Clinical Decision Support Lecture Brief History and State of the Art of Clinical Decision Support and relation to Terminology and Electronic Healthcare Records (EHRs) Available at

1
Clinical Decision Support Lecture Brief History
and State of the Art of Clinical Decision Support
and relation to Terminology and Electronic
Healthcare Records (EHRs)Available at
http//www.cs.man.ac.uk/rector/modules/cds/Notes
-1-HI-general-cds-2006.ppt
2
The Hype of the Time

• Guidelines
• Evidence Based Medicine
• Clinical Errors (reducing)
• Improving prescribing practice
• Reducing adverse drug reactions
• Protocols
• Knowledge Management
• ...

3
Clinical Judgement and Clinical Errors

• To Err is Human http//www.nap.edu/books/030906837
  1/html/
• Supporting a Humanly Impossible Task
  http//www.cs.man.ac.uk/rector/papers/Humanly-Imp
  ossible-Task.pdf
• Johnson Articles - see resourceshttp//www.cs.man
  .ac.uk/rector/modules/cds/cds_links.htm(NB some
  links may be broken because of University merger)
• OpenClinical Web sitehttp//www.openclinical.org/
  
• OpenEHR web sitehttp//www.openehr.org

4
Computer Aided Decision Support Works (sometimes)

• Evidence of effectiveness growing
• 25 years since Clem McDonalds Protocol-based
  computer reminders, the quality of care and the
  non-perfectability of man
• Use still limited
• Meta studies and reviews a decade old
• Elson R E and Connelly D P (1995). Computerized
  patient records in primary care Their role in
  mediating guideline-driven physician behaviour
  change. Archives of Family Medicine 4 698-705.
• Grimshaw J and Russell I (1993). Effect of
  clinical guidelines on medical practice a
  systematic review of rigorous evaluations. Lancet
  342 1317-1322.
• Johnston M, Langton K, Haynes R and Mathieu A
  (1994). Effects of computer-based clinical
  decision support systems on clinical performance
  and patient outcome. A critical appraisal of
  research. Archives of Internal Medicine 120
  135-142.

5
Important Recent Study

• Cristina Tural, Lidia Ruiz , Christopher Holtzer,
  Jonathan Schapiro, Pompeyo Viciana, Juan Gonzàlez
  , Pere Domingo,Charles Bouche, C. Rey-Jol.
  BonaventuraClotet and the Havana Study Group
  (2002) Clinical utility of HIV-1 genotyping and
  expert advice Havana trial, AIDS 16(2) 209-215

6
Types of Decision Support Information Tasks

• Informative
• Guidelines e.g. eBNF, BMJ Clinical Evidence,...
• Literature search - DxPlain
• Information structuring
• intelligent records (EPRs)
• PENPAD, Medcin vocabulary, ...
• Triggers and warnings
• MLMs, McDonalds original work, HELP, ...
• Critiquing - Perry Miller
• Advising

7
Types of Decision Support Clinical Tasks

• Management Protocols(often effective, Johnston
  et. al 1994)
• Prescribing
• Protocol based care
• Oncocin, T-Helper, etc.
• Referral
• Diagnosis(rarely effective, Johnston et. Al
  1994)
• Mycin
• Internist I
• Knowledge Couplers

8
Reasons for success and failure(1)

• Understanding of problem
• Meeting real and recognised needs
• Forsythe D E (1992). Using ethnography to build a
  working system rethinking basic design
  assumptions. Sixteenth Annual Symposium on
  Computer Applications in Medical Care (SCAMC-92),
  Baltimore, MD, Baltimore, MD 505-509.
• Meeting them effectively
• The user is always right but the user is
  usually wrong
• The technology is still crude at best
• Implementing it successfully

