Working Group Meeting, Orlando, fl




Дата канвертавання24.04.2016
Памер59.25 Kb.
Clinical Decision Support TC

Arden Syntax SIG

GLIF SIG

Meeting Minutes

Working Group Meeting, Orlando, FL


January 9 - 12, 2001


January 9, 2001 0900-1700: Educational Day

Attendance: Mike Jones (Micromedex; mike.jones@mdx.com)

Pat Lyons (SMS;pat.lyons@smed.com),

Matt Sailors (UT-Houston; matthew.sailors@uth.tmc.edu),

Peter Haug (LDSHospital; LDPHaug@ihc.com),

Rick Schiffman (Yale Ctr for Medical Informatics;richard.shiffman@yale.edu),

Bob Greenes (Harvard U; greenes@harvard.edu),

Aziz Boxwala (Harvard U; aziz@dsg.harvard.edu),

Ying-Chou Sun (Harvard U; ysun@dsg.harvard.edu),

Mor Peleg (Stanford U; peleg@smi.stanford.edu),

Carol Broverman (Fast Track Systems; cbroverman@fast-track.com),

Samson Tu (Stanford U; tu@smi.stanford.edu),

Ian Shepherd (Royal Pharm Society; ishepherd@rpsgb.org.uk),

Eric Connor (Pharmaceutical P; eric@rpsgb.org.uk),

Alex Ruggieri (Mayo Foundation; ruggieri.alexander@mayo.edu),

Tim Benson (Abies eHealth; tb@abies.co.uk),

Paul van Arragon (Mitra Imaging; arragon@mitra.com),

Sven Tiffe (Siemens; sven.tiffe@med.siemens.de),

John Dulcey (Applied Medical Informatics; john.dulcey@alum.mit.edu),

Rob Moe (EDS/VHA; bob.moe@eds.com),

Sharon Smart (SCHIN Newcastle University;s.r.s.smart@ncl.ac.uk),

Neill Jones (SCHIN Newcastle University; n.k.jones@ncl.ac.uk),

Ian Purves (SCHIN; ian.purves@ncl.ac.uk),

Robert Jenders (Columbia U; jenders@columbia.edu),

Anders Thurin,

David Rowed (HL7 Australia; drowed@bigpond.net.au).


1. Announcement: The TSC yesterday approved creation of an Arden SIG and the resultant changes in TC and SIG charter/mission statements.

2. Educational Presentations

a. Robert Jenders: Arden Syntax review [presentation].

b. Neill Jones: Procedural Guidlines (Sophie, PRODIGY II)



  1. Mor Peleg and Aziz Boxwala: GLIF. [presentation]

1. Hypothesis: Translating among guidelines implemented under different models

is difficult because the guideline models are incompatible.

2. Goals:

a. Interested in created a shareable model (important elements) of guideline models.

b. Creating joint tools that support the guidelines lifecycle: authoring, execution,

validation.

d. Ian Purves, Sharon Smart, Neill Jones and Samson Tu: PRODIGY[presentation1,

presentation2]

1. Facilitating sharing of guidelines by providing "standard services" (data queries, data

storage,

carrying out guideline-initiated acts, etc)

2. "Term mapper": Map internal PRODIGY concept to a standard terminology.

3. Guidelines mapped to a "virtual CPR".

e. Rick Shiffman: GEM [presentation]

1. Focus = guideline document, with rich set of markup tokens.

2. Possibly Arden as the non-declarative part of an implemented guideline.

3. HL7 Harmonization for Clinical Decision Support (Gunther Schadow) [presentation]

a. Brief overview of HL7 organization and RIM.

4. Initial Discussion of Tasks/Priorities

a. Topics: Bidirectional communication with CPR (queries, data entry, guideline position);

act messages; agent/entity interaction; criterion language; terminology mediation services;

guideline representation; library (use cases, etc); open-source software

b. Candidate high-priority tasks: criterion language; specification of derived observations

(promised in HL7 v2 but never fully completed).



