Enduring Understandings 7-9

1
(No Transcript)
2
Enduring Understandings 7-9

• Explaining associations
• and
• judging causation

3

• EU7 One possible explanation for finding an
  association is that the exposure causes the
  outcome. Because studies are complicated by
  factors not controlled by the observer, other
  explanations also must be considered, including
  confounding, chance, and bias.
• The Not everything that glitters is gold
  Principle

4

• EU8 Judgments about whether an exposure causes a
  disease are developed by examining a body of
  epidemiologic evidence, as well as evidence from
  other scientific disciplines.

5

• EU9 While a given exposure may be necessary to
  cause an outcome, the presence of a single factor
  is seldom sufficient. Most outcomes are caused
  by a combination of exposures that may include
  genetic make-up, behaviors, social, economic, and
  cultural factors and the environment.
• The Just because your friend sleeps in class and
  never fails her courses does not mean that
  sleeping in class does not cause F grades
  Principle

6
Reasons for associations

• Confounding
• E is associated with C and C causes D
• Bias
• F causes D, but we thought F was an E
• Reverse causality
• D causes E
• Sampling error (chance)
• Causation
• E1 ?D
• E1 E2 ? D
• E1 or E2 ? D
• E1 E2 OR E3E4?D

7

• Osteoporosis risk is higher among women who live
  alone.
• Death rates are low in AK and high in FL.
• African American women have higher infant
  mortality than others in the US.

8
Confounding

• Confounding is an alternate explanation for an
  observed association of interest.

Number of persons in the home
Osteoporosis
Age
9
Confounding

• Confounding is an alternate explanation for an
  observed association of interest.

Exposure
Outcome
Confounder
10
Confounding

• YES confounding module example
• Hypothetical cohort study
• 20,000 men followed for 10 yrs
• RQ Are bedsores related to mortality among
  elderly patients with hip fractures?

11
Bedsores and Mortality
D D-
E 79 745 824
E- 286 8290 8576
365 9035 9400
RR (79 / 824) / (286 / 8576) 2.9
12
Bedsores and Mortality

• Avoid bedsoresLive forever!!
• Could there be some other explanation for the
  observed association?

13
Bedsores and mortality

• If severity of medical problems had been the
  reason for the association between bedsores and
  mortality, what might the RR be if all study
  participants had very severe medical problems?
• What about if the participants all had problems
  of very low severity?

14
Bedsores and Mortality
Died Did not die
Bedsores 55 severe 24 not 51 severe 694 not 824
No bedsores 5 severe 281 not 5 severe 8285 not 8576
365 9035 9400
15
Bedsores and Mortality (Severe)
Died Did not die
Bedsores 55 51 106
No bedsores 5 5 10
60 56 116
RR (55 / 106) / (5 / 10) 1.0
16
Bedsores and Mortality (Not severe)
Died Did not die
Bedsores 24 694 718
No bedsores 281 8285 8566
305 8979 9284
RR (24 / 718) / (281 / 8566) 1.0
17
Bedsores and Mortality stratified by Medical
Severity
SEVERE Died Didnt die
Bedsores a b
No sores c d
RR 1.0
SEVERE- Died Didnt die
Bedsores a b
No sores c d
RR 1.0
SEVERE Died Didnt die
Bedsores a b
No sores c d
RR 2.9
SEVERE- Died Didnt die
Bedsores a b
No sores c d
RR 2.9
18
Bedsores

• So.
• Bedsores are unrelated to mortality among those
  with severe problems.
• Bedsores are unrelated to mortality among those
  with problems of less severity.
• .
• the adjusted RR 1, and the unadjusted RR 2.9

19
Confounding

• Confounding is an alternate explanation for an
  observed association of interest.

