Dale will take a slide deck previously prepared in 2006, from a lecture entitled, "The Power of an Enterprise Data Warehouse in Clinical Decision Support", presented to several informatics masters classes at Northwestern University and the University of Victoria. He won’t change anything about the slide deck, including the content and the old school graphics. The concept with this webinar is to give a “time capsule” perspective on past thinking and contrast that against current thoughts and trends in the market. Some of the information will be laughably wrong and naive, and some of the information will still be relevant. The hope is, by regularly reviewing our past, we will better inform our future.
Importance of Assessing Level of Consciousness in Medical Care | The Lifescie...
Culturally-Driven Process Improvement Enabled By Technology
1. Culturally-Driven Process Improvement
Enabled By Technology
Guest Lecture for Health Information Science
HINF 551
University of Victoria
May 2008
Clinical Decision Support
and Data Warehousing
Dale Sanders
312-695-8618
dsanders@nmff.org
2. 2
• Complex, life-critical, time-critical computerized decision support
• It all boils down to managing false positives and false negatives, then
optimizing your intervention and response
My background
US Air Force
Command,
Control,
Communications,
Computers &
Intelligence (C4I)
Officer
TRW/National
Security Agency
• START Treaty
• Nuclear Non-
proliferation
• US nuclear
weapons threat
reduction
Director of Medical
Informatics, LDS
Hospital/Intermountain
Healthcare
CIO,
Northwestern
CIO, Cayman Islands
National Health System
Product
Development,
Health Catalyst
20161983
Reagan/Gorbachev
Summits
Nuclear Warfare
Planning and
Execution–
NEACP &
Looking Glass
3. 3
Acknowledgements & Thanks
Robert Jenders, MD, MS
Associate Professor, Dept of Medicine, Cedars-Sinai Medical
Center & UCLA
Co-chair, HL7 Clinical Decision Support TC & Arden Syntax SIG
R. Matthew Sailors, PhD
Assistant Professor, Dept of Surgery, UT-Houston
Co-chair, HL7 Clinical Decision Support TC & Arden Syntax SIG
Clinical Decision Support and Arden Syntax
4. Overview
• Patient information systems trends & concepts
• Enterprise Data Warehouse (EDW)
– Basic Terms and Concepts
– Case Study Examples
– Intermountain Healthcare
– Northwestern University
• Clinical Decision Support
5. 5
Information Systems:
The Three Perspectives
Transaction Systems:
Collecting data that
supports analytics &
efficient workflow
Analytic Systems:
Aggregating and exposing
data to improve workflow
Knowledge Systems:
Organizing, sharing,
and linking
information
• Query and reporting tools
• Enterprise data warehouses
• Benchmarking data
• Document imaging
• Videoconferencing
• Collaboration tools
• Intranets/Internet access
• Search engines
• EMR’s
• Billing systems
• GL systems
• HR systems
• Scheduling systems
• Inventory management systems
Goal
Measurement
Goal
achievement
Goal
Achievement
Designed to support
6. 6
Patient Information
Systems Trends
Transportability and Interoperability
– Information moves with the patient
Real-time alerts and reminders
– Drug-drug and drug-allergy interactions
Data-driven treatment planning
Disease management at the point-of-care
Payer-driven data collection
– Pay for Performance (P4P)
Quality of care reporting
Transparency of cost is coming
7. 7
Health consumerism movement
– Demands for improved and more transparent
information access
– Demands for more security and privacy
– The “credit report” phenomenon
Computerized patient records
– Legislation and state and federal initiatives are
supporting investment in collaborative software
Regional health information networks are receiving
funding
