Decision%20Analysis

1
Decision Analysis
2
Real Case

• 63 year old housewife with 6 grown children.
• 10 year history of stable angina pectoris.
• 8 years ago she had total hip replacement.
• Had post operative pulmonary embolism but
  recovered and was pain free and fully mobile 1
  year later.

3

• For the last 12 months she had experienced
  increasing pain in her hip on weight bearing.
• Is now mostly confined to a wheelchair.
• 8 months ago she had an uncomplicated non-Q wave
  anterior wall MI. Her stable angina limits her
  ability to get about on crutches.
• Orthopedic surgeon reviewed the case and
  concluded she most likely had loosening of the
  femoral component.
• He spelled out the risks and benefits of
  reoperation

4
If only the femoral component needed replacement

• Chances of that are 65.
• The probability of a good result would be 60.
  (Probability of poor result 40).
• Operative mortality is 10.

5
If only the acetabular component needed
replacement

• Chances of that are 25.
• The probability of a good result would be 80.
  (Probability of poor result 20).
• Operative mortality is 5.

6
If both components needed replacement

• Chances of that are 10.
• The probability of a good result would be 45.
  (Probability of poor result 55).
• Operative mortality is 15.

7
If no surgery

• She would remain the same (chances of that 20).
• Or she would get worse and become permanently
  confined to her wheelchair (80).

8
Debate ensued within the care team

• Some advocated surgery.
• Others recommended against it.
• Moreover, although they wanted to involve the
  patient, they feared they would sway her by the
  fashion in which they presented the facts to her.

9
The debate was occurring

• Despite agreement about the risks and benefits.
• This suggested that a decision analysis may
  provide
• An objective resolution.
• Would allow the team to involve the patient
  without unfairly swaying her decision.

10
Step 1Create a decision tree

• A map of all the courses of action and all their
  consequences.
• Boxes are decision nodes.
• Circles are chance nodes.
• Ovals are end nodes.

11
(No Transcript)
12
Step 2 Assign probabilities

• To all the branches all the way to end.

13
(No Transcript)
14
Step 3Assign utilities

• To all the potential outcomes (the end nodes)
• First rank the outcomes from best to worst.
• Assign numerical values to them. This is where it
  gets tough, not because its artificial (for we
  always assign utilities when dealing with
  patients) but because it forces us to be explicit.

15
Standard reference gamble

• To assign utility to an intermediate outcome we
  offer a gamble between the two extremes versus
  the guaranteed intermediate.

16
Utilities

• Ranking
• Good result 1.0
• Same as now 0.4
• Poor result 0.25
• Worse 0.2
• Death 0

17
(No Transcript)
18
Step 4Fold back

• Multiply utilities by probabilities.
• The utility of a chance node is the weighted sum
  of its branches.
• The utility of a decision node (if there are any
  intermediate ones) is that of the highest branch.

19
(No Transcript)
20
The Verdict

• Operate 0.66 (0.81, 0.63, 0.50).
• Do not operate 0.24.

21
Step 5Pick the decision

• That has the highest utility.

22
Step 6Sensitivity analysis

• Test the decision for its vulnerability to
  clinically sensible changes in probabilities and
  utilities.

23
Sensitivity analysis

• After the patient pondered with family she
  revised utilities upward for current status and
  future without surgery.
• That did not reverse the preference for surgery.
• Some team members objected to some utilities.
• All adjustments made the decision to operate
  stronger.

24
Introduction

• Aortic stenosis often requires the replacement of
  the aortic valve with a prosthesis(AVR).
• Both short-term survival and long-term survival
  are influenced by the presence or absence of CAD,
  which may be asymptomatic.
• The prevalence of occult CAD in women in this age
  and race is 0.067.

25
CABG

• Coronary artery bypass graft (CABG), performed at
  the same time as the AVR improves the survival,
  both long-term and short term, of patients with
  CAD requiring AVR.
• Of course, CABG is an operative procedure that
  increases both the peri-operative mortality and
  the cost for a patient without CAD, and
  accordingly, one wishes to avoid it in a patient
  with NO CAD.

26
Cath

• Catheterization is the "gold standard" for the
  diagnosis of CAD. However, it too has a mortality
  associated with it, (.001 in a cohort of
  unselected patient but .002 in patients with
  aortic stenosis) and should be avoided if this
  can be done without increasing the estimated
  expected risk to the patient.

27
DTH

• In an attempt to improve the overall management
  strategy, a technique known as dipyridamole
  thallium imaging has been devised. This technique
  involves the administration of dipyridamole
  thallium, accompanied by hand-grip exercise (DTH
  test).
• Images of the heart are obtained, and the degree
  of perfusion is observed. This procedure is
  essentially free of mortality or serious
  side-effects.

28
DTH

• In populations with coronary artery disease, 85
  out of 100 patients will have a positive test
  (sensitivity .85).
• In patients without coronary artery disease, 14
  out of 100 patients will have a positive test
  (specificity .86).

29
Operative Mortality for AVR replacement

• Patients with CAD receiving CABG...............
  0.061
• Patients with CAD not receiving CABG.........
  0.122
• Patients without CAD..............................
  ........ 0.028

30
The Case

• A 63 year old white woman with symptomatic AS
  requires an aortic valve replacement. She has no
  angina or evident signs of CAD.
• Potential Management Strategies.
• 1.
• Perform AVR in the AS patient, with no CABG. No
  catheterization is performed in this strategy.

31
2. Catheterize this patient

• and if CAD is shown, perform a CABG and AVR.
  Several consequences
• Those patients whose occult coronary artery
  disease is detected will receive the benefit of
  longer long-term survival by virtue of CABG.
• All patients in this group are also subject to
  the additional risk of the catheterization itself.

32
3. Carry out the dipyridamole thallium handgrip
test (DTH).

• If DTH positive perform catheterization, and if
  that shows CAD, then perform CABG and AVR. If
  catheterization negative, only perform AVR.
• If DTH negative, only perform AVR.

33
Utilities

• No coronary artery disease (CAD).....13.3 years
• Coronary artery disease present - No CABG
  performed.....6.17 years
• Coronary artery disease present - Simultaneous
  CABG.....8.8 years

34
The assignment is

• 1. Make a decision tree, incorporating the 3
  strategies named above and the associated
  probabilities of each outcome.Label each node,
  associated probabilities and utilities.
• 2. Fold back the tree(s)showing the folded back
  values at each node.

35
Assignment cont.

• What is the preferred management strategy, based
  solely on expected longest length of survival as
  the utility.
• What is the estimated number of person-months of
  additional life that the preferred strategy
  provides over the second best strategy?
• For the preferred strategy, carry out a
  sensitivity analysis with respect to the
  prevalence of CAD versus the expected survival.

36
(No Transcript)
37
(No Transcript)
38
(No Transcript)
39
(No Transcript)
40
(No Transcript)
41
(No Transcript)
42
(No Transcript)
43
(No Transcript)
44
(No Transcript)
45

• At any prevalence of CAD below .050, it would be
  preferable to do a DTH test, and catheterize or
  not based on the results of that test. At
  prevalences above that threshold, catheterization
  of all candidates for AVR is preferred

46
(No Transcript)
47
(No Transcript)