Anda di halaman 1dari 14

Ecological Study Design

Sandro Galea, MD, DrPH


Professor of Epidemiology
University of Michigan, School of Public Health
& Columbia University, MSPH

Email: sgalea@umich.edu

Lecture Learning Objectives

Describe ways in which epidemiological findings affect policy


decisions made by governments, public health agencies, and
international organizations

Review definition of epidemiology and name epidemiological


study designs

Give definition and discuss different types of ecological studies

Interpret measures of effect estimated in ecological studies

List advantages and limitations of ecological studies

1
CORE
Epidemiology Ecological Study Design

Learning Objectives
• Provide classification and discuss different types of
epidemiological study designs
Ecological Study Design • Discuss ecological study design:
− Describe essential characteristics
Sandro Galea, MD, DrPH
Professor of Epidemiology − Interpret measures of effect
University of Michigan, School of Public Health − List strengths and limitations
& Columbia University, MSPH
− Give examples of uses of ecological study
Email: sgalea@umich.edu
designs for policy decisions by governments
and public health agencies

CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

Learning Objectives Epidemiological Methods


¾ Provide classification and discuss different types of
Classifications by:
epidemiological study designs
1.approach to data collection
• Discuss ecological study design:
− Describe essential characteristics 2.goal

− Interpret measures of effect 3.timing and directionality


− List strengths and limitations 4.unit of analysis
− Give examples of uses of ecological study
designs for policy decisions by governments
and public health agencies

CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

1. Classification by approach 2. Classification by goal


to data collection

• Experimental • Descriptive

– RCTs, field trials, community intervention – ecological correlational studies, case reports, case

series, cross-sectional surveys


and cluster randomized trials
• Analytic
• Non-experimental or observational
– observational studies and intervention studies (RCTs)
– cohort, case-control
CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

3. Classification by timing and 3. Classification by timing and


directionality directionality
Retrospective
Diseased
− Directionality: "Which did you observe first, the cohort study
exposure or the disease?“ exposed Non-diseased
– forward (RCT, cohort)
– backwards (case-control)
Diseased
unexposed
Non-diseased
− Timing: “Has the information being studied already x
occurred before the study actually began?"
past present future
– retrospective and prospective cohort studies Diseased
exposed Non-diseased
Diseased
unexposed
RCT Non-diseased

CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

Example 1:
4. Classification by unit of analysis
A hospital-based case-control study of association
− What is a unit?
between consumption of artificial sweeteners and
diet beverages and bladder cancer.
• Observations for which outcome and exposure are
measured What is the unit of analysis in this study?
− Individual-level variables are properties of
individuals A. Individual-level
− ecological variables are properties of groups, B. Ecological (group-level)
organizations or places

CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

Example 2: Example 3:

A study, based on the records of the Connecticut A study, based on national cancer registries, looks
population cancer registry, looks at the effects of at the relation between population consumption of
prior radiation treatment for primary breast cancer extra virgin olive oil and incidence of
on subsequent development of lung cancer. gastrointestinal cancers across countries.
What is the unit of analysis in this study? What is the unit of analysis in this study?

A. Individual-level
B. Ecological (group-level) A. Individual-level
B. Ecological (group-level)
CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

Ecologic studies Learning Objectives


Refers to unit of analysis NOT type of study • Provide classification and discuss different types of
design epidemiological study designs
In context of epidemiology, ecologic means ¾ Discuss ecological study design:
group level
Units of analysis are groups, e.g., countries,
¾ Describe essential characteristics
neighborhoods − Interpret measures of effect
Multiple study designs possible − List strengths and limitations
− Give examples of uses of ecological study
designs for policy decisions by governments
and public health agencies

CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

Ecological studies Therefore, main properties of ecological


studies:

“Studies in which measures that represent • Units of analysis are groups


characteristics of entire populations are used to
describe disease in relation to some factor of
interest.” • Both exposure and outcome are measured for
groups
C. Hennekkens

CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

Levels of analyses

Country

State

Persons

Group-level study
CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

Examples of ecological studies

• Exploratory studies
• Multiple-group studies
− differences among groups
• Time-trend studies
− changes over time within groups
• Mixed studies
− combination of the above

CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

Example of exploratory ecological study

Cotterill et al., (2001) Eur J Cancer; 37: 1020-26.

CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

Example of multi-group ecological study Measures of exposure in ecological studies:

− Aggregate – summaries of observations derived


from individuals in each group
• per-capita alcohol consumption in a given country

• the proportion of smokers and median family income

• proportion of the population under the age of 18 and


rate of thyroid cancer

Prisyazhniuk et al., Lancet (1991); 338: 1334-35.


CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

Measures of exposure in ecological studies: Measures of exposure in ecological studies:

− Aggregate – summaries of observations derived − Aggregate – summaries of observations derived from


individuals in each group
from individuals in each group
− Environmental (compositional) – physical characteristics
− Environmental (compositional) – physical
of the place in which members of each group live or work;
characteristics of the place in which members of with an analog at the individual level
each group live or work; with an analog at the − Global (contextual) – attributes of groups, organizations
individual level or places for which there is no distinct analogue at the

• air pollution level and hours of sunlight individual level


• population density
• well water arsenic concentration and skin lesion rate in
• existence of special law or type of health-care system
each village in Bangladesh

CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

Classification of ecologic variables

Key take away

In ecological studies, exposure represents an


average consumption for a population rather
than the actual consumption patterns of the
individuals

Blakely TA, Woodward AJ. Ecological effects in multi-level studies. J Epidemiol Community Health. 2000; 54:367-74

CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

In fact...individual-level studies Learning Objectives


show:
• Provide classification and discuss different types of
Alcohol consumption
epidemiological study designs
• Discuss ecological study design:
− Describe essential characteristics
¾ Interpret measures of effect
− List strengths and limitations
− Give examples of uses of ecological study
designs for policy decisions by governments
and public health agencies
Mortality from CHD
CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

Measure of association is correlation If association is linear…


coefficient, r
y = b0 + b1x, where b1 is slope (regression coefficient)
− Quantifies the extent to which two variables (exposure and

outcome) are associated − Proportionate increase or decrease in disease frequency for every

unit change in level of exposure

− r varies between –1 and 1

CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

55%

45%

30% better well off


70% least well off

Post Tax and


Benefit Money, U.S.

Wilkinson et al., (1992) BMJ; 304:165-8. Wilkinson et al., (1992) BMJ; 304:165-8.

CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

Learning Objectives
77% • Provide classification and discuss different types of
epidemiological study designs
23%
• Discuss ecological study design:
− Describe essential characteristics
50 % better well off
households
− Interpret measures of effect
50% least well off ¾ List strengths and limitations
households
− Give examples of uses of ecological study
Post Tax and Benefit
designs for policy decisions by governments
Household Income, Utah and public health agencies
Kaplan et al., (1996) BMJ; 312 (7037): 999-1003.
Epidemiology CORE Ecological Study Design CORE
Epidemiology Ecological Study Design

Reminder...levels of analyses Strengths of ecological studies:


1. Low cost and convenience
• Examples of secondary data sources: population
Country registries, vital records, large surveys

2. Ability to overcome measurement limitations of


State individual-level studies
• When exposures cannot be measured accurately for large
numbers of subjects
• When there is too much within-person variability in
Persons exposures (e.g., dietary factors)

3. Ability to overcome design limitations of individual-


level studies
• When there is not enough variability within the study area

Epidemiology CORE Ecological Study Design CORE


Epidemiology Ecological Study Design

Strengths of ecological studies, continued:


1
4.Interest in ecological effects % CHD
• Some variables can only be defined and measured at
the group level.
2
• Example: effect of helmet-use laws on the motorcycle-
related mortality rate of riders in different states

