Day 4. Teach Epidemiology. Professional Development Workshop. Centers for Disease Control and Prevention Global Health Odyssey Museum Tom Harkin Global Communications Center June 610, 2011. Teach Epidemiology. Teach Epidemiology.
Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.
4
Teach Epidemiology
Professional Development Workshop
Centers for Disease Control and PreventionGlobal Health Odyssey MuseumTom Harkin Global Communications Center June 610, 2011
Teach Epidemiology
Teach Epidemiology
http://www.cdc.gov/
8:15 AM
Teach Epidemiology
Teachers TeamTeaching Teachers (TTTT)
Teach Epidemiology
Enduring Epidemiological Understandings
Knowledge that “… is connected and organized, and … ‘conditionalized’ to specify the context in which it is applicable.”
National Research Council , Learning and Understanding
Teach Epidemiology
Making Group Comparisons and Identifying Associations
The goal of every epidemiological study is to harvest valid and precise information about the relationship between an exposure and a disease in a population.
The various study designs merely represent different ways of harvesting this information.
Essentials in Epidemiology in Public Health
Ann Aschengrau and George R. Seage III
Teach Epidemiology
9:00 AM
Teach Epidemiology
P – First people living near ports, then further inland.
P – Europe
T –1340’s1350’s
huron2.aaps.k12.mi.us
*Activity!
*Hypotheses???
1. Bacon fat?? What kind of study?
2. Genotypes?? What kind of study?
“A previously healthy 36 yearold man with clinically diagnosed CMV infection in September 1980 was seen in April of 1981 because of a 4month history of fever, dyspnea and cough. On admission, he was found to have P. carinii pneumonia, oral candidiasis, and CMV retinitis. A complementfixation CMV titer in April 1981 was 1928. The patient has been treated with 2 short courses of TMP/SMX that have been limited because of a sulfainduced neutropenia. He is being treated for candidiasis with topical nystatin” –MMWR 1981
P – healthy homosexual males, intravenous drug users
P – World wide (including U.S.)
T – Early 1980’s
*Maybe related to black plague resistance.
*Maybe related to HIV resistance
*What kind of study?
Enduring Epidemiological Understandings
Knowledge that “… is connected and organized, and … ‘conditionalized’ to specify the context in which it is applicable.”
National Research Council , Learning and Understanding
Teach Epidemiology
Making Group Comparisons and Identifying Associations
The goal of every epidemiological study is to harvest valid and precise information about the relationship between an exposure and a disease in a population.
The various study designs merely represent different ways of harvesting this information.
Essentials in Epidemiology in Public Health
Ann Aschengrau and George R. Seage III
Teach Epidemiology
9:45 AM
Teach Epidemiology
EPI501
Marian R Passannante, PhD
Associate Professor
University of Medicine and Dentistry of New Jersey
New Jersey Medical School
School of Public Health
Enduring Epidemiological Understandings
Chance
Confounding
Bias
Real
46
Teach Epidemiology
Enduring Epidemiological Understandings
First , choose a statistical method to test the association between an exposure and an outcome.
47
Teach Epidemiology
Enduring Epidemiological Understandings
48
Teach Epidemiology
Enduring Epidemiological Understandings
First , choose a statistical method to test the association between an exposure and an outcome.
Exposure: Statin use
status
Outcome: Colorectal
Cancer
status
49
Teach Epidemiology
Enduring Epidemiological Understandings
First , choose a statistical method to test the association between an exposure and an outcome.
“background
Statins are… effective lipidlowering agents. Statins inhibit the growth of coloncancer cell lines, and secondary analyses of some, but not all, clinical trials suggest that they reduce the risk of colorectal cancer.”
50
Teach Epidemiology
Enduring Epidemiological Understandings
First , choose a statistical method to test the association between an exposure and an outcome.
“methods
The Molecular Epidemiology of Colorectal Cancer study is a populationbased case–control study of patients who received a diagnosis of colorectal cancer in northern Israel between 1998 and 2004 and controls matched according to age, sex, clinic, and ethnic group. We used a structured interview to determine the use of statins in the two groups and verified selfreported statin use by examining prescription records in a subgroup of patients for whom prescription records were available."