– For collaborative clinical information sharing and for
pay-for-performance initiatives
Patient Information
Systems Trends
8. 8
Patient Care Data “Customers”
Patient Care Data
Financial
HIS Coding (HDM)
A/R Management
Standard Costing
Materials Management
Case Mix
Clinical
Patient Safety
Clinical Programs
Clinical Support Services
Case Mix
Accreditation/Regulatory
JCAHO, NCQA, HEDIS
HIPAA, EMTALA, OSHA, CLIA
Third-party Payers
Claims information
Utilization management
Case management
9. 9
Meaningful,
maintainable point-of-care
clinical decision support
•Registration
•Scheduling
•Accts Receivable
•Patient/payer billing
•Reporting
•HIPAA claims,
eligibility, remittance
•Benefit plan tracking
•Co-pay tracking
•Referral management
•COB
•Risk management
•Patient education
•Encounter
documentation
•Charge capture
•Diagnostic coding
•ePrescribing
•Allergy alerts
•D-D interactions
•Medical history
•Messaging & real time
collaboration
•Patient portal
• Self-scheduling
• Self-registration
• Account management
• Results & history
• Rx refills
• Credit card payment
•Lab interfaces
•Payer/clearinghouse interfaces (HIPAA)
•Integrated orders
•Integrated results
•ePrescribing
•Patient education
•Clinical references within context
•Affiliated referring partners
Business Intelligence/”Pay for Performance” Metrics
Workflow & Handoff Between Clinical and Business Processes
Core
Best Practices Reminders Meaningful Alerts
Advantage Differentiator Off The Edge
Regional/External Entities
Functional Framework:
Electronic Health Record
Leading Edge
• Rare & difficult
• The next frontier
10. The Future EHR User Interface
• Patient specific data
– Much like current EHRs
– “Tell me about this patient.”
• Disease management data
– “Tell me about managing patients like this.”
• Treatment options data
– “Tell me about my options for treating this patient.”
– “Tell me about the most common tests and medications ordered for patients like this.”
• Cost of care data
– “Tell me about how much these treatment options cost.”
• Clinical outcomes data
– “Tell me how satisfied patients were with these treatment options.”
10
14. 14
Multiple, Collaborative Organizations
EDW
A single data perspective
on the patient care process
Diagnostic systems
•Lab System
•Radiology
•Imaging
•Pathology
•Cardiology
•Others
Diagnosis
Registration &
Scheduling
Patient
Perception
Orders &
Procedures
Results &
Outcomes
Billing &
Accounts
Receivable
Claims
Processing
Encounter
Documentation
•ADT System
•Master Patient Index
Pharmacy Electronic
Medical Record
Surveys•Diagnostics
•Pharmacy
Billing and AR
System
Claims Processing
System
Diagnostic systems
•Lab System
•Radiology
•Imaging
•Pathology
•Cardiology
•Others
Diagnostic systems
•Lab System
•Radiology
•Imaging
•Pathology
•Cardiology
•Others
Diagnosis
Registration &
Scheduling
Patient
Perception
Orders &
Procedures
Results &
Outcomes
Billing &
Accounts
Receivable
Claims
Processing
Encounter
Documentation
•ADT System
•Master Patient Index
Pharmacy Electronic
Medical Record
Surveys•Diagnostics
•Pharmacy
Billing and AR
System
Claims Processing
System
Diagnosis
Registration &
Scheduling
Patient
Perception
Orders &
Procedures
Results &
Outcomes
Billing &
Accounts
Receivable
Claims
Processing
Encounter
Documentation
•ADT System
•Master Patient Index
•ADT System
•Master Patient Index
PharmacyPharmacy Electronic
Medical Record
Electronic
Medical Record
SurveysSurveys•Diagnostics
•Pharmacy
•Diagnostics
•Pharmacy
Billing and AR
System
Billing and AR
System
Claims Processing
System
Claims Processing
System
Diagnostic systems
•Lab System
•Radiology
•Imaging
•Pathology
•Cardiology
•Others
Diagnosis
Registration &
Scheduling
Patient
Perception
Orders &
Procedures
Results &
Outcomes
Billing &
Accounts
Receivable
