Alschuler.spinosa copyright 2003, Liora Alschuler HL7::CDISC::RIM::CDA:: DMIM:: ODM:: SDS:: The World Liora Alschuler CDISC Interchange Bethesda September.

alschuler.spinosa

c op

y r igh

t 20

03

, Li o

ra A

l sc hu

ler

HL7::CDISC::RIM::CDA:: DMIM:: ODM:: SDS::

The World

HL7::CDISC::RIM::CDA:: DMIM:: ODM:: SDS::

The World

Liora AlschulerCDISC Interchange

BethesdaSeptember 30, 2003

2

alschuler.spinosa

c op

y r igh

t 20

03

, Li o

ra A

l sc hu

ler

• Liora Alschuler• Current

– alschuler.spinosa, consultants– Co-chair HL7 Structured Documents TC & Marketing

Committee– Co-editor, Clinical Document Architecture– [email protected]

• Background– Helped bring XML to HL7 in 1997– Wrote “ABCD… SGML: A Manager’s Guide to Structured

Information” in 1995– Electronic text: 1986 -- 1998

3

alschuler.spinosa

c op

y r igh

t 20

03

, Li o

ra A

l sc hu

ler

The Universal “need a hub”diagram

Protocol

SDS

CDA

aECG

Lab modelODM

RIM DMIM

Janus

4

alschuler.spinosa

c op

y r igh

t 20

03

, Li o

ra A

l sc hu

ler

SGML & the5 New Yorker DTDS:

1. Am. Assoc. of Publishers

2. Content-specific

3. For print output

4. For hypermedia

5. For database population

5

alschuler.spinosa

c op

y r igh

t 20

03

, Li o

ra A

l sc hu

ler

SGML & Conclusions:

“This session of the conference helped attendees to see the relationship between the design of an SGML document structure and the use or uses projected for the information.”

Example:

Db: content-specific naming

Hypermedia: short, generic names

“…when an organization is designing an SGML information database it must relate its goals for the use and reuse of its information to the new and old product applications it plans.”

6

alschuler.spinosa

c op

y r igh

t 20

03

, Li o

ra A

l sc hu

ler

Mapping XML

• XML is extremely munge-able• We can extract data, transform it,

convert it

someXML

otherXML

newXMLoldXML

yourXML

myXML

7

alschuler.spinosa

c op

y r igh

t 20

03

, Li o

ra A

l sc hu

ler

8

alschuler.spinosa

c op

y r igh

t 20

03

, Li o

ra A

l sc hu

ler

Use of ODM & CDA in Single-source

ODM

Initiate document

Complete study-required data entry

Complete data entry for patient chart

Finalize, sign and archive

Clinical trial management

Clinical document repository (EHR)

CDA

• Why not ODM2CDA?– CDA

• Looser• General• Human

readable• Compreh

ensive

FDA

SDSinODM/define.XML

9

alschuler.spinosa

c op

y r igh

t 20

03

, Li o

ra A

l sc hu

lerAre transforms hand-crafted, or can they

be automated? Does it scale? How do you scope a standard? How many transforms are too many? And, ultimately, what does

that tell us about “semantic interoperability” (eg, doing useful work)?

10

alschuler.spinosa

c op

y r igh

t 20

03

, Li o

ra A

l sc hu

ler

Make the ODM into

a RIM-derived

schema?•Requirements:

•RMIM, HD, ITS, datatypes, vocab, tweak, iterate

•Won’t look like ODM

•Won’t look like CDA

11

alschuler.spinosa

c op

y r igh

t 20

03

, Li o

ra A

l sc hu

ler

Import CDA directly into Janus and

operational dbs?

•Requirements:

•Re-engineer existing schemas to RIM-derived or

•Transform CDA to existing-schema-compatible XML

12

alschuler.spinosa

c op

y r igh

t 20

03

, Li o

ra A

l sc hu

ler

Sanity check

• Re-engineer existing databases? not• Optimize XML for particular task? yup• Translate and extract data as required

at the instance level, XML2XML• Ensure that cross-walk between

schemas feasible• Ensure that essential information

captured in all models such that transformation is possible

13

alschuler.spinosa

c op

y r igh

t 20

03

, Li o

ra A

l sc hu

ler

Recommendation• View HL7 RIM-defined data as canonical

source, as captured at source• Design CDISC exchange schemas for

business requirements (operations, reporting, analysis, validation, …), just as CDISC has already done/is doing