9
Reasons for success and failure(2)

• Most projects fail at implementation!
• The technology only works if people want it and
  use it
• Requires emphasis on participation, ownership,
  training, respect for practicalities
• Implementation begins with design
• Evaluation begins with design
• Formative evaluation essential
• See Shortliffe Shortliffe The Adolescence of AI
  in Medicine Will the Field Come of Age in the
  1990's? Artificial Intelligence in Medicine,
  593-106, 1992. http//smi-web.stanford.edu/pubs/S
  MI_Abstracts/SMI-92-0449.html

10
Potted History (1)

• Bayesian stream
• 1968 Ledley and Lusted Diagnosis using Idiot
  Bayes discriminant
• Followed by Pauker Decision Support using utility
  theory
• 1970-1985 - de Dombal Idiot Bayes abdominal
  pain and other surgical diagnostic problems
• Meanwhile RCP Computer Workshop refined
  discriminants and then stimulated Spiegelhalter
  to come up with practical algorithms for belief
  nets in early 1990s
• 1980s Society for Medical Decision Making formed
  and statistical work largely separated from rule
  based work

11
Idiot Bayes

• A simple statistical means to use databases to
  determine weights.
• Collect a sample of patients with each disease,
  e.g.Acute Abdominal Pain patients 100 each
  ofAppendicitis, Cholecystitis, Pancreatitis,
  Perforated Ulcer, Obstruction, GI Cancer, Tubal
  pregnancy (in women only)
• Add a catch-all for everything else Non
  specific Abdominal pain
• Assume that all symptoms are caused independently
  by each disease e.g. that the mechanisms for
  rebound tenderness and nausea are different.
• Derive a table of probabilities to be combined
  using the Idiot Bayes formula
• Proved much more robust than less idiotic
  methods

12
Potted History (2)

• Rule based stream
• 1972 - Shortliffe Mycin First rule based system
• 1970s US AIM Workshop produced Big 4
• Mycin/Oncocin/Puff - Backwards chaining shells
• Interist I - NEJM CPCs from a large network
• Became QMR as a general reference
• Casnet - Multilayer causal reasoning (glaucoma)
• Abel - Complex causal networks (acid-base
  metabolism)
• 1990s Protocol based reasoning
• Protégé/Eon successors to Mycin/Oncocin at
  Stanford
• Musen MA. Domain ontologies in software
  engineering use of Protégé with the EON
  architecture. SMI Technical Report 97-0657.
  Methods of Information in Medicine 37540-550,
  1998.
• ProForma at ICRF
• ASBRU
• PRODIGY III

13
Typical Mycin Rules

• IF the gram-stain is gram-negativeAND if the
  culture-site is sterileAND if the culture-site
  is bloodAND if the aerobicity is anerobicTHEN
  there is strong (.8) evidence that the organism
  is enterobacter
• Based on expert opinion rather than data

14
Potted History (3)

• Reminders
• 1970 - Homer Warner, HELP, LDS
• 1980s - Arden Syntax
• 1990s - MLMs - standardised Arden
• 1970s - Clem McDonald - reminders and the
  nonperfectability of man
• Regenstrief laboratory systems
• Many variations
• PRODIGY II
• Systematic Review Johnston M, Langton K, Haynes
  R and Mathieu A (1994).

15
Potted History (4)

• Offshoots and Idiosyncratics
• Critiquing - Perry Miller
• Also Johan van der Lei
• Quick Medical Reference - Chip Masari
• Intelligent Records - Alan Rector and Anthony
  Nowlan
• Knowledge Couplers - Larry Weed

16
Potted History (5)

• Knowledge Management and the Web
• 1980s Grateful Med and DxPlain
• Quick access to Medline abstracts and related
• 1990s The Web with everything
• Rise of Evidence Based Medicine
• Cochrane, NICE, NELH, Health on the Web (HoN),
• Indexing and meta data
• How do you find it
• Portals and certification
• How do you know if it is any good
• Information for Public and Patients
• Its an open world out there
• Type Diabetes Support at Google 776,000 hits,
  AllTheWeb 295,000Yahoo 26, Netscape 2000
• Classic Information Retrieval and Librarianship
• Digital Libraries
• Different fields with little contact