51
Teach Epidemiology
Enduring Epidemiological Understandings
First , choose a statistical method to test the association between an exposure and an outcome.
Exposure: Statin use status
Outcome: Colorectal
Cancer status
Statistical method: Chisquare
To test whether the proportion of cases who took statins differs from the proportion of controls who took statins.
52
Teach Epidemiology
Enduring Epidemiological Understandings
Second, choose a level of risk (called a or the alpha level) that we are willing to take when conducting that test.
Most common alpha levels: 0.05 or 0.01
Third, conduct the statistical test and get a p value.
53
Teach Epidemiology
O = observed # of events in a cell
E = expected # of events in a cell if there were
no association between the independent and dependent variables
Expected Value = (Row Total x Column Total) Grand Total
Enduring Epidemiological Understandings
Chisquare (c2) test54
Teach Epidemiology
Enduring Epidemiological Understandings
If the percentage of cases who took statins is the same as the percentage of controls who took statins, what % would
you expect to find among cases and controls?
Exposure: Statin use status
Outcome: Colorectal
Cancer status
Statistical method: Chisquare
55
Teach Epidemiology
Enduring Epidemiological Understandings
If the percentage of cases who took statins is the same as the percentage of controls who took statins, what % would
you expect to find among cases and controls?
Exposure: Statin use status
Outcome: Colorectal
Cancer status
Statistical method: Chisquare
56
Teach Epidemiology
Enduring Epidemiological Understandings
If the percentage of cases who took statins is the same as the percentage of controls who took statins, what % would
you expect to find among cases and controls?
Exposure: Statin use status
Outcome: Colorectal
Cancer status
Statistical method: Chisquare
57
Teach Epidemiology
Enduring Epidemiological Understandings
If the percentage of cases who took statins is the same as the percentage of controls who took statins, what % would
you expect to find among cases and controls?
Exposure: Statin use status
Outcome: Colorectal
Cancer status
Statistical method: Chisquare
58
Teach Epidemiology
Enduring Epidemiological Understandings
Expected Values appear in red below:
Exposure: Statin use status
Outcome: Colorectal
Cancer status
Statistical method: Chisquare
c2 = S [(0E)2/E]
59
Teach Epidemiology
Enduring Epidemiological Understandings
Expected Values appear in red below:
Exposure: Statin use status
Outcome: Colorectal
Cancer status
Statistical method: Chisquare
To test whether the proportion of cases who took statins differs from the proportion of controls who took statins, chisquare compares what was observed and what one would expect to find if there were no association. The chisquare test result will provide a p value. We will compare our p value to our preset level of risk or alpha level.
60
Teach Epidemiology
Enduring Epidemiological Understandings
What does a p value tell us?
If you conducted a statistical test and got a p value of .04 it would mean that 4 times out of 100 you might find a significant association like the one observed in your study by chance alone.
Significance testing provides a p value and this value tell you how likely the result you found is due to chance alone.
Let’s look at the analysis results for the statin study….
62
Teach Epidemiology
Enduring Epidemiological Understandings
Third, we conduct the statistical test and get a p value.
63
Teach Epidemiology
Enduring Epidemiological Understandings
Third, we conduct the statistical test and get a p value.
If the p value is less than alpha
(say .05) we say that our
exposure and outcome are
significantly associated.
With p <.0001, we can
say that there is a strong association between statin
use and disease outcome. How would you describe the direction of the relationship?
64
Teach Epidemiology
Enduring Epidemiological Understandings
How would you measure the strength of the association between statin use and colorectal cancer in this
casecontrol study?
65
Teach Epidemiology
Enduring Epidemiological Understandings
How would you describe the relationship?
Which measure of risk would you calculate?
The odds ratio:
120 x 1781 = .50
234 x 1833
“In analyses including 1953 patients with colorectal cancer and 2015 controls, the use of statins for at least five years (vs. the nonuse of statins) was associated with a significantly reduced … risk of colorectal cancer (odds ratio, 0.50; 95 percent confidence interval, 0.40 to 0.63).”