January 10, 2001 0900-1230: Discussion of TC and SIG scope/mission

1. Dan Russler: Discussion of Scope between Clinical Decision Support and Patient Care TCs

[presentation]

a. Presented summary of tutorial regarding "acts" from USAM.

1. Each act has a mood (HL7 term table): goal, event, order, etc. A result can be

changed to a request for the result by changing the mood.

2. "Derivation expression": Statement of clinical logic, e.g., MLM, still incompletely

modeled.

b. Proposed Patient Care scope: point-of-care issues, diagnostic observations, clinical lists,

workflow messages.

c. Proposed Clinical Decision Support scope: automated decision support (including

guidelines,

clinical trials, protocols, etc); also messages (precondition/master file, compliance,

guideline

variance, alerts/reminders).

d. Proposed Orders/Observations scope: ancillary system observations and treatments.

2. Discussion of TC Work Products (tasks/priorities)

a. Bi-directional communication with the CPR: Queries (C/Q); Guideline position storage;

data entry services (request for decision services; unsolicited output [alert/reminder, etc,

to various destinations] = act messages [possible overlap with O/O TC]; KB update).

b. expression language (including criterion language):

1. examples of things accomplished using a criterion language: eligiblity determination;

triggers.

2. May have some overlap with C/Q TC.

c. guideline representation (common model)

d. library of use cases and (computable) guidelines

e. demonstration and validation software tools: open-source software that performs part

of the CDSS process (e.g., syntax checker; authoring tool; demonstration of applications; etc).

1. concern from vendors regarding economic competition.

2. needs to be discussed by the TSC: prepare a position paper with benefits (increase use of standards) and risks (precedent-setting for HL7).

3. HL7: provide links, access to software but not necessarily sponsor the software

f. outreach activities: meetings, tutorials

g. TC approved additions to Mission/Charter Statement

1. Action item: Synthesize additions with current list and circulate to TC (Jenders).

2. Action item: Present Statement to ARB and TSC for approval by spring meeting

(Jenders).



January 10, 2001 1330 - 1700: GLIF SIG

Attendance: Mike Jones (Micromedex; mike.jones@mdx.com), Pat Lyons (SMS;

pat.lyons@smed.com), Matt Sailors (UT-Houston; matthew.sailors@uth.tmc.edu), Peter Haug (LDS

Hospital; LDPHaug@ihc.com), Rick Schiffman (Yale Ctr for Medical Informatics;

richard.shiffman@yale.edu), Bob Greenes (Harvard U; greenes@harvard.edu), Aziz Boxwala

(Harvard U; aziz@dsg.harvard.edu), Mor Peleg (Stanford U; peleg@smi.stanford.edu), Carol

Broverman (Fast Track Systems; cbroverman@fast-track.com), Samson Tu (Stanford U;

tu@smi.stanford.edu), Ian Shepherd (Royal Pharm Society; ishepherd@rpsgb.org.uk), Sven Tiffe

(Siemens; sven.tiffe@med.siemens.de), John Dulcey (Applied Medical Informatics;

john.dulcey@alum.mit.edu), Ian Purves (ian.purves@ncl.ac.uk), Robert Jenders (Columbia U;

jenders@columbia.edu), David Rowed (HL7 Australia; drowed@bigpond.net.au).
1. Reviewed proposed goals/objectives [Greenes presentation].

a. Overall goal: facilitate availability and use of high-quality guidelines in clinical practice.

b. Specific goals: define common representations, etc.

c. Possible early tasks: guideline retrieval based on patient eligibility criteria; expression

language.

d. Other early tasks: identify apps by vendors to identify use cases, functional requirements and sample guidelines; engagement of vendor community; establishment of distribution lists and workgroups; possibly another guideline workshop.

e. Guideline workshop

1. Decision: Defer workshop for 2 HL7 meetings before deciding to arrange this.

2. HL7 registration fee may be barrier to holding this as part of working group meeting.

3. Probably insufficient time available to do this at HIMSS or AMIA meetings.

2. Use of shared guideline model

a. Advantage: incorporate these features into local guideline to enable larger audience.

b. Advantage: be able to use guidelines authored elsewhere

c. Is neither a subset nor a superset of any particular guideline formalism.

d. Discussion regarding utility of shared model. Helps to identify services required by CDSS

from other parts of the information system to enable platform that could accommodate

many models.