17
Examples of Web Based Initiatives

• DxPlain
• PaperChase
• Health on the Net (HoN)
• OpenClinical
• Baby CareLink
• Guardian Angels
• and of course PubMed and the NLM initiatives

18
Why isnt decision support in routine use?

• Hypothesis one Pearls before swine
• Doctors are resistant
• Hypothesis two The Emporers new clothes
• Systems are not clinically worthwhile
• Not clinically useful
• Too time consuming - too hard to learn
• Too expensive
• Too inaccessible
• Too sparse
• How many diabetic patients does a GP see per
  week?
• Easier ways to get help
• The technology is still primitive
• Developers misunderstand medicine
• They think it is rational!

19
Why isnt decision support in routine use?

• Hypothesis 3 The invisible computer
• When it works, no one notices
• ECG interpretation
• Alerts and reminders
• NHS Direct
• Simple but effective?
• Junior doctors PDAs
• Convergence of communication and computing
• Upmarket PDAs have 10-100 times the power of the
  machine that first ran Mycin!
• Why Web technology and XML are critical to this
  course
• divorce content and presentation

20
What would you want from decision support?

• Discussion break

21
Some Technical Issues

• Technical
• Re-use, transfer, and Terminology
• Links to medical records
• Protocols and Problem Solving Methods
• Combinatorial explosions
• Context and common sense
• Cognitive utility
• The demise of the oracle
• The difficulty of mixed initiative systems

22
The Interface of Three Technologies /Modelling
Paradigms

• Terminology and Ontologies
• Electronic Patient Records
• Decision Support/Inferencing
• including abstraction
• Plus Information Management/Information
  Retrieval

23
(No Transcript)
24
A Protocol
25
Who Should Be Evaluated for UTI? Under the
assumptions of the analysis, all febrile children
between the ages of 2 months and 24 months with
no obvious cause of infection should be evaluated
for UTI, with the exception of circumcised males
older than 12 months. Minimal Test
Characteristics of Diagnosis of UTI To be as
cost-effective as a culture of a urine specimen
obtained by transurethral catheter or suprapubic
tap, a test must have a sensitivity of at least
92 and a specificity of at least 99. With the
possible exception of a complete UA performed
within 1 hour of urine collection by an on-site
laboratory technician, no other test meets these
criteria. Performing a dipstick UA and
obtaining a urine specimen by catheterization or
tap for culture from patients with a positive LE
or nitrite test result is nearly as effective and
slightly less costly than culturing specimens
from all febrile children. Treatment of UTI The
data suggest that short-term treatment of UTI
should not be for lt7 days. The data do not
support treatment for gt14 days if an appropriate
clinical response is observed. There are no data
comparing intravenous with oral administration of
medications. Evaluation of the Urinary Tract
Available data support the imaging evaluation of
the urinary tracts of all 2- to 24-month-olds
with their first documented UTI. Imaging should
include VCUG and renal ultrasonography. The
method for documenting the UTI must yield a
positive predictive value of at least 49 to
justify the evaluation. Culture of a urine
specimen obtained by bag does not meet this
criterion unless the previous probability of a
UTI is gt22. FOOTNOTES The recommendations in
this statement do not indicate an exclusive
course of treatment or serve as a standard of
medical care. Variations, taking into account
individual circumstances, may be appropriate.
26
Semi Structured in GEM as seen in Gem Cutter
27
(No Transcript)
28
Terminology, Medical Records, and the curly
bracket problem

• Re-use
• Why should everyone start from scratch?
• Attempts to transplant HELP complete did not work
• Could we transfer fragments of Help?
• Workshop at IBM centre at Arden near New York
  City produced generalisation of HELP syntax
• The Arden Syntax - now renamed Medical Logic
  Modules, MLMs

