DECISION MAKING

Suppose your patient (from the Brazilian rainforest) has

tested positive for a rare but serious disease. Treatment
exists but is risky and costly.
Therefore, it is important to make sure the disease is
present prior to giving treatment.

Previous studies have shown that Disease A occurs in 10 out

of 10,000 persons within the Brazilian rainforest. The
test for the disease is positive in 90% of patients with the
disease. The test is negative in 90% of the patients
without the disease.

In light of this information what is the probability that your

patient has Disease A?
1
DECISION MAKING

Much of the early decision-making work (e.g., from

economics) assumed that the decision-maker will take
all the relevant factors into account and make the best
decision possible.

However, research on decision making shows that

people are far from optimal decision makers.

Today, we will examine a variety of ways in which

people do not give optimal decisions (including in
medical settings) and try to understand what they are
doing and why.
2
 DECISION MAKING
(a sampling)

(I) Cognitive Heuristics and Decision Making

(II) Other Factors Leading to Cognitive Bias

* Framing, Confirmation bias, overconfidence

(III) Probabilistic Reasoning in Medicine

* Applications of Bayes Theorem

3
(I) Cognitive Heuristics & Decision Making

Problem: Too much information

Cognitive heuristics - often effective rules of thumb

that are used to simplify decision making & reduce
memory load.

Used frequently by “experts” (including physicians) &

“novices”

Can be effective in some cases and very misleading in other

cases.
4
Availability Heuristic

(A) Availability - refers to the tendency to make a judgment on

the basis of what can easily be brought to mind.

* Convenient way to estimate prior probability, in that more frequent events

are usually recalled more easily.

* Availability heuristic works well as long as the retrieval process is unbiased.

BUT … many factors can influence accessibility of information & consequently

judgments.
** recency ** salience ** simplicity

Examples: are there more English words that begin with t or k?

are there more English words that end in “_n_” or “ing”?

5
Availability and estimates of deaths

 How do people
estimate how
frequently deaths
occur?
 Influenced by
availability

6
(II) Other Factors Leading to Cognitive Bias

(A) ONLY for students with last names beginning A-L

Imagine that the U.S. is preparing for the outbreak of an
unusual Asian disease, which is expected to kill 600
people. Two alternative programs to combat the disease
have been proposed. Assume that the exact scientific
estimate of the consequences of the programs are as
follows:
If Program A is adopted, 200 people will be saved.
If Program B is adopted, there will be a 1/3 probability that
600 people will be saved and a 2/3 probability that no
people will be saved.
Which of these two programs would you favor?

7
(II) Other Factors Leading to Cognitive Bias

(A) ONLY for students with last names beginning M-Z

Imagine that the U.S. is preparing for the outbreak of an
unusual Asian disease, which is expected to kill 600
people. Two alternative programs to combat the disease
have been proposed. Assume that the exact scientific
estimate of the consequences of the programs are as
follows:
If Program C is adopted 400 people will die.
If Program D is adopted there is a 1/3 probability that
nobody will die and a 2/3 probability that 600 people will
die.
Which of these two program would you favor?

8
(II) Other Factors Leading to Cognitive Bias
(A) Framing effects - The same information presented in
different forms can lead to different decisions.

* Equivalent gains and losses should have the same effects

on selection of an action.
BUT … they do not.

• Choices involving gains (lives saved) are usually risk

averse. Program A favored over B.
• Choices involving losses (people dead) are usually risk
taking. Program D favored over C.

Evidence that physician (and patient) judgments of preference

for treatments affected by whether the treatment outcomes
were described in terms of the probability of living or the
probability of dying. (McNeil et al., 1982) 9
(B) Confirmation Bias - tendency to seek out information that
could only confirm our ideas.

* Produces an inertia which favors the initial hypothesis.

* Leads to the avoidance of information or tests which could
disconfirm (e.g. negative evidence) our hypothesis.
** for example, looking for the lack of a symptom to
disconfirm a hypothesis
* Occurs with scientists & nonscientists alike

10
(C) Overconfidence in judgment - both novices & experts appear to
be more confident in their judgments than is objectively justifiable.
Why? Partly due to lack of feedback concerning outcomes

11
(D) Hindsight bias - tendency for people to think after the fact that
they would have known something before the fact when in
actuality they would not.

Arkes et al. (1981) -- physicians given medical history and asked to assign
probability of each of 4 diseases, given this history.
Group A -- assigns probabilities
Group B -- told correct diagnosis and then assigns probabilities, but told to
just use medical history, as if they had not been given other information.

RESULT: Group B gives 2-3 times more weight to correct diagnosis than
Group A. Cannot ignore diagnosis. (Maybe leads to more overconfidence?)

12
(III) Probabilistic Reasoning in Medicine
Suppose your patient (from the Brazilian rainforest) has tested positive for a
rare but serious disease. Treatment exists but is risky and costly.
Therefore, it is important to make sure the disease is present prior to giving
treatment.

Previous studies have shown that Disease A occurs in 10 out of 10,000

persons within the Brazilian rainforest. The test for the disease is positive in
90% of patients with the disease. The test is negative in 90% of the patients
without the disease.

In light of this information what is the probability that your patient has
Disease A?
Disease
Present (true) Absent (false)
Positive a b
Test
Result
Negative c d
13
 Disease
Present Absent
Positive a b
Test
Result Negative c d

Sensitivity = a
Specificity = d
a+c
b+d

Base Rate = a+c

(of disease a+b+c+d
In population)
Probability = a
(Disease / a+b
Positive test) 14
Bayes theorem - formula for revising beliefs in light of new
information. New odds for some event are:
(1) the old odds (base rate) multiplied by
(2) odds associated with the new information
P (Disease / Positive Test) =

Base Rate of Disease x Sensitivity of Test

Overall P of a Positive Test
OR
a
a+c
P (D / PT) = a+c x a+b
a+b+c+d a+b+c+d

OR …… or more simply ……positive predictive value

P (D / PT) = a
a+b 15
To Return to the Brazilian Rain Forest ………….

Disease
Present Absent
Test a b
Result Positive 9 999
c d
Negative 1 8991

P ( Disease / Positive Test ) = a =

a + b
9 ≈ 1%
9 + 999

16
SO ……….. Bayes’ Theorem takes account of:

* Base rate information (disease prevalence)

* Diagnosticity of tests (sensitivity / specificity)

Bayes’ Theorem can also be used to:

* Integrate information from multiple tests

* Estimate the probability of one disease versus another

Most important - Take home message

Provides a way to think about importance of disease prevalence

and sensitivity / specificity of tests in estimating the
diagnostic utility of a test. -- USE TABLES 17
 Decision Making

 People far from perfect decision-makers.

 Limited by information available, amount of
information that can be processed, and ability
to combine information.
 However, often do very well.
 Important to realize when our judgments may
be biased so we can try to take bias into
account or rely on additional help.

18
 Readings and Key Terms

 Bernstein et al., Chapter 8, 288-291

 Cognitive heuristics Availability

 Framing effects
 Bias: overconfidence, hindsight,
confirmation
 Bayes’ Theorem -- positive predictive
value
19