Clinical Guidelines

1
Clinical Guidelines

• Veli Biçer

2
Outline

• Evidence-Based Medicine
• Clinical Guidelines
• Developing Guidelines
• Computerized Clinical Guidelines
• Arden Syntax
• GEM
• PROforma Arezzo

3
Outline contd

• Asbru DeGel
• GUIDE NewGuide
• MyHeart
• EON Athena
• GLIF
• Towards Standardization
• What is next?
• References

4
Evidence-Based Medicine

• Advocates the use of up-to-date best scientific
  evidence from healthcare research as the basis of
  making decisions. It offers
• Objective way to determine high quality and
  safety standards
• The process of transfering clinical findings into
  practice
• Potential to reduce healthcare costs.

5
Clinical Guidelines

• systematically developed statements to assist
  practitioners and patients on decisions about
  appropriate health care for specific
  circumstances" Field and Lohr 1990

6
Developing Guidelines

• Prioritizing Guideline Topic
• Major causes of mortality for a population
• Uncertainty about the appropriateness of
  healthcare
• Need to conserve resources in providing care
• Cardiovascular Diseases is a major category.

7
(No Transcript)
8
Developing Guidelines

• The topic is usually refined since the task of
  developing a guideline for Cardiovascular
  diseases is considerable
• Care Elements
• Primary (The initial and nonspecialized health
  care)
• Secondary (Specialist care in a hospital setting
  )
• Tertiary (Services provided by highly specialized
  providers and tech.)
• Aspects of Management
• Screening
• Diagnosis
• Drug Therapy
• Risk Factor Management

9
Developing Guidelines

• Setting
• Inpatient
• Outpatient
• Time Frame
• Emergency
• Acute
• Chronic

10
(No Transcript)
11
(No Transcript)
12
Developing Guidelines

• Identifying and Assessing the evidence
• Best done by systematic review.
• The Cochrane Library contains references to over
  218000 clinical trials
• http//www.cochrane.org/
• Once gathered, the evidence is interpreted and
  translated into CPG.

13
(No Transcript)
14
(No Transcript)
15
Computerized Clinical Guidelines

• Most clinical guidelines are text-based
• All of them is not accessible online
• Physicians have difficulties in deciding which of
  multiple guidelines best pertains to their
  patient
• A clear need for effective guideline-support
  tools at the point of care
• To be effective, these tools
• need to be grounded in the patient's record
• must use standard medical vocabularies
• should have clear semantics
• must facilitate knowledge sharing

16
Computerized Clinical Guidelines

• Approaches to Electronic Guideline Representation
• Formal Representation Specification
• Encoding logic into application-specific format
• Guideline Modeling Methodologies
• Rule-based Arden Syntax
• Logic-based PROforma
• Workflow GUIDE, GLIF

17
Arden Syntax

• HL7/ANSI standard
• Current approved version is 2.1
• Standard, formal procedural language that
  represents medical algorithms in clinical
  information systems as Medical Logic Modules
  (MLMs).
• MLM an independent unit in a health knowledge
  base. It contains
• Maintenance Information
• Links to other sources
• Logic to make a single decision

18
Arden Syntax

• maintenance
• title Hepatitis B Surface Antigen in Women
• mlmname hepatitis_B_mlm
• arden version 2.1
• ...
• library
• keywords hepatitis B
• citations
• 1. Goldman L, Cook EF, et al. A computer protocol
  to predict myocardial infarction. N Engl J Med
  1988318(13)
• ...
• knowledge
• data penicillin_storage event store
  penicillin order
• evoke penicillin_storage
• evoke 3 days after time of creatinine_storage
  
• var1 call my_interface_function with param1,
  param2
• logic
• if last_creat is not present then
• alert_text "No recent creatinine available.
  Consider ordering creatinine before giving IV
  contrast."
• conclude true

19
Arden Syntax

• Advantages
• Not a full-feature programming language Suitable
  for Clinicians.
• Provides explicit links to data, trigger events.
• Defines how an MLM can be called (evoked) from a
  trigger event.
• Brings particular support for time functions.
• HL7/ANSI standard
• Used by Commercial DSSs.

20
Arden Syntax

• The basic format is not appropriate for
  developing complete electronic guideline
  applications
• Not as declarative as GLIF
• In case of an interaction with a clinical
  database to provide alerts and reminders, the
  encoding of clinical knowledge (MLM) may vary due
  to database schema, clinical vocabulary.