29
Example Arden Syntax

• Data Slot
• creatinine read 'dam'"PDQRES2"
• last_creat read last select "OBSRV_VALUE"
  from "LCR" where qualifier in
  ("CREATININE","QUERY_OBSRV_ALL")
• Items in curly brackets are institution
  specific

Source MLM Tutorial AMIA 2001 here
30
Arden Syntax - Next bit but from another
institution

• data
• creatinine_storage event '32506','32752'
  / isolated creatinine / ...'32506','33801'
  / chem 20 /
• evoke
• creatinine_storage
• Items in curly brackets are institution
  specific

31
The Curly Bracket Problem

• Transfering the logic is easy
• Transfering the access rules in curly brackets is
  hard
• And it takes your most skilled people
• Subtle dependencies and system indiosyncracies
• The need for a common vocabulary

32
Where we come from
Best Practice
Best Practice
33
Controlled Vocabularies and Ontologies

• A common theme
• Affects Protégé/Eon, ProForma, ASBRU etc
• Protégé/Eon based on Shared Problem Solving
  Methods (PSM) and shared Ontology
• A library of PSMs. No reused ontologies!
• The glue to link Medical Records and Clinical
  Decision Making
• But only half the problem
• Systems must have the same concepts
• Doctors must use the same concepts
• But made worse because most vocabulary is so
  awful to use

34
The Link to Medical Records

• The Terminology provides the content for the
  boxes in the information model

35
Surgical
Procedure
Disease
has treatment
Surgical
has diagnosis
Patient
Disease
Procedure
has complication
36
Surgical
Procedure
Disease
Excision
Infection
Melanoma
has diagnosis
has treatment
Excision
Melanoma
has complication
37
Protocols and Problem Solving Methods

• Machines and people
• If it is easy for people it is hard to specify
  logically and program
• and vice versa
• A real Guideline from NICE here
• And from GEM site here
• What do you do with one of these?
• What does it mean operationally?
• See next page for extract from GEM site protocol
  on UTI in children

38
How might we do it?

• Can you make a simple Clinical Algorithm from
  the previous?
• Can you scale this up to a cancer chemotherapy
  protocol

39
Todays Standards

• EHRs
• HL7 v2 and v3
• OpenEHR / CEN 13606 / Ocean Informatics
  Archetypes
• Terminology
• SNOMED-CT
• Clinical Terms V2
• ICD 9/10 (CM)
• Specialist terminologies

40
HL7 - A Very Brief Intro
41
HL7 Reference Information Model (The RIM)
42
HL7 RIM Backbone (UML)
43
HL7 RIM Backbone as Block-Diagram
44
HL7 Data in XML

• ltact classCodeACT moodCodegt ltid
  root1.3.6.1.4.1.12009.3 extensionA1234/gt ltc
  ode code... codeSystem2.16.840.1.113883.6.1/
  gt
• ltparticipant typeCodegt
• ltparticipant classCodeROLgt
• ltid root1.3.6.1.4.1.12009.4
  extension1234567-8/gt ltcode code
  codeSystem2.16.840.1.113883.6.21/gt
• ltplayingEntity classCodeENTgt
• ltnamegt...lt/namegt
• lt/playingEntitygt
• ltscopingEntity classCodeENTgt
• ltnamegt...lt/namegt
• lt/scopingEntitygt
• lt/participantgt
• lt/participantgt
• ltsourceOf typeCodeRELgt
• lttarget classCodeACTgt
• ltid root1.3.6.1.4.1.12009.3
  extensionA1235/gt
• lt/targetgt
• lt/sourceOfgt
• lt/actgt