Bedsores
Death
Severity of medical problems
20
Reasons for associations

• Confounding
• E is associated with C and C causes D
• Bias
• F causes D, but we thought F was an E
• Reverse causality
• D causes E
• Sampling error (chance)
• Causation
• E1 ?D
• E1 E2 ? D
• E1 or E2 ? D
• E1 E2 OR E3E4?D

21
Bias

• Errors are mistakes that are
• randomly distributed
• not expected to impact the MA
• less modifiable
• Biases are mistakes that are
• not randomly distributed
• may impact the MA
• more modifiable

22
Types of bias

• Selection bias
• The process for selecting/keeping subjects causes
  mistakes
• Information bias
• The process for collecting information from the
  subjects causes mistakes

23
Selection bias

• Healthy worker effect
• People who are working are more likely to be
  healthier than non-workers
• Non-response
• People who participate in a study may be
  different from people who do not
• Attrition
• People who drop out of a study may be less
  different from those who stay in the study
• Berksons
• Hospital controls in a case-control study

24
Information bias

• Misclassification, e.g. non-exposed as exposed or
  cases as controls
• Recall bias
• Cases are more likely than controls to recall
  past exposures
• Interviewer bias
• Interviewers probe cases more than controls
  (exposed more than unexposed)

25
Birth defects and diet

• In a study of birth defects, mothers of children
  with and without infantile cataracts are asked
  about dietary habits during pregnancy.

26
Pesticides and cancer mortality

• In a study of the relationship between home
  pesticide use and cancer mortality, controls are
  asked about pesticide use and family members are
  asked about their loved ones usage patterns.

27
Induced abortion breast CA

• Positive association found in 5 studies
• No association found in 6 studies
• Negative association found in 1 study

28
Minimize bias

• Can only be done in the planning and
  implementation phase
• Standardized processes for data collection
• Masking
• Clear, comprehensive case definitions
• Incentives for participation/retention

29
Reasons for associations

• Confounding
• E is associated with C and C causes D
• Bias
• F causes D, but we thought F was an E
• Reverse causality
• D causes E
• Sampling error (chance)
• Causation
• E1 ?D
• E1 E2 ? D
• E1 or E2 ? D
• E1 E2 OR E3E4?D

30
Reverse causality

• Suspected disease actually precedes suspected
  cause
• Pre-clinical disease ? Exposure ? Disease
• For example Memory deficits ? Reading cessation
  ? Alzheimers
• Cross-sectional study
• For example Sexual activity/Marijuana

31
Minimize effect of reverse causality

• Done in the planning and implementation phase of
  a study
• Pick study designs in which exposure is measured
  before disease onset
• Assess disease status with as much accuracy as
  possible

32
Reasons for associations

• Confounding
• E is associated with C and C causes D
• Bias
• F causes D, but we thought F was an E
• Reverse causality
• D causes E
• Sampling error (chance)
• Causation
• E1 ?D
• E1 E2 ? D
• E1 or E2 ? D
• E1 E2 OR E3E4?D

33
Sampling error/chance

• E and D are associated in a sample, but not in
  the population from which the sample was drawn.

34
RR in the population
D D-
E 50 50 100
E- 50 50 100
100 100 200
35
RR in sample1
D D-
E 25 25 50
E- 25 25 50
50 50 100
36
RR in sample2
D D-
E 20 30 50
E- 30 20 50
50 50 100
37
RR in sample3
D D-
E 30 20 50
E- 15 35 50
45 55 100
38
Reasons for associations

• Confounding
• E is associated with C and C causes D
• Bias
• F causes D, but we thought F was an E
• Reverse causality
• D causes E
• Sampling error (chance)
• Causation
• E1 ?D
• E1 E2 ? D
• E1 or E2 ? D
• E1 E2 OR E3E4?D

39
Causal pathways

• Necessary, sufficientrare, if at all
• Not necessary, sufficientalso rare
• Necessary, not sufficientTB
• Not necessary, not sufficient--Most causes fall
  into this category--heart disease, obesity

40
Reasons for associations

• Confounding
• E is associated with C and C causes D
• Bias
• F causes D, but we thought F was an E
• Reverse causality
• D causes E
• Sampling error (chance)
• Causation
• E1 ?D
• E1 E2 ? D
• E1 or E2 ? D
• E1 E2 OR E3E4?D

41
The process of assessing causality

• Observe patterns
• Generate hypothesis
• Design study to test hypothesis
• Conduct study
• Interpret the resultsthe big question is did the
  exposure cause the disease?
• Are there alternate non-causal explanations for
  the results we found?
• If not, then is this the whole story?