• Extract or transform– from HL7 RIM-derived XML-encoded source

(CDA, lab, aECG…) – to CRO/CDISC/FDA-encoded target XML for

insertion into appropriate schema

14

alschuler.spinosa

c op

y r igh

t 20

03

, Li o

ra A

l sc hu

ler

15

alschuler.spinosa

c op

y r igh

t 20

03

, Li o

ra A

l sc hu

ler

Advantages• Current db designs retain value• Source data available in richly encoded,

complete, canonical form• Exchange schemas standardized for

interoperability• Submission and operational schemas

optimized for dataprocessing, analysis • Maintain link back to comprehensive

original documentation: original will always be more comprehensive than research schema

16

alschuler.spinosa

c op

y r igh

t 20

03

, Li o

ra A

l sc hu

ler

Disadvantages

• Translation required between clinical, operational, submissions XML

• Today, these translations, even with standards, must be designed on a case-by-case basis

17

alschuler.spinosa

c op

y r igh

t 20

03

, Li o

ra A

l sc hu

ler

How to automate translation?

HL7XMLSubmissionsXML

Operations XML

18

alschuler.spinosa

c op

y r igh

t 20

03

, Li o

ra A

l sc hu

ler

How to automate translation?

HL7XMLOperations XML

SubmissionsXML

• eProtocol dictates:– Identifiers– Vocabulary– Extensibility– Constraints on

source XML– Constraints on

target XML

19

alschuler.spinosa

c op

y r igh

t 20

03

, Li o

ra A

l sc hu

ler

The Universal “need a hub”diagram

Protocol

SDS

CDA

aECG

Lab modelODM

RIM DMIM

Janus

20

alschuler.spinosa

c op

y r igh

t 20

03

, Li o

ra A

l sc hu

ler

Clinical documents are the core of healthcare information

Documenting patient care

Shared care

Clinical trials

Public health

Orders

Quality improvement

Reimbursement

Diagnostic imaging

Pathway, guidelines

Decision support

Relationships:

• Equal

• Derive

• Incorporate

• Link

• Transform

21

alschuler.spinosa

c op

y r igh

t 20

03

, Li o

ra A

l sc hu

ler

Clinical documents are the core of healthcare and clinical trial

information

Documenting patient care

Operations

Submissions

Orders

Quality improvement

Reimbursement

Analysis

But different forms of XML will be required for different applications

22

alschuler.spinosa

c op

y r igh

t 20

03

, Li o

ra A

l sc hu

ler

Conclusions• If every data schema were modeled on

the RIM, all the data could be HL7 derived (lab, CDA, aECG…)

• All schemas are not modeled on the RIM• Therefore,

– Render unto the HL7 RIM that which originates in the clinical environment

– Render unto CDISC-optimized XML that which needs to travel to operational and submissions databases in the clinical research environment

– Keep the link between the two– Automate transformation through standard

protocol, reconciled with the RIM

23

alschuler.spinosa

c op

y r igh

t 20

03

, Li o

ra A

l sc hu

ler

Dum, dum, da DMIM?To DMIM or not to DMIM?

• From RIM to XML:– RMIM– Hierarchical

descriptor– XML

• Design of XML ITS• Tooling output• Tweak• Bulk of CDA XML

hand-crafted datatypes and “narrative block”

• Development of RMIM– First balloted in 2000

(3-4 cycles)– Reconciliation– Three years of

implementation (~12 countries, 20+ sites)

– Incremented through user requirements

– Second ballot 2003-4

• Struc doc DMIM?– Requires reconciliation

w/ HL7 TCs– Requires reconciliation

with “CMETs”– Requires refinement of

full RIM– Requires encyclopedic

understanding of expression of clinical information

– Still in our future

• CDA Implementations– EU framework (12+

countries) “PICNIC”– NHII: Belgium, Greece,

Ireland, Denmark, Finland, Germany

– National claims processing: Canada, US (prop)

– Referrals: US, Japan, New Zealand, Australia, Germany

– Research: Germany, Canada, US (VA, KP+)

24

alschuler.spinosa

c op

y r igh

t 20

03

, Li o

ra A

l sc hu

ler

Thank you

Liora [email protected]/785-2623East Thetford, VT

HL7.org, Structured Documents TCHL7IHEjointdemo.org, HIMSS demo