Claims
Processing
Encounter
Documentation
•ADT System
•Master Patient Index
Pharmacy Electronic
Medical Record
Surveys•Diagnostics
•Pharmacy
Billing and AR
System
Claims Processing
System
Diagnostic systems
•Lab System
•Radiology
•Imaging
•Pathology
•Cardiology
•Others
Diagnostic systems
•Lab System
•Radiology
•Imaging
•Pathology
•Cardiology
•Others
Diagnosis
Registration &
Scheduling
Patient
Perception
Orders &
Procedures
Results &
Outcomes
Billing &
Accounts
Receivable
Claims
Processing
Encounter
Documentation
•ADT System
•Master Patient Index
Pharmacy Electronic
Medical Record
Surveys•Diagnostics
•Pharmacy
Billing and AR
System
Claims Processing
System
Diagnosis
Registration &
Scheduling
Patient
Perception
Orders &
Procedures
Results &
Outcomes
Billing &
Accounts
Receivable
Claims
Processing
Encounter
Documentation
•ADT System
•Master Patient Index
•ADT System
•Master Patient Index
PharmacyPharmacy Electronic
Medical Record
Electronic
Medical Record
SurveysSurveys•Diagnostics
•Pharmacy
•Diagnostics
•Pharmacy
Billing and AR
System
Billing and AR
System
Claims Processing
System
Claims Processing
System
Diagnostic systems
•Lab System
•Radiology
•Imaging
•Pathology
•Cardiology
•Others
Diagnosis
Registration &
Scheduling
Patient
Perception
Orders &
Procedures
Results &
Outcomes
Billing &
Accounts
Receivable
Claims
Processing
Encounter
Documentation
•ADT System
•Master Patient Index
Pharmacy Electronic
Medical Record
Surveys•Diagnostics
•Pharmacy
Billing and AR
System
Claims Processing
System
Diagnostic systems
•Lab System
•Radiology
•Imaging
•Pathology
•Cardiology
•Others
Diagnostic systems
•Lab System
•Radiology
•Imaging
•Pathology
•Cardiology
•Others
Diagnosis
Registration &
Scheduling
Patient
Perception
Orders &
Procedures
Results &
Outcomes
Billing &
Accounts
Receivable
Claims
Processing
Encounter
Documentation
•ADT System
•Master Patient Index
Pharmacy Electronic
Medical Record
Surveys•Diagnostics
•Pharmacy
Billing and AR
System
Claims Processing
System
Diagnosis
Registration &
Scheduling
Patient
Perception
Orders &
Procedures
Results &
Outcomes
Billing &
Accounts
Receivable
Claims
Processing
Encounter
Documentation
•ADT System
•Master Patient Index
•ADT System
•Master Patient Index
PharmacyPharmacy Electronic
Medical Record
Electronic
Medical Record
SurveysSurveys•Diagnostics
•Pharmacy
•Diagnostics
•Pharmacy
Billing and AR
System
Billing and AR
System
Claims Processing
System
Claims Processing
System
Hospital X
Hospital Y
Physician Office Z
15. Sanders’ Hierarchy of Analytic Maturity
• Basic business reporting
– Financial and Human Resources
• Legal compliance reporting
– As required by state and federal law
– Cancer Registry, mortality rates, et al
• Professional accreditation reporting
– Joint Commission, Society of Thoracic Surgeons, et al
• Quality of care reporting
– Physician Quality Reporting Initiative, Leap Frog, et al
• Patient Relationship Management (PRM)
– Borrowing from Customer Relationship Management in retail
– Tailored to the entire context of the patient
– Simpler, faster patient satisfaction and outcomes feedback
– Clinical “Loose Ends”
• Real-time analytic fusion
– Blending patient specific data with general patient type data
– “Other physicians who saw patients like this, ordered these medications and tests.”
15
Increasing Maturity
16. Healthcare Analytics Adoption Model
Level 8
Personalized Medicine
& Prescriptive Analytics
Tailoring patient care based on population outcomes and
genetic data. Fee-for-quality rewards health maintenance.
Level 7
Clinical Risk Intervention
& Predictive Analytics
Organizational processes for intervention are supported
with predictive risk models. Fee-for-quality includes fixed
per capita payment.