66
Teach Epidemiology
Enduring Epidemiological Understandings
Although you may have found a statistically significant association, it is always possible that your sample is not really a good representation of the total population and the result you got was by chance alone.
How to avoid this problem?
67
Teach Epidemiology
Enduring Epidemiological Understandings
Although you may have found a statistically significant association, it is always possible that your sample is not really a good representation of the total population and the result you got was by chance alone.
How to avoid this problem?
Increase your sample size
68
Teach Epidemiology
Enduring Epidemiological Understandings
Chance
Confounding
Bias
Real
70
Teach Epidemiology
Enduring Epidemiological Understandings
Is alcohol consumption associated with lung cancer?
Exposure Outcome
Alcohol consumption lung cancer
Confounding occurs when a third factor is associated with both an exposure and an outcome. This third factor may create the appearance of a causal association between the exposure and outcome even if it isn’t really there.
71
Teach Epidemiology
Enduring Epidemiological Understandings
Confounding occurs when a third factor is associated with both an exposure and an outcome.
Exposure Outcome
Alcohol consumption lung cancer
Cigarette Smoking
Confounder
Smoking is associated with alcohol use and lung cancer.
72
Teach Epidemiology
Enduring Epidemiological Understandings
Can you think of another relationship between an exposure and an outcome where there might be a confounder?
Exposure Outcome
Potential Confounder
73
Teach Epidemiology
Enduring Epidemiological Understandings
Is there an association between OC use and MI?
OC use MI (heart attack)
Exposure Outcome
Cigarette Smoking
Potential Confounder
74
Teach Epidemiology
Enduring Epidemiological Understandings
Can you think of a way to take a confounder into consideration when looking at the relationship between an exposure and an outcome?
Exposure Outcome
Potential Confounder
75
Teach Epidemiology
Enduring Epidemiological Understandings
Randomization (in experimental studies, randomly assign participants so that an equal proportion of those with the possible confounding factor will be in the exposure groups)
Drug A Survival
Drug B
Age (Possible Confounder)
76
Teach Epidemiology
Enduring Epidemiological Understandings
Alcohol use lung cancer
cigarette smoking
Restriction (limit the study to those who do not have the confounding factor e.g. never smokers)
Matching(in a casecontrol study match cases with controls who have or do not have the potential confounder e.g. lung cancer cases matched to controls who have the same smoking status)
77
Teach Epidemiology
Enduring Epidemiological Understandings
Alcohol use lung cancer
cigarette smoking
Stratification (Conduct the analysis separately for different levels of the possible confounder smoking status)
Statistical methods (A number of statistical methods allow for the assessment of the relationship between an exposure and an outcome, while controlling for possible confounders.)
78
Teach Epidemiology
Enduring Epidemiological Understandings
Chance
Confounding
Bias
Real
79
Teach Epidemiology
Enduring Epidemiological Understandings
a systematic deviation of results or inferences from the truth or processes leading to such systematic deviation; any systematic tendency in the collection, analysis, interpretation, publication, or review of data that can lead to conclusions that are systematically different from the truth. In epidemiology, does not imply intentional deviation.
*Definition Source: Principles of Epidemiology in Public Health Practice Third Edition , U.S. DHHS, CDC
80
Teach Epidemiology
Enduring Epidemiological Understandings
Bias can appear in all types of epidemiologic studies
Many types of bias have been identified by epidemiologists
Two main forms of bias in epidemiologic studies
Information Bias
Selection Bias
81
Teach Epidemiology
Enduring Epidemiological Understandings
Information Bias*
systematic difference in the collection of data regarding the participants in a study (e.g., about exposures in a casecontrol study, or about health outcomes in a cohort study) that leads to an incorrect result (e.g., risk ratio or odds ratio) or inference.
Can you think of some situations that might result in information bias?
*Definition Source: Principles of Epidemiology in Public Health Practice Third Edition , U.S. DHHS, CDC
82
Teach Epidemiology
Enduring Epidemiological Understandings
Information Bias
Misclassification Bias
Nondifferential (inaccuracy of the data collection unrelated to exposure or disease status)
Differential (rate of misclassification is different in different study groups)
Recall Bias
Observer Bias
83
Teach Epidemiology
Enduring Epidemiological Understandings
Misclassification Bias
Nondifferential (inaccuracy of the data collection unrelated to exposure or disease status)
some controls might be called cases
some exposed people might be identified
as nonexposed
How might this impact ORs and RRs if there is a true relationship between exposure and outcome?