42
So, what should we do?

• Goal is to understand causality
• Use guidelines to help us make sense of the
  evidence

43
Key Guidelines

• Temporality a necessary condition
• Consistency
• Dose-response
• Consideration of alternate explanations
• Coherence

44
Enduring Understandings

• 7, 8, and 9

45

• EU7 One possible explanation for finding an
  association is that the exposure causes the
  outcome. Because studies are complicated by
  factors not controlled by the observer, other
  explanations also must be considered, including
  confounding, chance, and bias.
• The Not everything that glitters is gold
  Principle

46

• EU8 Judgments about whether an exposure causes a
  disease are developed by examining a body of
  epidemiologic evidence, as well as evidence from
  other scientific disciplines.

47

• EU9 While a given exposure may be necessary to
  cause an outcome, the presence of a single factor
  is seldom sufficient. Most outcomes are caused
  by a combination of exposures that may include
  genetic make-up, behaviors, social, economic, and
  cultural factors and the environment.
• The Just because your friend sleeps in class and
  never fails her courses does not mean that
  sleeping in class does not cause F grades
  Principle

48
(No Transcript)
49
Possible Explanations for Finding an Association
1.
Cause
2.
Confounding
3.
Reverse Time Order
Chance
4.
5.
Bias
50
Possible Explanations for Finding an Association
Cause
A factor that produces a change in another factor.
William A. Oleckno, Essential Epidemiology
Principles and Applications, Waveland Press, 2002.
51
Sample of 100
52
Sample of 100, 25 are Sick
53
Types of Causal Relationships
Diagram
2x2 Table
DZ
DZ
X
a
b
c
d
X
54
Types of Causal Relationships
Diagram
2x2 Table
DZ
DZ
X
a
b
c
d
X
55
Handout
56
Necessary and Sufficient
Diagram
2x2 Table
DZ
DZ
X
X
a
b
c
d
X
57
Necessary but Not Sufficient
Diagram
2x2 Table
DZ
DZ
X
X
a
b
c
d
X
58
Not Necessary but Sufficient
Diagram
2x2 Table
DZ
DZ
X
X
X
a
b
c
d
X
59
Not Necessary and Not Sufficient
Diagram
2x2 Table
DZ
DZ
X
X
a
b
c
d
X
60
Lack of fitness and physical activity causes
heart attacks.
a bc d
61
Lack of supervision of small children causes lead
poisoning.
a bc d
62
Is the association causal?
63
Ties, Links, Relationships, and Associations
Suicide Higher in Areas with Guns
Family Meals Are Good for
Mental Health
1.
Cause
Study Concludes Movies Influence
Youth Smoking
Study Links Iron
Deficiency to Math
Scores
2.
Confounding
Lack of High School Diploma Tied to
US Death Rate
Study Links Spanking
to Aggression
3.
Reverse Time Order
Chance
4.
Depressed Teens More Likely to Smoke
Snacks Key to Kids TV- Linked Obesity China
Study
5.
Bias
Pollution Linked with Birth Defects in US Study
Kids Who Watch R-Rated Movies More Likely to
Drink, Smoke
64
(No Transcript)
65
Possible Explanations for Finding an Association
1.
Cause
2.
Confounding
3.
Reverse Time Order
Chance
4.
5.
Bias
66
Possible Explanations for Finding an Association
Population
All the people in a particular group.
67
Possible Explanations for Finding an Association
Sample
A selection of people from a population.
68
Possible Explanations for Finding an Association
Inference
Process of predicting from what is observed in a
sample to what is not observed in a population.
To generalize back to the source population.
69
Inference
Population
Sample
Process of predicting from what is observed
to what is not observed.
70
Population
Deck of 100 cards
71
Population
72
Population
Total
a
b


c
d
73
Population
Population
Total


74
Population
Total



Total
75
Population


Total
Risk
25 / 50 or 50
25 / 50 or 50
76
Population


Total
Relative Risk
25 / 50 or 50
50
____
25 / 50 or 50
50
77
Population
78
Possible Explanations for Finding an Association
Chance
To occur accidentally.
To occur without design.
A coincidence.
79
Chance
80
Chance
81
Sample
Sample of 20 cards
82
Sample
Sample of 20 cards
Total
83
Sample
Sample of 20 cards
Total
5 / 10 or 50
5 / 10 or 50
84
Sample
Sample of 20 cards
Total
Risk
5 / 10 or 50
50
____
5 / 10 or 50
50
85
Sample
CDC
By Chance
Total
___