3. Discussion of GLIF as product of SIG

a. Goal of Intermed is to transfer GLIF to HL7.

b. Need to determine development and release schedule for HL7 GLIF specification.

c. Want to maintain GLIF name because of recognition that it has achieved.

4. Expression language: Provide consistent mechanism for delineation of conditions ("if"

statements).

a. Ideally Arden, GLIF and other models would share a common expression language.

b. Can work on this in parallel with shared guideline model.

5. Discussion of name of SIG

a. GLIF is not an interchange language (for translating into/between executable

guidelines) but a common, shared guideline format.

b. Some have expressed concerns that proponents of other models would be put off by

perceived emphasis on this particular work product originating in another group.

c. Possibile Names

1. Consensus of options: GLIF, Common GLIF, Clinical Guidelines.

2. Decision: Clinical Guidelines SIG [7 votes], Common GLIF [3 votes], GLIF [0 votes].

January 11, 2001 0900 - 1230: Arden Syntax SIG

Attendance: Mike Jones (Micromedex; mike.jones@mdx.com),

Pat Lyons (SMS; pat.lyons@smed.com),

Mor Peleg (Stanford U; peleg@smi.stanford.edu),

Rick Schiffman (Yale Ctr for Medical Informatics; richard.shiffman@yale.edu),

Joyce Sager (HealthVision),

Aziz Boxwala

(Harvard U; aziz@dsg.harvard.edu),

Sven Tiffe (Siemens; sven.tiffe@med.siemens.de),

John Dulcey (Applied Medical Informatics; john.dulcey@alum.mit.edu),

Samson Tu (Stanford U;

tu@smi.stanford.edu),

Neill Jones (SCHIN Newcastle University; n.k.jones@ncl.ac.uk),

Matt Sailors(UT-Houston; matthew.sailors@uth.tmc.edu),

Robert Jenders (Columbia U; jenders@columbia.edu),

Bob Greenes (Harvard U; greenes@harvard.edu).


1. Elections: Elections for Co-Chair of the Arden SIG will be held at the next meeting in Minneapolis.

Persons interested in the Co-Chair position should make this interest known to the HL7 Staff in order for their name to be placed on the ballot. Although there is no provision in the HL7 charter, it was informally agreed that one of the Arden Co-Chairs would serve as one of the Co-Chairs of the Clinical Decision Support TC.

a. Decision(unanimous): The SIG will have 2 co-chairs, one of which will be a co-chair of the TC.

b. Action Item: Robert Jenders will notify HL7 regarding need for balloting for 2 co-chairs

at the 4/2001 meeting.

2. SIG Mission/Charter Statement:

a. Robert Jenders presented the mission statement of the Arden SIG for discussion.

The SIG edited the document..

b. Action Item: Robert Jenders will present the revised mission statement to the

Architectural Review Board and the Technical Steering Committee.

3. Publicity: Robert Jenders reviewed recent past and possible upcoming outreach efforts.

a. Action Item: Develop standard, high-level presentation reviewing importance of

clinical decision support and standardization efforts (e.g., Arden Syntax) (Jenders).

b. Action Item: Submit the names of meetings at which delivery of a presentation of clinical decision support would be desirable (all SIG members).

4. Completion of Arden v 2.1: Robert Jenders presented the Arden Roadmap developed by Peter

Haug. The purpose of the road map, proposed at the 9/2001 meeting, is to describe future activities planned for the Arden Syntax for inclusion in the version 2.1 document.

a. Action Item: John Dulcey will create a section for Arden 2.1 that lists the changes (new

operators and the structured write statement) from v 2.0.

b. Action Item: Once the structured write statement is finalized, Matt Sailors will

update the Bakus-Nauer Form (BNF) of the Arden 2.1.