21
GEM

• Guideline Elements Model
• XML-based guideline markup model
• International ASTM (American Society for Testing
  and Materials) standard.
• The free-text is markup in XML.

22
(No Transcript)
23
PROforma

• A formal knowledge representation language
• EU 4th Framework Health Telematics PROMPT project
  
• Guideline is modeled as a set of
• Tasks
• Data Items
• Tasks are divided into
• Actions
• Enquiries
• Decisions
• Plans
• PROforma software consists of a graphical editor
  to support the authoring process, and an engine
  to execute the guideline specification.
• Two major tools AREZZO, TALLIS

24
PROforma
25
AREZZO

• Software to create and run clinical guidelines
  based on PROforma
• Commercial
• Two main components Composer, Performer
• PROforma provides some rules supported by AREZZO
• Performer has Microsoft COM Interface

26
(No Transcript)
27
(No Transcript)
28
(No Transcript)
29
Asbru

• The Asgaard project led by the Vienna University
  of Technology and Stanford Medical Informatics,
  1998
• A task-specific and intention-based plan
  representation language
• Embody clinical guidelines and protocols as
  time-oriented skeletal plans
• Regarding the timing, the plans can be
  Sequential, Parallel, Any-order, Unordered.

30
(No Transcript)
31
Asbru

• ltplan-librarygt
• ltlibrary-info title"Skeleton of a Plan
  Library/gt
• ltlibrary-defsgt
• ltdomain-defsgtlt/domain-defsgt
• ltvariable-def name"List-1" scalar-or-not"list"
  type"string"gt
• ltcomment text"List-1 is a list of strings"/gt
• lt/variable-defgt
• ltconstant-def name"PI"gt
• ltnumerical-constant unit"amount"
  value"3.1415"/gt
• lt/constant-defgt
• ltfunction-def class-name"asgaard.checkit
  method-name"add_em_up name"add
  return-type"length"/gt
• lt/library-defsgt
• ltplansgt
• ltplan-groupgt
• ltplan name"Plan-A"gtlt/plangtltplan
  name"Plan-B"gtlt/plangt
• lt/plan-groupgt
• lt/plansgt
• lt/plan-librarygt

32
(No Transcript)
33
Asbru

• Records can also be defined in domain definitions
  and used as an interface to plans

34
DeGel

• Digital Electronic Guideline Library
• Developed tools to support the development and
  implementation of guideline applications.
• Expert physicians cannot program in guideline
  specific language, while engineers do not
  understand the clinical semantics
• Problem How will the large mass of free text
  guidelines be converted to a formal
  machine-readable language?

35
DeGel

• Based on a hybrid (multiple-format) electronic
  representation of guidelines
• A guideline is first converted from free text
  into semantically semi-structured text
• Then from semi-formal language by a medical
  expert using a markup editor, to a fully formal
  representation by a knowledge engineer
• The current default target language is Asbru

36
DeGel
37
DeGel

• The framework provides the following tools
• Uruz - Gradual conversion of free-text clinical
  guidelines into a machine-comprehensible
  representation in a given target guideline
  ontology
• IndexiGuide - Manual or automated classification
  of clinical guidelines along multiple semantic
  axes
• Vaidurya - Search and retrieval of clinical
  guidelines represented in free text, or in a
  semi-structured format that uses the labels of a
  given target ontology
• VisiGuide - Visualization and browsing of a set
  of guidelines in a target ontology

38
Uruz

• Guideline markup tool
• Similar to GEM Cutter Editor
• Source guideline (free-text) is loaded and marked
  up with semantic labels of the target ontology.
• The target ontology can only be Asbru or GEM
• The result is an XML document

39
DeGel
URUZ
40
(No Transcript)
41
Uruz

• Plan Body Builder
• Specific to Asbru
• Used for defining guidelines control structure
• Decompose actions into atomic actions and other
  sub-guidelines

42
(No Transcript)
43
(No Transcript)
44
IndexiGuide

• Allows medical experts to index the guidelines
  with semantic axes
• Semantic axes can be signs, symptoms, diagnostic
  findings, disorders, treatments and so on.
• Semantic axes are headers of standardized
  vocabularies such as MeSH, ICD-9, CPT

45
(No Transcript)
46
Vaidurya

• Guideline search and retrieval tool
• The user can search based on semantic axes
• The marked-up guidelines can also be queried for
  the existence of the terms within internal context