86
Chance
How many students picked a sample with 5 people
in each cell?
No Marijuana
No Marijuana
Total
Risk
Relative Risk
10
5
5
5 / 10 or 50
Odd
50
____
10
5
5
5 / 10 or 50
50
Even
By Chance
87
Ties, Links, Relationships, and Associations
Association is not necessarily causation.
Suicide Higher in Areas with Guns
Family Meals Are Good for
Mental Health
1.
Cause
Study Concludes Movies Influence
Youth Smoking
Study Links Iron
Deficiency to Math
Scores
2.
Confounding
Lack of High School Diploma Tied to
US Death Rate
Study Links Spanking
to Aggression
3.
Reverse Time Order
Chance
4.
Depressed Teens More Likely to Smoke
Snacks Key to Kids TV- Linked Obesity China
Study
5.
Bias
Kids Who Watch R-Rated Movies More Likely to
Drink, Smoke
88
(No Transcript)
89
An Association TV and Aggressive Acts
90
Worksheet
91
the study of the distribution and determinants
of health-related states or events
92
Study Designs
Experimental Studies
Randomized Controlled TrialsOther Experimental
Studies
Observational Studies
Cohort StudiesCase-Control StudiesCross-Sectio
nal StudiesEcologic Studies
Cohort Studies
93
Cohort Study

• A study in which a group of people is followed
  over time
• The group is made up of people who have the
  exposure of interest and people who do not have
  the exposure of interest
• Exposed and unexposed people are followed over
  time to determine whether they experience the
  outcome

94
Exposure - Outcome
When epidemiologists ask a question,
it is often of the form Does
______________ cause ______________?
(exposure)
(outcome)
95
Exposure - Outcome
When epidemiologists ask a question,
it is often of the form Does
______________ cause ______________?
(exposure)
(outcome)
For example
Do diesel exhaust fumes from school buses cause
asthma? Does eating chocolate cause acne? Are
males at higher risk of automobile
accidents? Does immunization with the measles
vaccine prevent
measles? Does acupuncture result in pain relief?
96
Cohort Study Flow Diagram
Cohort
Time
A designated group of persons
who are followed or
traced over a period of time
97
Cohort Study Flow Diagram
Adolescents Young Adults
By age 22
At age 14
A designated group of persons
who are followed or
traced over a period of time
98
Express it in Numbers
At age 14
Watched TV
gt 1 hour per day
154 reported
aggressive acts
465 did not report
aggressive acts
99
Express it in Numbers
At age 14
Watched TV
gt 1 hour per day
154 reported
aggressive acts
465 did not report
aggressive acts
100
Express it in Numbers
At age 14
Watched TV
gt 1 hour per day
154 reported
aggressive acts
465 did not report
aggressive acts
No Aggressive Acts
No Outcome
Aggressive Acts
Outcome
Total
Total
Watched TV
gt 1 hour per
day
Exposed
154
465
619
101
Risk
154
154

(154 465)
619
No Aggressive Acts
No Outcome
Aggressive Acts
Outcome
Total
Total
Watched TV
gt 1 hour per
day
Exposed
154
465
619
102
Hypothesis
An educated guess
An unproven idea,
based on observation or reasoning,
that can be proven or disproven
through investigation
Watching TV causes aggressive acts.
103
Does watching TV cause aggressive acts?
154
154