5. Arden v3 development: Fuzzy Logic: Sven Tiffe presented his work on Fuzzy Logic.

[presentation] This included a description of fuzzy vs crisp sets and incorporation of membership

functions and their output in Arden Syntax. The proposal detailed overloading of operators to handle both fuzzy and classic Boolean values.

a. Action Item: Sven Tiffe is working on an Arden compiler to which he will add the

proposed fuzzy operators.

b. Action Item: Sven Tiffe, Mike Jones and Matt Sailors will develop use cases for

fuzzy operators. These will include examples of MLMs using the current Arden version and

proposed additions.

6. Arden v3 Development: Harmonization of Arden Syntax Data Model and the HL7 RIM

[Peleg and Boxwala presentation]

a. Need: Arden data elements have only 2 parts (value and primary time), but RIM

objects have many additional attributes (time interval, units, etc).

b. Two suggested solutions:

1. Allow "dot" notation for Arden variables to represent additional attributes.

2. Object-oriented: Provide for inclusion of standard methods to access

attributes without using "dot" notation.

c. Preliminary advice to GLIF developers wanting to proceed with development that

will be consistent with Arden and the RIM: "dot" notation [7 votes], object-oriented [4

votes].

d. Action Item: Mor Peleg and Aziz Boxwala will circulate examples of implementing



both approaches and elicit comments.

  1. Sager suggestion: Can implement this as parallel lists.



January 11, 2001 1330 - 1700: Joint Meetings with Control/Query and Structured Documents TC
Attendance: Mike Jones (Micromedex; mike.jones@mdx.com), Pat Lyons (SMS;

pat.lyons@smed.com), Mor Peleg (Stanford U; peleg@smi.stanford.edu), Rick Schiffman (Yale Ctr

for Medical Informatics; richard.shiffman@yale.edu), Joyce Sager (HealthVision), Aziz Boxwala

(Harvard U; aziz@dsg.harvard.edu), Sven Tiffe (Siemens; sven.tiffe@med.siemens.de), John Dulcey

(Applied Medical Informatics; john.dulcey@alum.mit.edu), Samson Tu (Stanford U;

tu@smi.stanford.edu), Neill Jones (SCHIN Newcastle University; n.k.jones@ncl.ac.uk), Matt Sailors

(UT-Houston; matthew.sailors@uth.tmc.edu), Robert Jenders (Columbia U; jenders@columbia.edu),

Ian Purves (ian.purves@ncl.ac.uk).



Control/Query [led by Mark Tucker (Regenstrief Institute), TC member]

1. Clinical Decision Support TC members described the need for standard queries.

a. Arden Syntax curly braces: data model, query language and vocabulary as important

components.

b. GLIF: Ability to populate complex USAM objects.

2. Possible query languages: OQL, SQL.

a. However, C/Q does not endorse any particular query language.

b. C/Q has developed standard messages for mediating requests for data but not a query

language.

3. Extensive debate ensued regarding feasibility of using RIM as the standard data model because of need to translate between the RIM and the actual data model at any particular institution.

a. C/Q members felt that this translation at run-time would guarantee poor performance.

b. Alternate solution: First, do a bulk normalization of database (logical EPR) and then do a run-time query translation.

c. Alternate solution: Compile-time translation of RIM-based query to a query against

an institution's actual database. (Preferred by members of the CDS TC.)

4. Potential tasks: Specify expression language (with query syntax) and "logical EPR"

(data model, related to RIM). Action Items deferred until 12 January session.


Structured Documents [led by Bob Dolin (Kaiser-Permanente), TC Co-Chair]
1. Presentation of Clinical Document Architecture (CDA) Level One [Dolin presentation].

a. Higher levels (Two and Three) = less coarse/more finely granular resolution of documents (e.g., progress notes, guidelines, etc).

b. CDA Level One: Header + (coarsely granular) Body.