Level 6
Population Health Management
& Suggestive Analytics
Tailoring patient care based upon population metrics. Fee-
for-quality includes bundled per case payment.
Level 5 Waste & Care Variability Reduction
Reducing variability in care processes. Focusing on
internal optimization and waste reduction.
Level 4 Automated External Reporting
Efficient, consistent production of reports & adaptability to
changing requirements.
Level 3 Automated Internal Reporting
Efficient, consistent production of reports & widespread
availability in the organization.
Level 2
Standardized Vocabulary
& Patient Registries
Relating and organizing the core data content.
Level 1 Enterprise Data Warehouse Collecting and integrating the core data content.
Level 0 Fragmented Point Solutions
Inefficient, inconsistent versions of the truth. Cumbersome
internal and external reporting.
16
18. 18
Examples of Clinical Goals
• Decrease the total number of
nulliparous elective inductions with
a Bishop Score <10 by 50%
• Keep the variable cost increase of
deliveries without complications
resulting in normal newborns to
5.73% for 2003
• For all adult patients with diabetes,
increase the percent of patients with
LDL less than 100 to >=45.5%.
(Currently 44.5%)
• Measured glucose values will be
between 60 and 155 mg/dl 80% of
the time for all ICU patients
• 100% compliance to post-surgery
radiation therapy protocols for
breast cancer cases with >4
positive nodes and tumor size
>=5cm
• Compliance with the timing of
administration of Pre-surgical
Prophylactic Antibiotic Usage will
exceed 91%
• For patients being treated for
depression, increase the
percentage continuing on
prescribed antidepressant for 6
months after filling first prescription
to >=44.6%
25. 25
Structured vs. Unstructured Data
Representation of Human
Experience & Knowledge
ComputableAnalyticValue
• Text
• Video
• Recorded
Audio
• Structured,
discrete data
• Face-to-Face
Audio
27. 27
Case Study
• Primary Care: Diabetes
– Motive: Improved long-term management of diabetes patients
– RAND Study 2002: “64% of diabetic patients receive inadequate care.”
– Integrates five disparate data sources
– Lab
– Problem list
– Insurance claims: CPT’s and pharmacy
– In-patient pharmacy
– Hospital ICD-9
– This one hits home
– Winner
– National Exemplary Practice Award 2002
– American Association of Health Plans
28. Measure Goal
HbA1c (test at least 2 times a
year)
<7.0%
Blood Pressure
(check at each office visit)
<130/80 mm
Hg
LDL Cholesterol
(test at least every 2 years)
<100 mg/dL
Triglycerides
(test at least every 2 years)
<150 mg/dL
Foot Exam (perform at least
annually)
normal
Urine Microalbumin/Creatinine
Ratio (test at least annually)
<30
Dilated Eye Exam (check
annually,
or every 2 years if well
controlled)
normal
Diabetes CPM:
Key Indicators
28
36. Case Study
• Labor and Delivery - Elective Inductions
– Continue to educate physicians and patients on the safe
and efficacious practice of elective labor induction.
– To maintain at ≤5% elective deliveries that do not meet
strict criteria (39 weeks gestation) developed by the
Intermountain Obstetrical Development Team.
– To measure clinical outcomes of care and report
quarterly by provider.
36
40. Data Loaded to Date
Metric Value
Number of Rows 3,173,632,200
Terabytes 2.2
Truckloads 1,233
Complete works of Shakespeare 252,483
41. 41
Early Adopters and Value of the EDW
Customer Analytic Use
NUgene Relating genomic data and clinical profiles for phenotyping high risk
diseases such as diabetes and asthma
Neurosurgery A summary of new patients, encounters and diagnoses from the
EDW is import daily into MDAnalyze, a Neurosurgery outcomes
database
Alan Peaceman, MD Creation of a perinatal patient registry for studying clinical quality
outcomes; BMI relationships to complications
Bill Grobman, MD Statistics of deliveries at NMH in preparation for a grant proposal
Dana Gossett, MD Application of Systemic Inflammatory Response Syndrome (SIRS)
criteria to pregnant and postpartum women with infectious
complications
Andrew Naidech, MD First adopter of the Research Patient Data Aggregator for use in
research and clinical quality assessment of subarachnoid
hemorrhage, intracerebral hemorrhage, and stroke patients
NMH Process Improvement A DMAIC project aimed at improving the quality of care for patients
seen with bone fractures at NMH. Used the EDW to help narrow
and speed their search for bone fracture patients using a query of
text-based Radiology reports.