24.9
(154 465)
619
No Aggressive Acts
No Outcome
Aggressive Acts
Outcome
Total
Risk
Total
Watched TV
gt 1 hour per
day
Exposed
154
465
619
24.9
104
Does watching TV cause aggressive acts?
24.9 risk of committing
an aggressive act
Watching TV for gt 1 hrs per day
Adolescents Young Adults
? risk of committing
an aggressive act
Watching TV for lt 1 hr per day
By 22 years
At 14 years
105
Does watching TV cause aggressive acts?
24.9 risk of committing
an aggressive act
Watching TV for gt 1 hrs per day
Adolescents Young Adults
? risk of committing
an aggressive act
Watching TV for lt 1 hr per day
Comparison Group
By 22 years
At 14 years
106
Comparison Group
5 reported
aggressive acts
83 did not report
aggressive acts
No Aggressive Acts
No Outcome
Aggressive Acts
Outcome
Total
Risk
Total
Watched TV
gt 1 hour per
day
Exposed
154
465
619
24.9
107
Comparison Group
At age 14
By age 22
5 reported
aggressive acts
83 did not report
aggressive acts
Watched TV
lt 1 hour per day
No Aggressive Acts
Aggressive Acts
Outcome
Total
Risk
Total
Watched TV
gt 1 hour per
day
Exposed
154
465
619
24.9
5
83
88
5.7
108
Contingency Table
No Aggressive Acts
No Outcome
Aggressive Acts
Outcome
Total
Risk
Total
Watched TV
gt 1 hour per
day
Exposed
154
465
619
24.9
5
83
88
5.7
109
Does watching TV cause aggressive acts?
No Aggressive Acts
No Outcome
Aggressive Acts
Outcome
Total
Risk
Total
Watched TV
gt 1 hour per
day
Exposed
154
465
619
24.9
5
83
88
5.7
110
Does watching TV cause aggressive acts?
No Aggressive Acts
No Outcome
Aggressive Acts
Outcome
Total
Risk
Total
Watched TV
gt 1 hour per
day
Exposed
154
465
619
24.9
4.4
5
83
88
5.7
Compared to those who watched TV for lt 1 hour per
day, those who watched TV for gt 1
hours per day were ____ times as likely to commit
aggressive acts.
111
Relative Risk
A way of quantifying the relationship between two
risks
Tells us the number of times one risk is larger
or smaller than another
Cartoon from Larry Gotnicks The Cartoon Guide to
Statistics, HarperPerennial, 1993
112
the control of health problems
What should be done?
No Aggressive Acts
Relative Risk
No Outcome
Aggressive Acts
Outcome
Total
Risk
Total
Watched TV
gt 1 hour per
day
Exposed
154
465
619
24.9
4.4
5
83
88
5.7
113
Association
When things turn up together
114
Confounding
Another Exposure
Drinking Alcoholic Beverages
Association
Cause
Association of Interest
When an observed association between
an exposure and an
outcome is distorted because
the exposure of interest is associated with
some other
exposure that causes the outcome
115
Confounding

• Confounding is the distortion of an
  exposure-outcome association brought about by the
  association of another factor with both outcome
  and exposure.
• A confounder confuses our conclusions about the
  relationship between an exposure and an outcome.