84
Teach Epidemiology
Enduring Epidemiological Understandings
Differential Misclassification (rate of misclassification is different in different study groups)
Recall Bias (CaseControl Study)
(differential recollection of exposure among cases and controls)
Observer Bias (Cohort study/Trial)
(if those assessing the disease outcome look more closely at the exposed group compared to the unexposed group)
85
Teach Epidemiology
Enduring Epidemiological Understandings
Differential Misclassification (rate of misclassification is different in different study groups)
Recall Bias (CaseControl Study)
(differential recollection of exposure among cases and controls)
Observer Bias (Cohort study/Trial)
(if those assessing the disease outcome look more closely at the exposed group compared to the unexposed group)
86
Teach Epidemiology
Enduring Epidemiological Understandings
Differential (rate of misclassification is different in different study groups)
Recall Bias (CaseControl Study)
may be minimized with careful interview/survey item construction
Observer Bias (Cohort study/Trial)
can be controlled by blinding those who are assessing the outcome regarding the exposure status
87
Teach Epidemiology
Enduring Epidemiological Understandings
Blinding (masking) intervention assignment is hidden from participants, trial investigators, or assessors.
Singleblind one of the three categories of individuals (normally participant) remains unaware of intervention assignment
Doubleblind participants, investigators, and assessors usually all remain unaware of the intervention assignments
Tripleblind usually means a doubleblind trial that also maintains a blind data analysis.
* F Schulz, David A Grimes, Blinding in randomised trials: hiding who got what THE LANCET • Vol 359 • February 23, 2002 • www.thelancet.com
88
Teach Epidemiology
F Schulz, David A Grimes, Blinding in randomised trials: hiding who got what THE LANCET • Vol 359 • February 23, 2002 • www.thelancet.com
Potential Benefits accruing dependent on those individuals successfully blindedF Schulz, David A Grimes, Blinding in randomised trials: hiding who got what THE LANCET • Vol 359 • February 23, 2002 • www.thelancet.com
Enduring Epidemiological Understandings
Selection Bias*
systematic difference in the enrollment of participants in a study that leads to an incorrect result (e.g., risk ratio or odds ratio) or inference.
*Definition Source: Principles of Epidemiology in Public Health Practice Third Edition , U.S. DHHS, CDC
91
Teach Epidemiology
Enduring Epidemiological Understandings
Selection Biassystematic difference in the enrollment of participants in a study that leads to an incorrect result (e.g., risk ratio or odds ratio) or inference.
Crosssectional study
Prevalence study of asthma
Nonrespondents
Higher proportion of smokers
Higher prevalence of asthma symptoms
How might this impact prevalence estimates?
92
Teach Epidemiology
Enduring Epidemiological Understandings
Selection Biassystematic difference in the enrollment of participants in a study that leads to an incorrect result (e.g., risk ratio or odds ratio) or inference.
CaseControl Study:
Study of smoking Lung Cancer
Cases: Lung CA in hospital
Controls: Heart attack (MI) patients in hospital
How might this impact the OR?
93
Teach Epidemiology
Enduring Epidemiological Understandings
Selection Biassystematic difference in the enrollment of participants in a study that leads to an incorrect result (e.g., risk ratio or odds ratio) or inference.
Cohort Study:
Study of H20 Exercise Heart Attack
Participants in H20 Class at Y
Neighborhood nonparticipants
How might this impact the Relative Risk?
94
Teach Epidemiology
Enduring Epidemiological Understandings
Necessary and SufficientRare to have a factor be both necessary and sufficient.
E.g. When a group of people are exposed to someone with active TB disease, not all of them get infected.