42. 42
Specific Research Example
For the last year for the women who deliver, provide…
• mean age and standard deviation
• percent between 18-34, inclusive
• ethnic breakdown, at least by white, black, latino
• % smokers
• % singletons (i.e. no twins or triplets)
• % who receive their prenatal care with an NMH doc
• mean BMI and standard deviation
• % BMI < 19
• % BMI 19 - 29.9
• % BMI > 29.9
• % who start prenatal care in the first trimester
Rapid turnaround (<2 days) to meet a grant submission deadline…
43. 43
Other Examples
• How many patients were prescribed an NSAID and who also had a lab
value which indicated reduced renal function (lab result of GFR < 50 or
Creatinine > 1.5)?
– Answer: 725 out of 16214 in calendar year 2007
• What percentage of patients diagnosed with multiple myeloma in
remission over age 18 were prescribed bisphosphonates in the past 12
months?
– Answer: 18%
• How many patients who have had 1 or more low LVEF (<40) measurements in
our outpatient echo system (Xcelera) and who have received a low LVEF
measurement within the last 180 days and who have not seen one of the
following doctors in a Northwestern clinic office visit within the last 120 days?
– 'KADISH, ALAN H.'
– 'GOLDBERGER, JEFFREY J.'
– 'PASSMAN, ROD S.'
– 'DENES, PABLO'
– 'JACOBSON, JASON‘
– Answer: 309
44. Changes in quality measures during the first 3 months of the study
MEASURE Satisfied (%)
Sept 301, 2007
Satisfied (%)
Dec 31, 2007
Satisfied (%)
April 30, 2008
Coronary Heart Disease
Beta blocker in MI 0.89 0.91 0.91
Antiplatelet drug 0.90 0.89 0.91
Lipid lowering drug 0.88 0.88 0.89
ACE inhibitor/ARB in DM or LVSD 0.84 0.84 0.85
Heart Failure
ACE inhibitor/ARB in LVSD 0.86 0.87 0.85
Anticoagulation in atrial fibrillation 0.63 0.64 0.72
Beta blocker in LVSD 0.83 0.84 0.85
Hypertension control 0.76 0.75 0.76
Diabetes Mellitus
Blood pressure management 0.60 0.60 0.63
HbA1c control ( < 8) 0.63 0.65 0.64
LDL control 0.51 0.51 0.52
Aspirin for primary prevention 0.76 0.77 0.83
Nephropathy screening/management 0.81 0.82 0.83
Examples
45. Prevention
Screening mammography 0.79 0.80 0.84
Cervical cancer screening 0.80 0.81 0.80
CRC screening 0.49 0.48 0.47
Pneumococcal vaccination 0.49 0.52 0.54
Osteoporosis screening or
therapy
0.76 0.79 0.82
Changes in quality measures during the first 3 months of the study
MEASURE Satisfied
(%)
Sept
301,
2007
Satisfied
(%)
Dec 31,
2007
Satisfied
(%)
April 30,
2008
49. Why Didn’t the Patient
Follow the Protocol?
• 167 patient reasons for not following advice for
preventive service
– 9 have resulted in patient having service
• 2 patients unable to afford medication
• 14 patients refused medication
– 0 have started medication
50. Why Didn’t the Physician
Follow the Protocol?
• 147 cases in which medical exceptions or modifiers
were given
– 132 appropriate on initial review
– 5 discussed with another reviewer and judged
appropriate
– 4 discussed with another reviewer and judged
inappropriate: feedback given
– 6 reviewed with peer reviewer and expert and
recommended change in management
52. 52
Clinical DSS Structure
Point-of-Care DSS
– Alerts, reminders
Retrospective
– What happened?
