Clinical Decision Support Systems
HIMA 4160
Fall 2009
Outline

Definitions

Methodologies

Applications

Probabilistic reasoning

Decision tree
2
CDSS

Providing clinicians or patients with
clinical knowledge and patientrelated information, intelligently
filtered or presented at appropriate
times, to enhance patient care
• NOT just physicians …
• Not just rules and alerts …
• (NOT just computer-based …)
3
Categories

Generating alerts and reminders

Diagnostic assistance

Therapy critiquing and planning

Image recognition and interpretation

And many others …
4
Need for CDSS



Limited resources - increased demand
Need for systems that can improve health
care processes and their outcomes in this
scenario
The marriage of medical and technological
advances - to produce a child called Frugal
Efficiency?
5
Generalized Structure
Knowledge
Base
Inference
Engine
6
Knowledge base, inference
engine, and interface
7
Application Areas
8
Workflow Opportunities
9
Possible Disadvantages of
CDSS




Changing relation between patient and the
physician
Limiting professionals’ possibilities for
independent problem solving
Legal implications - with whom does the
onus of responsibility lie?
Information fatigue
10
Issues for success or failure

Evaluation of User Needs

Top management support

Commitment of expert

Integration Issues

Human Computer Interface

Incorporation of domain knowledge

Consideration of social and organizational context 11of
the CDS
Evaluation of Clinical Decision
Support Systems


Criteria for success of CDSS
Aspects for consideration during
evaluation
12
Criteria for a clinically useful
DSS









Knowledge based on best evidence
Knowledge fully covers problem
Knowledge can be updated
Data actively used drawn from existing
sources
Performance validated rigorously
System improves clinical practice
Clinician is in control
The system is easy to use
The decisions made are transparent
13
Aspects for Evaluation of a
CDSS





The clinician need that the CDSS is
intended to address
The process used to develop the system
The system’s intrinsic structure
Evidence of accuracy, generality and clinical
effectiveness
The impact of the resource on patients and
other aspects of the health care
environment
14
Methodology
Major Use
Key developments
Information Retrieval
Finding information, answering
questions
Taxonomies, ontologies, textbased methods, automatic
invocation
Evaluation of logical
conditions
Alerts, reminders, constraints,
inference system
Decision tables, event-conditionaction-rules, production rules
Probabilistic and data
driven classification or
prediction
Diagnosis, technology
assessment, treatment
selection, classification and
prediction, prognosis estimation,
evidence-based medicine
Bayes theorem, decision theory,
ROC analysis, data mining,
logistic regression, artificial
neural networks, belief networks,
meta-analysis.
Heuristic modeling
and export systems
Diagnostic and therapeutic
reasoning, capturing nuances of
human expertise
Rule-based systems, framebased reasoning
Calculations,
algorithms and
multistep processes
Execution of computational
processes, flow-chart-based
guideline and consultations,
interactive dialogue control,
biomedical image and signal
processing
Process flow and workflow
modeling, guideline formalisms
and modeling languages
Associative groupings
of elements
Structured data entry,
structured reports, order sets,
other specialized presentations
and data views
Report generators and document
construction tools, document
architectures, templates, markup
languages, ontology tools, 15
ontology languages
Computerized
Physician/Provider
Order Entry
The Two Sides of Errors
• 44,000+ hospital deaths due to
medical error
• 50 adverse events/1000
outpatient pt-years (Gurwitz
2003)
• Patients receive 55% of
recommended care (McGlynn,
2003)
17
Our Solution to Safety
physician
nurse
pharmacis
t
Bedside
team
BMJ 2000;320:768–70
18
What is CPOE?


Computer application
which replaces
traditional paper order
sheets
Care / computerized
provider is a key part
of the name
19
Key Advantages to CPOE




Data aggregated for clinical use
Clinician can interact with medical
record away from the bedside
Immediate routing of orders and
requisitions to ancillary departments
Smart prompts and checks can
enhance safety and quality of care
21
Important OE work
1. El Camino Hospital, 1971
First clinician order entry system, (TDS)
2. Warner, Pryor, Clayton, Gardner, et al
HELP System, LDS Hospital, 1970++ (3M)
3. McDonald, Tierney et al, 1974++
Regenstrief order entry / reminders / (~SMS)
4. Glaser, Teich, Bates, Kuperman et al, 1994++
Brigham & Women’s order entry (~Eclipsys)
22
Commercial Order Entry (80s)
23
CPOE Integration
Decision
support
Pharmacy
Terminology
Lab
System
Data Server
Or Interface
External
Knowledge
Sources
ADT
CPOE System
EHR
(documents)+
Internal
Knowledge
Sources
24
WizOrder Main Screen Layout: Simple, fixed format: functionally oriented, designed with users
1) Active orders
2) Common useful
orders based on
patient location
3) What to do next in WizOrder
4) Buttons for
commonly used
features
Physician enters order for antibiotic,
Gentamicin, by partially typing its name
Copyright (C) 2003 Vanderbilt University Medical
Center
25
26
Time
Savings:
New
method
for
summarizing
“active”
orders &
current
information
“What you
need to know
about patient”
printed on one
piece of paper
Copyright (C) 2003 Vanderbilt University Medical Center
Active orders
Recent
Labs
27
28
29
Issues

People

Process
30
Bayesian Network
31
Review of Probability

P(A) = p, P(not A) = 1 – p

P(A, B) = P (A | B)* P(B)


P(A, B) = P (A | B) * P(B) = P(A) *
P(B)
P(A) =  ( A | B )
B
32
Probability


Frequentist
Bayesian
33
Bayes’ Theorem
Likelihood
P (h / e) 
Prior
P (e / h ) P (h )
P (e)
Posterior
Probability
of Evidence
Probability of an hypothesis, h, can be updated when evidence, e, has
been obtained.
34
A Simple Example
Consider two related variables:
1. Disease (D) with values y or n
2. Test (T) with values +ve or –ve
And suppose we have the following probabilities:
P(D = y) = 0.001
P(T = +ve | D = y) = 0.8
P(T = +ve | D = n) = 0.01
These probabilities are sufficient to define a joint probability distribution.
Suppose an athlete tests positive. What is the probability that he has
the disease?
P(D  y|T   ve)



P (T   ve | D  y ) P ( D  y )
P (T   ve | D  y ) P ( D  y )  P (T   ve | D  n ) P ( D  n )
0 .8  0 .001
0 .8  0 .001  0 .01  0 .999
0 . 074
35
Sensitivity, Specificity,
Prevalence and Probabilities
Consider two related variables:
1. Disease (D) with values y or n
2. Test (T) with values +ve or –ve
And suppose we have the following probabilities:
P(D = y) = 0.001 (Prevalence)
P(T = +ve | D = y) = 0.8 (Sensitivity)
P(T = +ve | D = n) = 0.01(1-specificity)
These probabilities are sufficient to define a joint probability distribution.
Suppose an athlete tests positive. What is the probability that he has
taken the drug?
P(D  y|T   ve)



P (T   ve | D  y ) P ( D  y )
P (T   ve | D  y ) P ( D  y )  P (T   ve | D  n ) P ( D  n )
0 .8  0 .001
0 .8  0 .001  0 .01  0 .999
0 . 074
36
Bayesian Network Demo
Decision Tree
Descargar

Slide 1