returns a CDA documents (also as a java.String) if one exists

44. Integrating the CDA record
A reverse process to that taken to generate the data at the data layer could be followed in order to integrate the data with a local system.
For this project it was enough to be able to view the record

45. Viewing the Record
Once the record is saved the user can finally view the file.
A standard XSL transform is attached to the file.
The transform creates an HTML document that can be viewed by any browser.
The document can also be printed out.

48. Application areas Out of Hours Services – telephone triage etc
Emergency Services
Community Nursing

49. Conclusions: Advantages
GP buy in
can maintain their systems of choice (open source or commercial)
GPs can maintain their records locally (pull model)
Healthcare Standardization
CDA for clinical document exchange
Other summaries can be implemented easily
Technology Standardizations
SOA and Web Services
Distributed MPI with location information

50. Conclusions: Challenges
Technical barriers
Each GP system requires an interface Engine to generate CDA documents
GP database may not hold required information
GP Communications System is a hybrid SOA/P2P solution
Security
Cost
Distributed MPI with location information

51. Thank You!
Questions?