2. Outline
Decision Process
What Causes Poor Decisions?
The Decision Level & Decision Milieu
Decision Making under Uncertainty
Payoff Table
Methods
Decision Making under Risk
EVM, EOL, EVPI
Decision Tree Approach
Rollback Procedure
What if Payoff Values are Cost?
Decision Analysis with Non-Monetary Values & Multiple-
Attributes
Dominance
Minimum Attribute Satisfaction
Most Important Attribute
3. Identification of the problem and its nature
Specification of objectives and decision criteria
Development of alternatives
Analysis and comparison of alternatives
Selection of the best alternative
Implementation of the choice
Controlling and monitoring the results
The Decision Process
4. What Causes Poor Decisions?
Mistakes in the Decision Process
Bounded Rationality: is the limits imposed on decision
making by costs, human abilities and errors, time,
technology, and the tractability of data.
Suboptimization: is characterized by Decisions are often
departmentalized as separate organizational units
compete for scarce resources. Individual departments
often seek solutions that benefit their own department,
but not necessarily the healthcare organization as a
whole.
5. Decision Theory
represents a general approach to decision
making which is suitable for a wide range of
operations management decisions,
including: capacity, service design, location
planning, equipment selection, etc.
6. The Decision Level & Decision Milieu
Certainty-- Known values for parameters of interest
Uncertainty-- Impossible to assess the likelihood of
various possible future events
Risk-- Certain parameters have probabilistic
outcomes
7. Certainty
Certainty rarely exists, especially in
health care decisions. But if it does,
simply choose the best available option
(highest profit/least cost).
8. Decision Making under Uncertainty
Maximin-- best of the worst (pessimist)
Maximax-- best of the best (optimist)
Hurwitz-- allows you to adjust the
probabilities/weighing between maximin and
maximax or pessimist vs. optimist
Laplace-- best average payoff
Minimax Regret-- best of the worst regrets
10. Example 3.1: A major imaging center is not able to meet the
increased demand from patients for MRIs. The administration is willing
to explore the possibilities by evaluating such alternatives as adding
one or two additional units or out sourcing to other image centers and
earning a commission of $30.00 per MRI.
A feasibility analysis showed that three major demand chunks
could occur in the future, summarized as 500, 750 and 1000
additional MRI requests. The financial analysis of the potential
business summarizes profits/losses under additional MRI demand
chunks in a payoff table shown in Table below.
Alternatives
500 Cases 750 Cases 1000 Cases
Buy One MRI Unit -15* 200 300
Buy Two MRI Units -150 100 725
Outsource 15 22.5 40
* in $ USD
11. Maximin Solution
* in $ USD
Alternatives 500 Cases 750 Cases 1000 Cases Worst
Buy One MRI Unit -15* 200 300 -15
Buy Two MRI Units -150 100 725 -150
Outsource 15 22.5 40 15
12. Maximax Solution
* in $ USD
Alternatives 500 Cases 750 Cases 1000 Cases Best
Buy One MRI Unit -15* 200 300 300
Buy Two MRI Units -150 100 725 725
Outsource 15 22.5 40 40
13. Hurwitz Solution
For optimism with α = 0.5.
Then the HV value for the three alternatives would be:
HV (Buy one MRI unit) = .5(300,000)+(.5)(-15,000) = 142,500.
HV (Buy two MRI units)= .5(725,000)+(.5)(-150,000)= 287,500.
HV (Outsource) = .5(40,000)+(.5)(15,000) = 27,500.