Prospective
– What will happen?
53. 53
Where Does It Appear?
Organization of Data
– “checklist effect”
Stand-Alone Expert Systems
– often require redundant data entry
Data Repository: Mining
CDSS Integrated into Workflow
– push information to the clinician at the point
of care
– examples: EMR, CPOE
54. 54
The Revolutions in CDSS
Phase 1: Quality and safety of care
– What is “good care”?
– Did we provide good care?
– Barely entering this phase now
Phase 2: Economics of care
– What does good care cost?
– Did we provide good care at the most effective cost?
Phase 3: Genomics of care
– What are the genomic influences on good care?
– Did we provide personalized, tailored care?
55. 55
Key Architectural Elements
Data capture/display/storage
– EMR
– central data repository
Controlled, structured vocabulary
Knowledge representation (e.g., Arden)
Knowledge acquisition
Clinical event monitor: integrate the pieces
for many different uses (clinical, research,
administrative)
56. 56
Foundation and Rationale for
Decision Support Models
Mathematics, mathematical models and
decision making
Probability and statistics (Bayesian models)
Rule-based decision-making
– IF the patient has symptoms A or B or C
THEN
– Prescribe medication X and treatment Y and
schedule next visit for T weeks
Data-driven models
– Looks for patterns within a test set of data
and then generalize
57. 57
Justification for CDSS:
Medical Errors
Estimated annual mortality:
Air travel deaths 300
AIDS 16,500
Breast cancer 43,000
Highway fatalities 43,500
Preventable medical errors 44,000 -
(1 jet crash/day) 98,000
Costs of Preventable Medical Errors:
$29 billion/year overall
1999 Institute of Medicine (IOM) Report
58. 58
Definitions: What is an error?
Error of execution: Failure of an action to be
completed as planned
Error of planning: Use of a wrong plan to achieve an
aim
Adverse event: An injury caused by medical
management (and not the result of the patient’s
condition)
Preventable adverse event: An adverse event
attributable to error
Negligent adverse event: A preventable adverse event
that satisfies criteria for malpractice
59. 59
Errors in Medicine
Hospital admissions: 2.9% (UT/CO, 1992) -
3.7% (NY, 1984) have an adverse event
Proportion of preventable adverse events: 53%
(CO/UT) - 58% (NY)
Extrapolate to USA (33.6M admissions in
1997): 44,000 - 98,000 deaths
60. 60
Errors in Medicine
Types of adverse events (Harvard
Medical Practice Study, 1991):
– drug complications: 19%
– wound infections: 14%
– technical complications: 13%
50% associated with operations
61. 61
Clinical DSS: The Impact
Examined randomized and nonrandomized
controlled trials that evaluated the effect of a
CDSS compared with care provided without a
CDSS on practitioner performance or patient
outcomes.
CDSS improved practitioner performance in
62 (64%) of the 97 studies
JAMA. 2005;293:1223-1238.
62. 62
Case Studies:
Examples of CDSS Effectiveness
Perioperative Antibiotic Administration
– intervention: reminder re timing and type of abx
– period: 1988 - 1994
– result: perioperative wound infections dec 1.8% ->
0.9%
– avg # doses: 19 -> 5.3
– overall antibiotic cost (constant $) per treated
patient: $123 -> $52
Pestotnik SL, Classen DC, Evans RS, Burke JP. Implementing antibiotic practice
guidelines through computer-assisted decision support: clinical and financial
outcomes. Ann Intern Med 1996;124(10):884-90.
64. 64
Examples (continued)
Reminders of Redundant Test Ordering
– intervention: reminder of recent lab result
– result: reduction in hospital charges (13%)
– Tierney WM, Miller ME, Overhage JM et al. Physician inpatient order writing on
microcomputer workstations. Effects on resource utilization.
JAMA 1993;269(3):379-83.