2. In anticipation of 5/2001 meeting: Dolin and Shiffman will explore potential harmonization between GEM and the CDA.

3. Action Item: Ian Purves will coordinate work on this issue with Neill Jones and Bob Dolin before the 5/2001 meeting.
Clinical Decision Support TC: Planning for 5/2001 Meeting (May 7-11, 2001)
1. Decided tentative schedule:

Monday (May 7): Clinical Decision Support & Arden Syntax tutorial

Tuesday (May 8): Arden Syntax SIG (leftover v2.1 issues [structured write], fuzzy logic, XML).

Proposed Clinical Trials SIG (Linda Quade will initiate).

Wednesday (May 9): Clinical Decision Support TC (logical EPR, terminology [medication etc]

use cases).

Joint meeting with Patient Care TC (first quarter).

Joint meeting with Vocabulary TC (last quarter).

Thursday (May 10): Clinical Guidelines SIG (expression language, guideline representation)

Joint meeting with Structured Documents TC (last quarter).

Friday morning (May 11): Clinical Decision Support TC (guideline representation)

January 12, 2001 0900 - 1130: GLIF SIG

Attendance: Mike Jones (Micromedex; mike.jones@mdx.com), Pat Lyons (SMS;

pat.lyons@smed.com), Joyce Sager (HealthVision), Aziz Boxwala (Harvard U;

aziz@dsg.harvard.edu), Sven Tiffe (Siemens; sven.tiffe@med.siemens.de), John Dulcey (Applied

Medical Informatics; john.dulcey@alum.mit.edu), Samson Tu (Stanford U; tu@smi.stanford.edu), Neill

Jones (SCHIN Newcastle University; n.k.jones@ncl.ac.uk), Matt Sailors (UT-Houston;

matthew.sailors@uth.tmc.edu), Robert Jenders (Columbia U; jenders@columbia.edu), Ian Purves

(ian.purves@ncl.ac.uk), Carol Broverman (Fast Track Systems; cbroverman@fast-track.com).


1. Discussion of standard for human-readable guideline documents

a. Should include flowchart standard, specification of different levels of structure, definition

of attributes of specific kinds of steps and relationship between knowledge and

document views.

b. Action Item: Meet before next meeting to address these issues (Tu, Dulcey, Peter Johnson

[SCHIN, UK], Sharon Smart [SCHIN, UK], Jones).

c. Action Item: Apply these specification to user interface of current authoring tools (Lyons).

d. Action Item: Set up listserv list for this project.

2. Project: Enable guideline retrieval based on eligibility criteria (with limited/no execution).

a. Possible guideline collections: NGC, ACP, NCI cancer treatment protocol.

b. Action Item: Define the project, use cases (e.g., problem-list-based, triggered by results,

provider-initiated, data mining) and the corpora of guidelines before the

next meeting (N Jones, Broverman, possible non-HL7 members [John Gennari,

Lucila Ohno-Machado, Herbert Kaizer], Tu, Greenes).

c. Possible translation: Work on this project can produce additions to Arden Syntax. This

will require simultaneous work on a common expression language.

3. Action Item: Update SIG Mission Statement and submit to ARB and TSC in advance of next

meeting (Peter Johnson, Bob Greenes).

4. Action Item: Publicize projects and plans for next meeting to HL7 and vendor communities

(Peter Johnson, Bob Greenes).

a. AMIA workshop (Jenders).

b. AMIA panel (Purves, Greenes, Broverman, Bob Dolin/Rick Shiffman).

c. AMIA tutorial (Boxwala, Peleg).

d. Possible MEDINFO late submission (N Jones).

5. Deferred decision on project to identify ongoing applications by vendors to identify functional

standards.