α HV Decision Alternative
1.0 725,000* Buy Two MRI Units
.5 287,500 Buy Two MRI Units
.4 200,000 Buy Two MRI Units
.3 112.500 Buy Two MRI Units
.24 60,600 Buy One MRI Unit
.2 48,000 Buy One MRI Unit
.1 17,500 Outsource
0 15,000 Outsource
14. Opportunity Losses (Regrets)
* in $ USD
Alternatives 500 Cases 750 Cases 1000 Cases Worst
Buy One MRI Unit 30* 0 425 425
Buy Two MRI Units 165 100 0 165
Outsource 0 177.5 685 685
15. Laplace Strategy
* in $ USD
Probability 1/3 1/3 1/3 Expected
Value
Alternatives 500 Cases 750 Cases 1000 Cases
Buy One MRI Units -15* 200 300 161.67
Buy Two MRI Units -150 100 725 225
Outsource 15 22.5 40 25.89
16. Expected Value Model
Once the healthcare manager has assessed the probability
distribution, computation of the expected values for each
alternative is straightforward, as follows:
EMV(Ai) = Σj pj Oij
17. Payoff Table for EMV
* in $ USD
Probability .2 .6 .2 Expected
Value
Alternatives 500 Cases 750 Cases 1000 Cases
Buy One MRI Unit -15* 200 300 177
Buy Two MRI Units -150 100 725 175
Outsource 15 22.5 40 24.5
18. Expected Opportunity Loss
The probabilities can also be incorporated into the regrets (or
opportunity losses) calculated earlier. In this way the healthcare
manager can assess the expected losses and try to minimize
them with proper decision. Calculations of expected opportunity
loss follow the formula:
EOL(Ai) = Σj pj Rij
19. Expected Opportunity Loss
* in $ USD
Probability .2 .6 .2 Expected
Opportunity
Loss
Alternative 500
Cases
750
Cases
1000
Cases
Buy One MRI Unit 30* 0 425 91
Buy Two MRI Units 165 100 0 93
Outsource 0 177.5 685 243.5
20. Expected Value of Perfect Information (EVPI)
EVUC = Σj pj (Best Oij given Sj)
EVPI = EVUC-EMV
Probability .2 .6 .2
Alternatives 500 Cases 750 Cases 1000 Cases
Buy One MRI Unit -15* 200 300
Buy Two MRI Units -150 100 725
Outsource 15 22.5 40
* in $ USD
21. Expected Value of Perfect Information (EVPI)
EVUC = Σj pj (Best Oij given Sj)
EVUC = (.2*15000) + (.6*200000) + (.2*725000) = 268000.
EVPI = EVUC-EMV
EMV = $177,000
EVPI = $268,000 – $177,000 = $91,000
22. What if Payoffs are Costs?
Alternatives 500 Cases 750 Cases 1000 Cases
Buy One MRI Unit 2,050* 2,075 2,100
Buy Two MRI Units 4,050 4,075 4,100
Outsource 5 10 15
* in $ USD
23. Regret Table Using Costs
Alternatives 500 Cases 750 Cases 1000 Cases
Buy One MRI Unit 2,050-5=2,045* 2,075-10=2,065 2,100-15=2,085
Buy Two MRI Units 4,050-5=4,045 4,075-10=4,065 4,100-15=4,085
Outsource 5-5=0 10-10=0 15-15=0
* in $ USD
24. Decision Tools-- The Decision Tree
Decision
Node
Event
Node
Event
Node
Event
Node
Actions
Action A
Action B
Action C
Outcomes
Events
1
2
3
1
2
3
1
2
3
Outcome 1
Outcome 4
Outcome 7
Outcome 2
Outcome 5
Outcome 8
Outcome 3
Outcome 6
Outcome 9
25. 177
177
175
24.5
Buy Two MRI Units
750 Cases, p=.6
750 Cases, p=.6
750 Cases, p=.6
-$15*
$300
$200
-$150
$100
$725
$15
$22.5
$40
║
* in $USD
Analysis of the Decision Tree: Rollback Procedure
26. Multi-attribute Decisions
Dominance Procedure: compares a pair of alternatives
attribute by attribute.
Minimum Attribute Satisfaction Procedure: satisfactory
levels are set for each alternative
Most Important Attribute Procedure: attributes are
ranked in order of importance
Combination: combines two or more of the above
procedures.
27. Decision Analysis with Non-Monetary Values
and Multiple Attributes
Attributes*
Alternative
Cardinal McKesson Owens &
Minor
Importance
Ranking
Minimum
Acceptable
Level
Availability 7 7 7 1 >= 7
Reliability of IT
Technology
7 5 7 2 >= 6
Quality of
Products
8 9 8 3 >= 7
Cost in $000 per
year
23,749 24,195 23,688 5 <=25,000
On Time Delivery 97% 95% 97% 4 >=95%
*Attributes are scored on a 1-10 scale (with the exception of those associated with costs
and on-time-delivery percentage), score of 10 being most favorable.