47
(No Transcript)
48
Visiguide

• Visualization of multiple and single guidelines
• Free text, semi-structured text and formal
  language (Asbru).
• Organizes the guidelines along semantic axes

49
(No Transcript)
50
GUIDE- NewGuide

• GUIDE 1998
• Reengineered to NewGuide in 2002
• Laboratory for Medical Informatics, Department of
  Computer and System Science, University of Pavia,
  Italy
• The Guide environment integrates three main
  independent modules
• Guideline Management System (GlMS) (providing
  clinical decision support)
• Electronic Patient Record (EPR)
• Workflow Management System (WfMS or CfMS)
  (providing organisational support)

51
GUIDE- NewGuide
52
GUIDE- NewGuide

• Different views of the formalized knowledge to
  allow different people with different roles (e.g.
  clinicians, patients, administrators...) to have
  their own context-specific interactions with the
  system www.openclinical.org
• For example, if a guideline suggests taking a
  blood sample (Lab Test), the physician view would
  incorporate the interpretation of the examination
  results (CPG), while the patient view would
  provide a reminder and a facility to book the
  blood examination (Healthcare Process)

53
GUIDE- NewGuide

• Guideline management system for handling whole
  lifecycle of a CCPG
• Two main levels Central, Local
• The Components
• An Editor to formalize guidelines
• Repository to store
• Inference Engine to implement
• Reporting System to logging
• Implemented in Java and uses SOAP for the
  integration with HIS

54
GUIDE- NewGuide
Manage GLs by some health authority or scientific
organization
Healthcare Organization adopting one or more GLs
55
GUIDE- NewGuide

• GL Lifecycle
• Constructing GL with NewGuide Editor
• Storing GL by Repository Manager at local and
  central level. Two DB, one for metadata, one for
  GL Template
• Final user retrieves GL Template by Inference
  Engine and creates an instance with VMR of the
  patient
• Inference engine produces recommendations such as
  drug pres., lab. test by updating log

56
GUIDE- NewGuide

• NewGuide Editor
• Produces four XML data structure
• General properties in GEM
• The set of medical terms based on ICD and LOINC
• Abstractions
• GL Flow

57
GUIDE- NewGuide

• NewGuide Repository
• Manages the Guidelines
• GL general properties are used for querying

58
GUIDE- NewGuide

• Inference Engine
• An instance of a GL is created by using the VMRs
• Includes Instance Manager for the management of
  instances.
• Instance Manager can start, finish, drop,
  suspend, activate GL execution
• CfMS manages the flow and timing of the GL

59
(No Transcript)
60
GUIDE- NewGuide

• For a recommendation such as Wait for 2 days,
  CfMS decides to put the instance to stand by.
• GL represents medical knowledge, while CfMS is
  responsible for execution.
• When an info acquisition task is scheduled,
  inference engine can request through the SOAP
  from the HIS. GL is put on stand-by

61
(No Transcript)
62
MyHeart

• Project Acronym MYHEART Project Reference
  507816 Start Date 2003-12-31 Duration 45
  months Project Cost 34.92 million euro
  Contract Type Integrated Project End Date
  2007-09-29 Project Status Execution Project
  Funding 16.00 million euro
• PHILIPS GMBH AACHEN GERMANY
• http//www.extra.research.philips.com/euprojects/m
  yheart/

63
MyHeart

• aims to fight cardio-vascular diseases by
  prevention and early diagnosis. 
• Intelligent Biomedical Clothes The combination
  of functional clothes (including sensors) and
  integrated electronics.
• Intelligent Biomedical Clothes for monitoring,
  diagnosing and treatment.

64
MyHeart

• The main Technical Challenges are - Continuous
  Monitoring- Continuous Personalised Diagnosis-
  Continuous Therapy- Feedback to user- Remote
  Access and Professional Interaction

65
MyHeart

• Works to be done during the project.
• Applications and personalized algorithms.
• Functional Clothes including sensors with
  long-term monitoring capability.
• Developing on-body electronics integrated to the
  clothes.
• Developing a system architecture for user and
  professional interaction.
• No public results yet

66
EON

• A component-based suite of models and software
  components for the creation of guideline-based
  applications
• Stanford Medical Informatics
• Support the National Library of Medicine
• Uses Protégé

67
EON

• Provides an extensible set of ontologies covering
  different aspects of concepts and relations
  needed for encoding CPG
• Ontologies are
• Patient Data Model( the classes and attributes of
  patient data (EMR))
• Concept Model (like archetypes)
• Guideline Model
• Expression/Criterion Model
• Temporal Model

68
EON

• Patient Data Model
• Patient class hold demographic information
• Note_Entry class that describes qualitative
  observations about patients
• Numeric_Entry class that represent results of
  quantitative measurements
• Medication and Procedure model drugs and medical
  procedures
• Not try to create a data model that replicates
  everything that an EMR holds, but only those
  relevant for modeling guidelines.