6. Action Item: Request listserv list for SIG and publicize this to attendees (Jenders).

7. Deferred decision on open guidelines.

8. Action item: Consider tools for cooperative workflow (Sailors).

January 12, 2001 0900 - 1130: Arden SIG

Attendance: Mike Jones (Micromedex; mike.jones@mdx.com),

Pat Lyons (SMS; pat.lyons@smed.com),

Joyce Sager (HealthVision),

Aziz Boxwala (Harvard U;aziz@dsg.harvard.edu),

Sven Tiffe (Siemens; sven.tiffe@med.siemens.de),

John Dulcey (Applied Medical Informatics; john.dulcey@alum.mit.edu),

Samson Tu (Stanford U; tu@smi.stanford.edu),

Matt Sailors (UT-Houston; matthew.sailors@uth.tmc.edu),

Robert Jenders (Columbia U; jenders@columbia.edu).


1. Structured Write Statement [Sailors presentation]

a. 3 options:

1. Encode DTD directly into Arden BNF using reserved words.

a. Disadvantage: Many new tokens; lose flexibility of different DTDs.

b. Advantage: Same look and feel as current Syntax.

2. Put XML statement in MLM without changing Syntax, using standard tags to

enable special processing outside of main lexing/parsing modules.

a. Advantage: Less complex to implement than #1; quotations need not be escaped.

b. Disadvantage: Requires 2 new tokens. Requires external runtime

validation/processing.

3. Treat XML as (quoted) string.

a. Advantage: Minimal change to Syntax itself; DTD can be varied, with name

included in string.

b. Disadvantage: No validation at compile time; requires passing string to external processor at run time; double quotes must be escaped.

c. Form: Write at .

b. Discussion of suggested DTD 4 communication ack timeouts (submission, delivery,

display, closure). Some concern expressed that this may inhibit shareability. However, some systems may need such specification.

c. Generally discussed re-opening of suggested DTD for amendment.

d. Decision: Include Structured Write statement in Syntax as Option #3 (7 votes),

2 Abstentions, other options = 0 votes.

e. Action Item: Incorporate this decision into proposed v2.1 draft version and circulate to SIG (Dulcey).

f. Action Item: Update proposal document (DTD) to reflect decisions made at 9/2000 meeting (Jenders).

g. Action Item: Insert suggested DTD as Annex in v2.1 draft proposal (Sailors).


  1. Decision: Proceed to membership ballot with these changes incorporated.

Vote: 7 yes, 0 no, 2 abstentions.
Agreed deadline for completion of document changes:

March 15, 2001. In keeping with the decision of the 9/2000 meeting, will eventually harmonize the DTD structure

with the RIM but not necessarily in advance of the Arden v2.1 balloting.

2. Arden Syntax as XML [Sailors presentation]

a. Suggested at 9/2000 meeting as a possible goal of Arden v3.0.

b. Reviewed one possible model for transforming Arden to XML. Details were discussed.

c. One issue: some Arden constructs have other meanings in XML and may require

adjustment (e.g., "greater than" instead of ">").

d. Issue: Level of model--perhaps down to the statement level but not to the operator level.

e. Potential advantages: Validation of MLM (as valid XML against DTD); display/searching by

available engines; potential harmonization with other HL7 messages and the RIM.

f. Issue: Adding new structure to some slots (author [structured names], keywords, links,

citations).

g. Issue: Need to change some reserved words to codes (e.g., validation levels).

h. General discussion: Advantages vs disadvantages, especially wrt to different levels of

granularity.

1. Potential levels of granularity (akin to the CDA hierarchy):

a. Level 0: Current text document.

b. Level 1: Categories and slots are tagged, all else is the same.

c. Level 2: Structure of lists.

d. Level 3: Structure at the statement level.

e. Level 4: Structure at the operator level.

2. Different levels of granularity may be desired for different applications (searching

for guidelines, execution by an external event monitor, etc).

i. Action Item: Document the current Syntax in Levels 1 and 2 with MLM examples

and circulate in advance of 5/2001 meeting (Sailors).



j. Action Item: Circulate previously agreed TC interaction diagram for comment (Sailors).


База данных защищена авторским правом ©shkola.of.by 2016
звярнуцца да адміністрацыі

    Галоўная старонка