96
Teach Epidemiology
Enduring Epidemiological Understandings
Necessary but not sufficientcompromised Exposure to TB Bacillus
immune system (necessary but not sufficient)
Malnutrition
lack compromised
Crowded medical care immune system
Living conditions
Source: modified from Figure 5.1 Bonita, Ruth. Basic epidemiology 2nd edition
97
Enduring Epidemiological Understandings
Necessary and SufficientCan you think of a factor that is both necessary and sufficient to cause disease?
98
Teach Epidemiology
Enduring Epidemiological Understandings
Necessary and SufficientIf the title of this article is correct….
Nature Medicine11, 740  747 (2005) Published online: 12 June 2005;  doi:10.1038/nm1261 Smallpox vaccine–induced antibodies are necessary and sufficient for protection against monkeypox virus
Necessary: Without the antibodies you won’t be protected
Sufficient: With the antibodies you will be protected
99
Teach Epidemiology
Enduring Epidemiological Understandings
Assessing a Causal RelationshipSource: Principles of Epidemiology in Public Health Practice Third Edition , U.S. DHHS, CDC
102
Teach Epidemiology
Enduring Epidemiological Understandings
Odds Ratio of lung cancer by 75 by countryStrength of Association:
relationship must be clear
Look for large Odds Ratios and Relative Risks
Source: British Journal of Cancer (2004) 91, 1280–1286. doi:10.1038/sj.bjc.6602078 www.bjcancer.com Table 2
103
Teach Epidemiology
Enduring Epidemiological Understandings
Consistency: observation of the association must be repeatable in different populations at different times
Ecological Studies
The rising death rate from lung cancer in many counties along with a rise in cigarette consumption
Case Control Studies*
By 1959, 21 independent groups of investigators in 8 different countries. More studies followed with similar results.
Cohort Studies*
By 1959, large cohort studies in two countries by three independent groups.
*Source: Cornfeld et al J. Nat. Cancer Inst. 22:173–203, 1959
104
Teach Epidemiology
Enduring Epidemiological Understandings
Temporality:
the cause (exposure) must precede the effect (the disease).
Source: The American Cancer Society . Cancer Statistics 2010.http://www.cancer.org/research/cancerfactsfigures/cancerfactsfigures/cancerfactsandfigures2010
105
Teach Epidemiology
Enduring Epidemiological Understandings
Odds ratios of lung cancer by age 75, for current cigarette smokers stratified by amount smoked per dayBiological Gradient:
There must be a dose response.
Source: British Journal of Cancer (2004) 91, 1280–1286. doi:10.1038/sj.bjc.6602078 www.bjcancer.com Table 4
106
Teach Epidemiology
Enduring Epidemiological Understandings
Assessing a Causal Relationshiphttp://www.txtwriter.com/onscience/articles/smokingcancer2.html
Biological Plausibility: the explanation must make sense biologically.
107
Teach Epidemiology
Enduring Epidemiological Understandings
Assessing a Causal RelationshipSource: Principles of Epidemiology in Public Health Practice Third Edition , U.S. DHHS, CDC
108
Teach Epidemiology
10:15 AM
Teach Epidemiology
10:30 AM
Teach Epidemiology
11:30 AM
Teach Epidemiology
12:30 PM
Teach Epidemiology
1:00 PM
Teach Epidemiology
E
%
Hypothesis
or
DZ
%
or
Healthy People
Healthy People
DZ
E

?
DZ
Risk
Relative Risk
Total
Exposure
Outcome
a b
c d
Turned Up Together
Where are we?
126
Teach Epidemiology
Enduring Epidemiological Understandings
Teach Epidemiology
Enduring Epidemiological Understandings
Teach Epidemiology
Ties, Links, Relationships, and Associations
Suicide Higher in Areas with Guns
Family Meals Are Good for Mental Health
Study Links Iron Deficiency to Math Scores
Study Concludes: Movies Influence Youth Smoking
Lack of High School Diploma Tied to US Death Rate
Study Links Spanking to Aggression
Depressed Teens More Likely to Smoke
Snacks Key to Kids’ TV Linked Obesity: China Study
Pollution Linked with Birth Defects in US Study
Kids Who Watch RRated Movies More Likely to Drink, Smoke
Ties, Links, Relationships, and Associations
Suicide Higher in Areas with Guns
Family Meals Are Good for Mental Health
Study Links Iron Deficiency to Math Scores
Study Concludes: Movies Influence Youth Smoking
Lack of High School Diploma Tied to US Death Rate
Study Links Spanking to Aggression
Depressed Teens More Likely to Smoke
Snacks Key to Kids’ TV Linked Obesity: China Study
Pollution Linked with Birth Defects in US Study
Kids Who Watch RRated Movies More Likely to Drink, Smoke
Possible Explanations for Finding an Association
1.