Preventive Health Reminders in HIV
– intervention: reminders to perform screening tests or
vaccination (e.g., pap smear, HBV)
– result: sig decreased time to documentation (median = 11 vs
52 days)
– Safran C, Rind DM, Davis RB et al. Guidelines for management of HIV infection with
computer-based patient's record. Lancet 1995;346(8971):341-6.
65. 65
Examples (continued)
Systematic review
– 68 studies
– 66% of 65 studies showed benefit on physician
performance
• 9/15 drug dosing
• 1/5 diagnostic aids
• 14/19 preventive care
• 19/26 other
– 6/14 studies showed benefit on patient outcome
Hunt DL, Haynes RB, Hanna SE et al. Effects of computer-based clinical
decision support systems on physician performance and patient outcomes:
a systematic review. JAMA 1998;280(15):1339-46.
66. 66
Other CDSS Success Stories
Point-of-Care Decision Support
– Bilirubin Management in neonates
– Ventilator Management in ARDS
– Coumadin Management
– Glucose Management in the ICU
– Antibiotic Assistant
– Infectious Disease Monitoring
68. 68
Goals of AI
Study the thought processes of humans to
better understand the complexity of
human intelligence
Create computer systems which achieve
human levels of reasoning
69. 69
Knowledge Representation Formalisms:
Their Role
Express policies (institutional, national, international)
in computable format
Formulate interventions in medical practice
Make local variations in guidelines
Provide “intelligence” to a clinical expert system
71. 71
Roots of Medical AI
MYCIN (late 1070s)
– Shortliffe, et al, at Stanford
– 1970s, infectious disease and antibiotic
therapies
– Rules-based
PUFF (early 1980s)
– Based on MYCIN
– Pulmonary data interpretation
72. 72
Roots of Medical AI
APACHE (1981)
– http://www.cerner.com/public/Cerner_3.asp?id=3562
– Point of care in ICU
73. 73
Computers Are Good At…
Computational functions - add, subtract,
multiply, divide, compare
– The most familiar
Symbolic reasoning
Pattern recognition
74. 74
The Arden Syntax
A symbolic language for encoding medical knowledge
Adopted by HL7 and ANSI in 1999
Used to develop Medical Logic Modules (MLMs)
Each MLM can make a single medical decision
– MLMs can be chained
Can be used for variety of clinical decision support
functions
– E.g., alerting physicians of potential kidney failure
75. 75
Arden Syntax: Assessment
Incorporated into several vendors’ products
Growing number of installation sites
Facile for simple alerts/reminders
May not be sufficiently expressive for complex
guidelines
76. 76
Support for Arden Syntax
Institutions
Cedars-Sinai Medical Center
Software Vendors
Eclipsys/Healthvision
McKesson
Siemens
Knowledge Vendors
Micromedex
77. 77
Arden Syntax - History
HELP
LDS Hospital
Salt Lake City, UT
CARE
Regenstrief Institute
Indianapolis, IN
Arden Syntax
1989
78. 78
Arden Syntax - Rationale
Arden Syntax arose from the need to make medical
knowledge available for decision making at the point
of care.
Allow knowledge sharing within and between
institutions
Make medical knowledge and logic explicit
Standardize the way medical knowledge is integrated
into hospital information systems
79. 79
Pattern Recognition
Objects, events or processes are described by their
qualitative features, logical, and computational
relationships
Examples
– Computer matches pattern found in a new x-ray to
other cases to determine diagnosis
– Searching text for context-based key words
• Spam filters
80. 80
Wikipedia
Based on either a priori knowledge or on
statistical information extracted from the
patterns
Sensor
Feature
Extraction
Classification
Engine
Training Set
Real Data
81. 81
Other AI Methods
Genetic algorithms
– Selection, recombination, mutation
Search algorithms
Constraint-based problem solving
– When conditions in variables are met,
then execute
Frame-based reasoning
84. 84
In Summary
Enterprise Data Warehouses and
Electronic Medical Records work hand-
in-hand to address Clinical Decision
Support
Artificial Intelligence has yet to prove
itself scalable beyond informatics
research projects