45
Refined Model Observation on Patient
46
Observation on Patient in XML

• ltobservationEvent classCodeOBS
  moodCodeEVNgt ltid root1.3.6.1.4.1.12009.3
  extensionA1234/gt ltcode code...
  codeSystem2.16.840.1.113883.6.1/gt
• ltsubject typeCodegt
• ltpatient classCodeROLgt
• ltid root1.3.6.1.4.1.12009.4
  extension1234567-8/gt ltcode code
  codeSystem2.16.840.1.113883.6.21/gt
• ltpatientPerson classCodePSNgt
• ltnamegtltgivengtJohnlt/givengtltfamilygtDoelt/familygtlt
  /namegt
• lt/patientPersongt
• ltproviderOrganization classCodeORGgt
• ltnamegtSt., Josephs Hospitallt/namegt
• lt/providerOrganizationgt
• lt/patientgt
• lt/subjectgt
• ltcomponent typeCodeRELgt
• ltobservationEvent classCodeACTgt
• ltid root1.3.6.1.4.1.12009.3
  extensionA1235/gt
• lt/observationEventgt
• lt/componentgt
• lt/observationEventgt

47
Refined Model Observation on Trial Subject
48
Observation on Trial Subject in XML

• ltobservationEvent classCodeOBS
  moodCodeEVNgt ltid root1.3.6.1.4.1.12009.3
  extensionA1234/gt ltcode code...
  codeSystem2.16.840.1.113883.6.1/gt
• ltsubject typeCodegt
• ltresearchSubject classCodeROLgt
• ltid root1.3.6.1.4.1.12009.5
  extension1234567-8/gt ltcode code
  codeSystem2.16.840.1.113883.6.21/gt
• ltsubjectPerson classCodePSNgt
• ltnamegtltgivengtJohnlt/givengtltfamilygtDoelt/familygtlt
  /namegt
• lt/subjectPersongt
• ltresearchSponsor classCodeORGgt
• ltnamegtEli Lillylt/namegt
• lt/researchSponsorgt
• lt/researchSubjectgt
• lt/subjectgt
• ltcomponent typeCodeRELgt
• ltobservationEvent classCodeACTgt
• ltid root1.3.6.1.4.1.12009.3
  extensionA1235/gt
• lt/observationEventgt
• lt/componentgt
• lt/observationEventgt

49
Archetypes

• Find on openEHR web site
• Google OpenEHR

50
OpenEHR http//www.openehr.org
51
Ex General Biochemistry
52
Ex Blood Lipids
53
Computers can do anything you can tell them to

• How to write a perfect chess programme
• List all the possible first move
• For each first move, list all the possible
  answering moves
• For each answering move list all the replies
• ...
• A 10 line programme
• So why dont computers play perfect chess?

54
Combinatorial Explosion 20 questions
1
Q
2
no
yes
4
yes
no
yes
no
8
no
yes
no
yes
no
yes
no
yes
...
The legend of the Persian chess board
55
Combinatorial Explosion!

• 264 2106.4 1019
• 1019 milliseconds 1011 days 109 years
• 1 billion years
• 1019 nanoseconds 1000 years
• 1019 grains of wheat 1000 million metric tons
  of wheat
• Predicted world wheat production for 2001 567
  million metric tons

Brute force does not always work!(But dont
underestimate brute force cleverly applied -
consider Google)
56
The human brain

• How big? How fast?
• 1010 neurons
• 105 connections per neuron
• 103 firings per second
• ?1018 floating point operations / second to
  simulate and 1018 memory locations equivalent
• Probably a gross under estimate

57
Computers

• Current computers around 109-1012
• At least 106 to go!
• 106 ? 220
• Moores law says power doubles every 1.5 years
• Roughly 30 years to go!
• And then what?
• See recent controversy http//www.tecsoc.org/innov
  ate/focusbilljoy.htm (click here)
• Heuristics
• Rules of thumb
• As opposed to Algorithms - procedures with
  guaranteed solutions

58
Technologies

• Problem Solving Methods require
• Knowledge Representation
• Semantic nets, frames, description logics,
  ontologies
• Inference
• Rule based systems
• Planning
• Skeletal Plan Refinement
• Bayesian Reasoning
• Belief Nets
• Logic Engines

Programming by Search