Cause
2.
Confounding
3.
Reverse Time Order
Chance
4.
5.
Bias
Epidemiology
... the study of the distribution and determinants of healthrelated states or events in specified populations and the application of this study to the control of health problems.
Leon Gordis, Epidemiology, 3rd Edition, Elsevier Saunders, 2004.
Possible Explanations for Finding an Association
1.
Cause
2.
Confounding
3.
Reverse Time Order
Chance
4.
5.
Bias
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.
X1
X1
X1
X1
X1
X1
DZ
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
Necessary and Sufficient
Diagram
2X12 Table
DZ
DZ
X1
a
b
c
d
X1
X1
X1
X1
X1
X1
X1
X1
X1
+
X2
+
X3
DZ
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
Necessary but Not Sufficient
Diagram
2X12 Table
DZ
DZ
X1
a
b
c
d
X1
X1
X1
X1
+
X2
+
X3
X1
X1
X1
X4
+
X5
+
X6
DZ
X1
X1
X1
X7
+
X8
+
X9
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
X1
Not Necessary and Not Sufficient
Diagram
2X12 Table
DZ
DZ
X1
a
b
c
d
X1
X
X
X
+
X
+
X
X
X
X
X
+
X
+
X
DZ
X
X
X
X
+
X
+
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
Not Necessary and Not Sufficient
Diagram
2x2 Table
DZ
DZ
X
X
a
b
c
d
X
Heart Attack
Lack of Fitness
No Lack of Fitness
Lack of fitness and physical activity causes heart attacks.
a bc d
Lead Poisoning
Lack of Supervision
No Lack of Supervision
Lack of supervision of small children causes lead poisoning.
a bc d
Ties, Links, Relationships, and Associations
Suicide Higher in Areas with Guns
Family Meals Are Good for Mental Health
1.
Cause
Study Links Iron Deficiency to Math Scores
Study Concludes: Movies Influence Youth Smoking
2.
Confounding
Lack of High School Diploma Tied to US Death Rate
3.
Reverse Time Order
Study Links Spanking to Aggression
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 RRated Movies More Likely to Drink, Smoke
Enduring Epidemiological Understandings
Teach Epidemiology
Possible Explanations for Finding an Association
1.
Cause
2.
Confounding
3.
Reverse Time Order
Chance
4.
5.
Bias
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.
Not Observed
Inference
Population
Sample
Process of predicting from what is observed
to what is not observed.
Deck of 100 cards
a
b
c
d
No Marijuana
No Marijuana
Odd #
25 cards
25 cards
25 cards
Even #
25 cards
Population
Total
a
b
=
=
c
d
b
c
d
No Marijuana
No Marijuana
50
25
25
Odd #
25 cards
25 cards
50
25
25
25 cards
Even #
25 cards
Population
Population
Total
=
=
a
b
c
d
No Marijuana
No Marijuana
No Flu
Flu
50
50
25
25
25
25
Odd #
M&M’s
25 cards
25 cards
50
50
25
25
25
25
No M&M’s
25 cards
Even #
25 cards
Population
Total
=
=
=
Total
a
b
c
d
No Marijuana
No Marijuana
50
25
25
Odd #
25 cards
25 cards
50
25
25
25 cards
Even #
25 cards
Population
=
=
Total
Risk
25 / 50 or 50%
25 / 50 or 50%
a
b
c
d
No Marijuana
No Marijuana
50
25
25
Odd #
25 cards
25 cards
50
25
25
25 cards
Even #
25 cards
Risk
= 1
50 % / 50% =
Population
=
=
Total
Relative Risk
25 / 50 or 50 %
50 %
____
25 / 50 or 50 %
50 %
Possible Explanations for Finding an Association
Chance
To occur accidentally.