5.Simplicity of analysis and presentation 3


• Example: a national debate over who should be
screened for breast cancer is very important, but not
less important is the issue of who and where are being
screened
% High Fat Diet

Epidemiology CORE Ecological Study Design CORE


Epidemiology Ecological Study Design

Limitation of ecological studies (1)


No information on the cross-classification of
exposures and outcomes within groups 1
% CHD

CHD CHD CHD


2
High Fat Diet

Yes No Yes No Yes No


Y ? ? 50% Y ? ? 40% Y ? ? 10%
N ? ? 50% N ? ? 60% N ? ? 90%
3
5% 95% 3% 97% 1% 99%
Country 1 Country 2 Country 3

Example of % high fat diet and CHD rate % High Fat Diet
CHD = Coronary heart disease
CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

The problem with ecologic studies


Therefore...

We cannot be SURE that the people who got


CHD were the ones who had high fat diets

Greenland S. Ecologic versus individual-level sources of bias in ecologic estimates of contextual health effects. Int J Epidemiol.
2001;30(6):1343-50.

CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

The problem with ecologic studies The problem with ecologic studies
Smoking

Compare Group A vs. Group B for prevalence of cancer mortality

Mortality rate in group A=560/100=5.6 per 1,000 PY


Mortality rate in group B=560/100=5.6 per 1,000 PY
Therefore RRA/RRB=1
Lung cancer

Greenland S. Ecologic versus individual-level sources of bias in ecologic estimates of contextual health effects. Int J Epidemiol. Greenland S. Ecologic versus individual-level sources of bias in ecologic estimates of contextual health effects. Int J Epidemiol.
2001;30(6):1343-50. 2001;30(6):1343-50.

CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

There are two possibilities that contribute Limitation of ecological studies (2)
to this contextual null effect
Scenario 2 No information on individual-level variables which may
RR(X vs no X)=7.0/3.5 in A and 8.0/4.0 in B, be confounders (alternate explanations)
i.e. RRX=2
RR(A vs B)=7/8 State 1 State 2 State 3

Scenario 3 High Medium Low


RR(X vs no X)=4.0/8.0 in A and 3.5/7.0 in B, Income Income Income
i.e. RRX=2 Inequality Inequality Inequality
RR(A vs B)=4/3.5, or 8/7
Death
Rate 12 % 9% 5%
Therefore, ecologic analysis represents uncertainty and loses information
due to aggregation
Ecologic analysis must rely on prior information about distributional
properties of variables within populations to draw valid inference Example of Income Inequality and Mortality
IF we do not know RRX, then we cannot know the direction of the
individual-level confounding; hence cross-level confounding
Greenland S. Ecologic versus individual-level sources of bias in ecologic estimates of contextual health effects. Int J Epidemiol.
2001;30(6):1343-50.
CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

Therefore...

It is possible that high income inequality states


have higher death rates because they have MORE
people of lower income
We cannot be SURE that it is income inequality
itself that is related to mortality (could be
individual level income)

Blakely TA, Woodward AJ. Ecological effects in multi-level studies. J Epidemiol Community Health. 2000; 54:367-74

CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

The ecologic fallacy Ecological fallacy:


Fallacy of drawing inferences regarding associations at At the group level:
the individual-level based on data collected for groups - No relationship between OC use and CHD mortality in
young women
e.g., if data shows that countries with low income have
higher death rate, we infer that individuals with low
income have higher death rate At the individual level:
- two-fold increase in risk of CHD among OC users
The importance (and presence) of the ecologic fallacy compared to nonusers
may differ for different research questions
Summary:
• Impossible to detect from ecologic data
• Incorrect to assume that there is no relationship
between OC use and CHD mortality

CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

Ecological fallacy: Ecological fallacy, continued:


Fallacy of drawing inferences regarding
associations at the individual level based on the • This does not mean that every ecological study has
group-level data ecological fallacy!
Group level