116
the control of health problems
X
Another Exposure
Drinking Alcoholic Beverages
Association
Cause
X
Association of Interest
117
Association
When things turn up together
118
Confounding
Aggressive Acts
Watching TV
Association of Interest
When an observed association between
an exposure and an
outcome is distorted because
the exposure of interest is associated with
some other
exposure that causes the outcome
119
Confounding
Living in a Violent Neighborhood
Aggressive Acts
Watching TV
Association of Interest
When an observed association between
an exposure and an
outcome is distorted because
the exposure of interest is associated with
some other
exposure that causes the outcome
120
Confounding
Lack of Adequate Supervision
Aggressive Acts
Watching TV
Association of Interest
When an observed association between
an exposure and an
outcome is distorted because
the exposure of interest is associated with
some other
exposure that causes the outcome
121
the control of health problems
X
Lack of Adequate Supervision
X
Aggressive Acts
Watching TV
Association of Interest
When an observed association between
an exposure and an
outcome is distorted because
the exposure of interest is associated with
some other
exposure that causes the outcome
122
Assessment
In a study of the hypothesis that drinking orange
juice prevents the flu, 3,000 students at Wright
High School, who did not have the flu on December
31, 2000, were followed from January 1 through
March 31, 2001. By the end of the study, among
the 1000 students who drank orange juice, 123
students had developed the flu. Among the 2000
students who did not drink orange juice, 342
students had developed the flu. Display the
above data on a 2x2 table, calculate risks of
flu, calculate the relative risk, and explain
whether or not the results support the hypothesis
that drinking orange juice prevents the flu.
123
123
124
Explaining Associations and Judging Causation
Does evidence from an aggregate of studies
support a cause-effect relationship?
Guilt or Innocence?
Causal or Not Causal?
124
Teach Epidemiology
125
Explaining Associations and Judging Causation
Sir Austin Bradford Hill
The Environment and Disease
Association or Causation?
Proceedings of the Royal Society of Medicine
January 14, 1965
Teach Epidemiology
126
Explaining Associations and Judging Causation
In what circumstances can we pass
from this observed association
to a verdict of causation?
126
Teach Epidemiology
127
Explaining Associations and Judging Causation
Here then are nine different viewpoints
from all of which we should study association
before we cry causation.
127
Teach Epidemiology
128
Explaining Associations and Judging Causation
Does evidence from an aggregate of studies
support a
cause-effect relationship?
  1.   What is the strength of the association
between the risk factor and the disease? 2.  
Can a biological gradient be demonstrated? 3.  
Is the finding consistent? Has it been
replicated by others in other places? 4.   Have
studies established that the risk factor precedes
the disease? 5.   Is the risk factor associated
with one disease or many different
diseases? 6.   Is the new finding coherent with
earlier knowledge about the risk factor and the
m disease? 7.   Are the implications of the
observed findings biological sensible? 8.   Is
there experimental evidence, in humans or
animals, in which the disease has m been
produced by controlled administration of the risk
factor?
Teach Epidemiology
129
Explaining Associations and Judging Causation
Stress causes ulcers.
Helicobacter pylori causes ulcers.
Teach Epidemiology
130
Explaining Associations and Judging Causation
Teach Epidemiology
131
Explaining Associations and Judging Causation
Teach Epidemiology
132
(No Transcript)
133
In the News

• Assemble into three-person teams
• Select an article
• Use the article to create a lesson plan to teach
  one or more of the Enduring Understandings to a
  specified class for 30 minutes
• Teach the lesson
• Specify the student population and course
• Engage us as though we were the students
• Help us to understand what you did to generate
  the lesson plan

Teach Epidemiology
134
Article Choices

• Early childhood behavior and substance use
• Huffing and suicide
• Soft drinks and diabetes
• Circumcision and AIDS
• Prenatal smoking and attention deficit
• ADHD among girls
• Traffic and childhood asthma
• Breast-feeding and childhood obesity
• Depression and sexual risk-taking
• Family stress and childhood illness
• ADHD medications and mortality

Teach Epidemiology
135
(No Transcript)
136
Teaching Epidemiology Rules

• Teach epidemiology.
• As a group, create a 30-minute lesson during
  which we will develop a deeper understanding of
  an enduring epidemiological understanding.
• Focus on the portion of the unit that is
  assigned. Use that portion of the unit as the
  starting point for creating your 30-minute
  lesson.
• When teaching, assume the foundational
  epidemiological knowledge from the preceding days
  of the workshop.
• Try to get us to uncover the enduring
  epidemiological understanding. Try to only tell
  us something when absolutely necessary.
• End each lesson by placing it in the context of
  the appropriate enduring epidemiological
  understanding.
• Teach epidemiology.
• Metacognition--After the lesson, reflect on your
  preparation for and teaching of the lesson.

136
Teach Epidemiology
137
Teaching Epidemiology
Metacognition
They can then use that ability to think about
their own thinking to grasp
how other people might learn.
They know what
has to come first,
and they can
distinguish between foundational concepts
and elaborations or
illustrations of those ideas. They realize
where people are likely to face
difficulties developing
their own comprehension,
and
they can use that understanding
to
simplify and clarify complex topics for others,
tell the right story, or raise a powerfully
provocative question. Ken Bain, What the Best
College Teachers Do
Teach Epidemiology
138
Teaching Epidemiology
To create a professional community

that discusses new teacher materials and
strategies and
that supports the risk taking and struggle
entailed in
transforming practice.
Teach Epidemiology
139
Teaching Epidemiology
Group Assignments
Births Class 1, p. 6-12 War Qs
11-21 Case-control Class 1, p.
16-21 Confounding p. 32-36 Bias p. 25-29 and
30-32 Alpine Fizz Procs 2, 4, 5
139
Teach Epidemiology