69
EON

• Concept Model
• The concepts we want to model in the concept
  model are abstract entities that can be organized
  into taxonomic hierarchies.
• Concrete subclasses are created and used by the
  Guideline Model

70
EON

• Guideline Model
• Uses the patient data and concept models to
  create GL
• Classes to model Guideline
• Goal and Step
• Clinical algorithm
• Activity and Action Specifications

71
Athena

• Assessment and Treatment of Hypertension
  Evidence-based Automation
• Decision support system for the management of
  hypertension in primary care
• Mostly depend on Sixth report of the Joint
  National Committee on Prevention, Detection,
  Evaluation, and Treatment of High Blood
  Pressure(JNC6)
• Currently JNC7 is available
• Stanford Medical Informatics and VA Palo Alto
  Health Care System

72
Athena

• Encourages blood pressure control
• Recommends guideline-concordant choice of drug
  therapy
• Easily modifiable knowledge base that specifies
  eligibility criteria, risk stratification, blood
  pressure targets, relevant diseases,
  guideline-recommended drug classes for patients
  preferred drugs within each drug class, and
  clinical messages.
• Designed to allow clinical experts to customize
  the knowledge base to incorporate new evidence or
  to reflect local interpretations of guideline
  ambiguities.
• Database mediator, Athenaeum
• Physical and logical data independence from the
  legacy Computerized Patient Record System (CPRS)
  supplying the patient data

73
Athena

• Two major components
• A knowledge base that models hypertension
  independently of its use
• Guideline interpreter creating patient specific
  treatment recommendations

74
Athena

• Knowledge base
• Based on EON
• A computerized version of JNC6 (Prevention,
  Detection, Evaluation, and Treatment of High
  Blood Pressure) GL
• Clinicians can customize through Protégé

75
Athena

• The EON based system also determines
• Whether or not the GL is applicable to a patient
• Which portion is applicable
• Whether the goal has been reached
• Applies criteria for selecting an action
• For the EON based system to work, patient
  clinical data is needed

76
Athena

• Athenaeum Database Mediator
• Maps legacy database onto data model of Athena
  DSS
• In addition to data model, terminology is also
  mapped
• From ICD-9 to EON internal codes

77
(No Transcript)
78
Athena
79
Athena

• Advisory for clinicians
• Clinical Assumptions used for reasoning
• Recommendations

80
Athena

• Clinical Assumptions
• Patient Risk Class
• Patient Data considered in calculation
• Target Blood Pressure and whether it is achieved
  or not
• Additional Blood Pressure Readings entered by
  clinicians

81
Athena

• Recommendations
• Increasing/decreasing the dose of a specific drug
• Using a new drug
• Warnings to patient or clinician

82
Athena

• Tested in 100 cases
• 224 drug recommendations
• 87 disagreements between clinicians and Athena
• 12 ATHENA errors!!!

83
GLIF

• The Guideline Interchange Format
• InterMed Collaboratory (Stanford Medical
  Informatics, Harvard University, McGill
  University and Columbia University)
• GLIF
• Defines an ontology for representing guidelines,
  and a medical ontology for representing medical
  data and concepts.
• Tools are under development to support guideline
  authoring and execution.

84
GLIF

• Guidelines are represented as a flowchart of
  guideline steps
• Guideline steps
• Decision Step
• Action Step
• Medically oriented actions
• Programming-oriented actions
• Branch, Synchronization Step
• Patient State Step

85
GLIF

• Layers of abstraction
• Conceptual Level of Representation (Level A)
• Computable Level of abstraction (Level B)
• Implemental Level (Level C) (Not completed yet)

86
GLIF

• Level A When a guideline is first authored, a
  conceptual level of representation is created
• the guideline author to concentrate on
  conceptualizing a guideline as a flowchart
• the author is not concerned with formally
  specifying details, such as decision criteria,
  relevant patient data, and iteration information
  that must be provided to make the specification
  computable

