Lecture 8 clinical decision support systems

1
Lecture 8 clinical decision support systems
2
What are they and how do I know if they are any
good?

• To introduce the major dimensions of computerised
  clinical decision support systems (CDSSs)
• Suggested appraisal criteria for CDSS

3
A scenario

• Chief nurse in a PCT.
• audit results reveal significant variability,
  exception reports and low satisfaction around
  guidelines for monitoring and managing warfarin
  therapy.
• Whether specialist clinics and CDSS might be a
  better approach?
• Find a paper (Fitzmaurice 2000)

4
Some assumptions (humans and decisions)

• Decisions are choices
• Types of decisions merit types of research
  evidence
• All require the combination of information and
  injection of context to make knowledge
• Combining information is difficult and error
  prone.

5
Some assumptions (CDSS)

• Designed around the correct kind of knowledge for
  the problem faced
• Must take into account natural variations in
  patients and so must work with individual
  profiles and data
• Must offer tailored advice and actually inform
  decision making

6
Kinds of decision support

• AUDIENCE Public, professional and embedded
• Functionalist
• Information management
• Focussing attention
• Patient specific consultations
• Clinical role (Dx, Tx)
• Architectural approach

7
CDSS architecture
Decision models
Quantitative
qualitative
Neural networks
Bayesian, Fuzzy sets
Truth tables, Boolean logic
Decision trees
Expert systems, reasoning models
8
CDSS quantitative
Decision Model Decision Model Decision Model Decision Model
Truth -
TP FN 100
- FP TN 100
9
ROC
10
hypothyroidism
T4 Hypothyroid Euthyroid
5 or less 18 1
5.1 7 7 17
7.1 9 4 36
9 or more 3 39
totals 32 93
11
Decision thresholds
1 T4 of 5 or less
2 T4 of 7 or less
T4 Hypothyroid Euthyroid
5 or less 18 A 1 C
gt5 14 B 92 D
Totals 32 93
T4 Hypothyroid Euthyroid
7 or less 25 18
gt7 7 75
Totals 32 93
3 T4 of 9 or less
Impact
T4 Hypothyroid Euthyroid
9 or less 29 54
gt9 3 39
Totals 32 93
Cut point Sensitivity Specificity
5 0.56 0.99
7 0.78 0.81
9 0.91 0.42
12
roc
Cut point TP FP
5 0.56 0.01
7 0.78 0.19
9 0.91 0.58
13
interpretation

• .90-1 excellent (A)
• .80-.90 good (B)
• .70-.80 fair (C)
• .60-.70 poor (D)
• .50-.60 fail (F)

14
Qualitative approaches

• Symbolic rule based reasoning (?Boolean logic)
• Truth tables
• Flowcharts or algorythms
• Expert systems
• Forward driven reasoning
• Backwards reasoning

15
Truth tables
Rule E1 E2 E3 E4 E10 Dx
1 T F T T F VT
2 . . . 9 T T T T F d F d T T NR SR
T true F false D dont care
E1 all RR intervals are regular E2 All QRS
complexes are identical E3 QRS complexes longer
than 120 msec E4 P waves present E10 PR
intervals regular
16
algorithmic
start
E1
F
T
E2
F
VT
F
F
T
E3
E10
E4
T
F
T
T
17
Knowledge based systems
Practice guidelines Secondary or primary
Evidence base
Knowledge Base
Inference Methods
Patient Database
Knowledge Acquisition
explication
Electronic Health Records
NB may be called expert systems in Older
literature
18
(No Transcript)
19

20
(No Transcript)
21
(No Transcript)
22
(No Transcript)
23
(No Transcript)
24
(No Transcript)
25
http//www.isabel.org.uk/private/diagnostic/drsnav
.htm
26
How might CDSS help innovation?
Stage Barrier Possible Benefit
Predispose Clinicians dont know about innovation Might help as learning tool
(staff unwilling to change) Apathy CDSS might attract clinical interest/generate discussion
Peer resistance DSS might help in marketing
Patient resistance DSS may promote innovation to patients
were too busy DSS enables NP to take on new tasks freeing up medical time
Conflicting financial interest Unlikely to help
27
ENABLING
Stage Barrier Possible Benefit
Enable innovation (staff willing system is against them) Patient data not complete, poorly presented DSS issues problem-specific checklist or reminder to record relevant data, preinterprets complex patient data, carries out automatic case finding
Poor access to detailed knowledge DSS acts as intelligent front-end to literature, filtering knowledge according to current patient and problem
Clinicians find it hard to synthesise patient data and knowledge DSS carries out complex calculation or logical reasoning to link relevant patient data and clinical knowledge
Lack of skills DSS might help as a learning or simulation tool
Lack of space, drugs, money etc Unlikely to help
28
reinforcing
Stage Barrier Possible Benefit
Reinforce innovation (staff need encouragement) Forgetting Reminders for clinicians
Mistakes caused by action slips, capture errors Reminders and alerts to build a sfae operating environment preinterpreted patient data problem-specific workflow and record formats to lessen errors
Diminished motivation over time Reminders or alerts DSS can help support others (e.g. nurse practitioners) to carry out routine tasks
29
Critical appraisal of CDSS - validity

• Were study participants randomised
• If not, did the investigators demonstrate
  similarity in all known determinants of prognosis
  or adjust for differences in analysis?
• Was the control group uninfluenced by the CDSS?
• Were interventions similar in the two groups ?
• Was outcome assessed uniformly in the
  experimental and control groups?

30
Critical appraisal of CDSS results and
application

• What is the effect of the CDSS?
• What elements of the CDSS are required
• Is the CDSS exportable to a new site?
• Is the CDSS decision support system likely to be
  accepted in your setting?
• Do the benefits of the CDSS justify the risks and
  costs?