To occur without design.
A coincidence.
b
Sample
No Marijuana
No Marijuana
10
5
5
Odd #
10
5
5
Even #
25 cards
25 cards
25 cards
25 cards
Sample
Sample of 20 cards
Total
b
Sample
No Marijuana
No Marijuana
10
5
5
Odd #
10
5
5
Even #
Risk
25 cards
25 cards
25 cards
25 cards
Sample
Sample of 20 cards
Total
5 / 10 or 50 %
5 / 10 or 50 %
b
Sample
No Marijuana
No Marijuana
10
5
5
Odd #
10
5
5
Even #
Relative Risk
= 1
50 % / 50% =
25 cards
25 cards
25 cards
25 cards
Sample
Sample of 20 cards
Total
Risk
5 / 10 or 50 %
50 %
____
5 / 10 or 50 %
50 %
No Marijuana
No Marijuana
Sample of 20 cards
Odd #
Even #
Risk
Relative Risk
5 / 10 = 50 %
50 1
5 / 10 = 50 %
Sample
CDC
By Chance
Total
%
___
%
=
Chance
How many students picked a sample with 5 people in each cell?
No Marijuana
No Marijuana
Total
Risk
Relative Risk
5
5
10
5 / 10 or 50 %
Odd #
50 %
____
5
5
10
5 / 10 or 50 %
50 %
Even #
By Chance
Ties, Links, Relationships, and Associations
Study Links Having an Odd Address to Marijuana Use
Ties, Links, Relationships, and Associations
Study Links Having an Even Address to Marijuana Use
No Marijuana
Odd #
Even #
Risk
Relative Risk
5 / 10 = 50 %
50
5 / 10 = 50 %
Different Sample Sizes
b
Sample of 20 cards
50
Total
%
___
%
=
By Chance
By Chance
25 cards
25 cards
25 cards
25 cards
Chance
50 cards
Relative Risks
Greater than 1
Less than 1
No Marijuana
Odd #
Even #
Risk
Relative Risk
5 / 10 = 50 %
50
5 / 10 = 50 %
Different Sample Sizes
b
Sample of 20 cards
75
Total
%
___
%
=
By Chance
By Chance
25 cards
25 cards
25 cards
25 cards
Chance
75 cards
Relative Risks
Greater than 1
Less than 1
No Marijuana
Odd #
Even #
Risk
Relative Risk
5 / 10 = 50 %
50 1
5 / 10 = 50 %
Different Sample Sizes
b
Sample of 20 cards
99
Total
%
___
%
=
By Chance
By Chance
25 cards
25 cards
25 cards
25 cards
Chance
99 cards
Relative Risks
Greater than 1
Less than 1
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 Links Iron Deficiency to Math Scores
Study Concludes: Movies Influence Youth Smoking
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 RRated Movies More Likely to Drink, Smoke
Enduring Epidemiological Understandings
Teach Epidemiology
Enduring Epidemiological Understandings
Teach Epidemiology
Explaining Associations and Judging Causation
Teach Epidemiology
Explaining Associations and Judging Causation
Coffee and Cancer of the Pancreas
1.
Cause
2.
Confounding
3.
Reverse Time Order
Chance
4.
5.
Bias
Teach Epidemiology
Explaining Associations and Judging Causation
Does evidence from an aggregate of studies support a causeeffect relationship?
Causal or Not Causal?
Guilt or Innocence?
197
Teach Epidemiology
Explaining Associations and Judging Causation
Handout
Sir Austin Bradford Hill “The Environment and Disease: Association or Causation?” Proceedings of the Royal Society of Medicine January 14, 1965
Teach Epidemiology
Explaining Associations and Judging Causation
“In what circumstances can we pass from this observed association to a verdict of causation?”
199
Teach Epidemiology
Explaining Associations and Judging Causation
“Here then are nine different viewpoints from all of which we should study association before we cry causation.”
200
Teach Epidemiology
Explaining Associations and Judging Causation
Does evidence from an aggregate of studies support a causeeffect relationship?