87
GLIF

• Guideline Model
• Guideline class Actual GL subGL class
• Algorithm a flowchart of GL steps
• Maintenance Info metadata about GL
• A GL uses the instances of the Medical Ontology
  through its data_items and parameters_passed
  attributes

88
(No Transcript)
89
(No Transcript)
90
(No Transcript)
91
(No Transcript)
92
(No Transcript)
93
(No Transcript)
94
GLIF

• Expressions
• Guideline_Expression class
• Guideline Expression Language (GEL) is developed
  based on the Arden Syntax grammar
• A new language, GELLO, is developed based on the
  decision support execution model proposed in HL7
  Clinical Decision Support TC
• This standard with GELLO will be adopted when it
  is published by HL7

95
(No Transcript)
96
(No Transcript)
97
(No Transcript)
98
GLIF

• Medical Ontology
• GLs and GL Components (Logical expressions and
  action specifications) use the Patient Data and
  Medical Concepts
• The concepts are defined by referencing
  controlled vocabularies (UMLS) and standard
  medical data models (HL7 RIM)

99
GLIF

• Layers of Medical Ontology
• Core GLIF
• Reference Information Model (RIM)
• Medical Knowledge Layer (Under Development)

100
GLIF

• Core GLIF
• Defines medical data model
• Data types are classified into
• Primitive Data Item
• Data Item
• Knowledge Item

101
(No Transcript)
102
(No Transcript)
103
GLIF

• Reference Information Model (RIM)
• Adopted from HL-7 RIM Version 0.94
• The Act class is renamed to Patient_Data
• Extension?
• Patient Data in HIS message ontology can be
  mapped to current RIM if it is adequate
• The sensor data may require additional classes
• A new RIM can be adopted by defining a new
  ontology

104
GLIF
105
GLIF

• A Draft Standard
• The tools are not available
• For creating GLs in three layers of abstraction
• For validation and testing
• Protégé is not so user-friendly for GL definition
• Medical Knowledge Layer is not implemented yet
• RIM may require extensions
• Guideline Expression Languages
• GEL
• GELLO (Not adopted)

106
GLIF

• Consensus based multi-institutional process
• Open process
• Planning to support the use of multiple
  vocabularies and data models
• Incorporates complementary specifications such as
  Arden Syntax, HL7

107
Towards Standardization
108
Towards Standardization

• International Workshop, Toward Sharable
  Guideline Representation by InterMed
  Collaboratory
• Near-term goals
• To move toward a common standard
• To create prototype authoring tools
• Provide mechanisms to link GL to the EHRs

109
What is next?

• Intelligent Clinical Decision Support Systems
• Decision support systems survey
• ATHENA, CEMS, DXplain, ERA , PRODIGY, RetroGram
• Agent based clinical decision support systems
• Design an engine
• Glif3 Guideline Execution Engine (GLEE)
• GELLO, An Object-Oriented Query and Expression
  Language for Clinical Decision Support
• HL7 CDSTC HL7 Clinical Decision Support Technical
  Committee

110
References

• Field MJ, Lohr KN (Eds). Guidelines for clinical
  practice from development to use. Institute of
  Medicine, Washington, D.C National Academy
  Press, 1992.
• Shekelle, P. Woolf, S. Eccles, M. Grimshaw, J.
  Clinical guidelines developing guidelines /
  British Medical Journal (BMJ) , 1999
• GLIF 3.5 Technical Specification, InterMed
  Collaboratory

111
References

• Sutton D, Fox J. The syntax and semantics of the
  PROforma guideline modelling language.
• Shakar et. al. DeGel A hybrid, multiple ontology
  framework for specification and retrieval of
  Clinical Guidelines
• Elkin et. al. Toward Standardization of
  Electronic Guideline Representation

112
References

• Coiera E., Clinical Decision Support Systems,
  Book Chapter, Guide to Health Informatics 2nd
  Edition
• Samson et. al., Modelling Data and Knowledge in
  the EON Guideline Architecture
• Goldstein et al., Implementing Clinical Practice
  Guidelines while taking account of changing
  evidence

113
References

• Peleg et al., Comparing Computer-Interpretable
  Guideline Models A Case-study approach
• Ciccarese et al., A guideline management system
• Sackett et al, Evidence-based medicine what it
  is and what it isnt.
• www.openclinical.org
• Peter et al, A virtual Medical Record for
  Guideline-Based Decision Support

114
References

• Specifications of
• PROforma
• Asbru
• EON
• Arden
• GEM

115
Thanks for your attention