− The group-level data: • The importance of the ecological fallacy may differ for
• inverse linear relationship between different research questions
alcohol consumption and CHD mortality
• Those who consume large quantities of • Potential strategies to reduce ecological fallacy:
alcohol have the smallest mortality

• Use smaller units to make groups more homogeneous


− The individual-level data: • Supplement ecological variables with individual-level
• relationship is J-shaped variables
• non-drinkers and those who consume
large quantities have higher mortality
Individual
than those who consume small to
moderate amounts of alcohol. level
CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

Conversely, the atomistic fallacy Atomistic fallacy:


− drawing inferences at a higher level from
Fallacy of drawing inferences regarding associations at analyses performed at a lower level
the group-level based on data collected for individuals
− Example:
e.g., if the data shows that individuals who eat a high • in a case-control collect information on
various possible exposures but ignore the
fat diet have a greater risk of liver cancer, we infer that geographic, spatial, and social context in
in countries with more fat consumption there is an which a person lives
overall greater rate of liver cancer Group level

Individual
level

CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

Bottom line...levels of analyses Bottom line...levels of analyses

Country Country

State State

Persons Persons

CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

Bottom line...levels of analyses Summary of limitations of ecological studies:


1. No information on the cross-classification of
exposures and outcomes within groups
Country
? ? A+B
? ? C +D
State
2. Lack of ability to control for the effects of possible
alternate explanations (confounders)

Persons • Exposure can be associated with a number of factors that are


related to the elevated risk of disease; it is not possible to
separate their effects using ecological data
CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

But…
“the possibility of bias does not demonstrate the
presence of bias, and…a conflict between ecologic
and individual-level estimates does not by itself
demonstrate that the ecologic estimates are the
more biased…”
Ecologic estimates incorporate a contextual
component that “may be viewed as both a key
strength and weakness of ecologic studies”

Greenland S. Ecologic versus individual-level sources of bias in ecologic estimates of contextual health effects. Int J Epidemiol.
2001;30(6):1343-50.

CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

Case-control meta-analysis Cohort study

RCT

CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

Saga continues… testing new hypotheses


generated by ecological studies Learning Objectives
Cancer Causes Control. 2005 Apr;16(3):225-33. Dietary intakes of fruit, • Review definition of epidemiology, its origins and uses
vegetables, and fiber, and risk of colorectal cancer in a prospective cohort
of women (United States). Lin J et al.
• Provide classification and discuss different types of
CONCLUSIONS: “Our data offer little support for epidemiological study designs
associations between intakes of fruit, vegetables, and
fiber, and colorectal cancer risk. However, our data • Discuss ecological study design:
suggest that legume fiber and/or other related sources
may reduce risk of colorectal cancer. “ − Describe essential characteristics
Int J Cancer. 2006 Oct;119(12):2938-2942 Dietary intake of calcium, − Interpret measures of effect
fiber and other micronutrients in relation to colorectal cancer risk:
Results from the Shanghai Women's Health Study. Shin A et al. − List strengths and limitations
CONCLUSIONS: “No apparent associations were found for Give examples of uses of ecological study
fiber, total vitamin A, carotene, vitamins B1, B2, B3, C and
designs for policy decisions by governments
E with colorectal cancer risk. Our results suggest that
calcium may be protective against colorectal cancer and public health agencies
development …”
CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

Washington Heights/ Inwood community Washington Heights/ Inwood community

CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

Health profile of the Washington Heights/


Inwood (WaHI) community

“My Community’s Health”

http://www.nyc.gov/html/doh/html/home/home.shtml

CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

Snapshots from the community


CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

Possible Reasons?

CORE
Epidemiology Ecological Study Design CORE
Epidemiology Ecological Study Design

Possible Reasons? Possible Reasons?

CORE
Epidemiology Ecological Study Design

Anda mungkin juga menyukai