Teach Epidemiology
O
E
O
E
Healthy People
Healthy People
O
E
O
O
Random Assignment
O
O
E
E
O
E
O
Healthy People
Healthy People
O
O
E
E
E
E
Timeline
Timeline
Timeline
Timeline
Explaining Associations and Judging Causation
CaseControl Study
Randomized Controlled Trial
Cohort Study
CrossSectional Study
Teach Epidemiology
Explaining Associations and Judging Causation
Stress causes ulcers.
Helicobacter pylori causes ulcers.
Teach Epidemiology
Explaining Associations and Judging Causation
Teach Epidemiology
Epidemiology
... the study of the distribution and determinants of healthrelated states or events in specified populations and the application of this study to the control of health problems.
Leon Gordis, Epidemiology, 3rd Edition, Elsevier Saunders, 2004.
209
X
… and you avoided or eliminated the hypothesized cause, what would happen to the outcome?
Control of Health Problems
If an association was causal, ….
?
Hypothesized Exposure
Outcome
X
210
Unobserved Exposure
X
… and you avoided or eliminated the hypothesized cause, what would happen to the outcome?
Control of Health Problems
If the association was found due to confounding, ….
?
Hypothesized Exposure
Outcome
211
found due to reversed time order, ….
X
… and you avoided or eliminated the hypothesized cause, what would happen to the outcome?
Control of Health Problems
If an association was found due to reversed timeorder, ….
?
Hypothesized Exposure
Outcome
212
… and you avoided or eliminated the hypothesized cause, what would happen to the outcome?
Control of Health Problems
If an association was found due to chance, ….
found due to chance, ….
?
Hypothesized Exposure
Outcome
213
… and you avoided or eliminated the hypothesized cause, what would happen to the outcome?
Control of Health Problems
If an association was found due to bias, ….
found due to bias, ….
?
Hypothesized Exposure
Outcome
214
If an association was causal, ….
causal, ….
Hypothesized Exposure
X
Outcome
X
… and you avoided or eliminated the hypothesized cause, what would happen to the outcome?
Control of Health Problems
... the study of the distribution and determinants of healthrelated states or events in specified populations and the application of this study to the controlof health problems.
215
... the study of the distribution and determinants of healthrelated states or events in specified populations and the application of this study to the controlof health problems.
1.
Cause
2.
Confounding
3.
Reverse Time Order
Chance
4.
5.
Bias
216
Ties, Links, Relationships, and Associations
Suicide Higher in Areas with Guns
Family Meals Are Good for Mental Health
1.
Cause
Study Links Iron Deficiency to Math Scores
Study Concludes: Movies Influence Youth Smoking
2.
Confounding
Lack of High School Diploma Tied to US Death Rate
3.
Reverse Time Order
Study Links Spanking to Aggression
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 RRated Movies More Likely to Drink, Smoke
217
Enduring Epidemiological Understandings
Teach Epidemiology
2:45 PM
Teach Epidemiology
3:00 AM
Teach Epidemiology
3:30 PM
Teach Epidemiology
4:00 PM
What do you mean  Teach Epidemiology?
Leverage the Science Olympiad Competition
http://soinc.org/
Teach Epidemiology
Enduring Epidemiological Understandings
Knowledge that “… is connected and organized, and … ‘conditionalized’ to specify the context in which it is applicable.”
National Research Council , Learning and Understanding
Teach Epidemiology
Authentic Assessment
Teach Epidemiology
Finish
http://www.njscienceolympiad.org/content/events/c/websites/epidemiology/index.html
Teach Epidemiology
Think Like an Epidemiologist Challenge
New Jersey Science Olympiad, March 16, 2010
Finish
Name
School
Thank you for stepping up, being a pioneer, and competing in the first Think Like an Epidemiologist Challenge trial event.
You worked with others, developed epidemiologic knowledge and skills, and used judgment and innovation to actually "do" epidemiology under pressure.
We hope you enjoyed the challenge.
Detectives in the Classroom
Teach Epidemiology
Robert Wood Johnson Foundation
Special thanks to the Epidemiology Section of the American Public Health Association for allowing us to distribute their Section pins to the student participants in the 2010 Think Like an Epidemiologist Challenge.