Observational Research
EPI4921
LECTURES
➔Notes from Syllabus
- Observational research → Aims at studying the occurrence of phenomena that are
“naturally” present in the population or society, and characteristics that are
associated with these phenomena.
- As experimental research is not feasible for many relevant health issues and also
sometimes unethical, observational research is often needed to answer these
questions.
- The following topics will be dealt with:
1. Observational research designs and their advantages and disadvantages:
cross-sectional research, case-control studies, prospective and historical
cohort studies, nested designs
2. Applications of measures of disease frequency and association in various
designs
3. Exposure measurement in observational research and misclassification
4. Criteria for evaluating the quality of research
5. Sources of bias
6. Ways to deal with bias
7. Effect modification and procedures to detect effect modification
8. Simple statistical analysis techniques in observational research
9. Application of these analysis techniques in different observational
research designs
10.Principles of causality, causal reasoning and causal diagrams
11.Reporting of observational research.
➔ Introduction:
- Epidemiology research can be divided into observational and experimental
research.
- Observational research → information is being collected about certain
characteristics of one or more groups of individuals, but the researcher does not
manipulate the determinants.
- The design and execution of observational research (that aims to simulate
the results of experimental research), is aimed at achieving comparability
of the groups of individuals that are being studied. If this comparability is
not achieved, several types of bias can occur, roughly divided into
, selection bias, information bias and confounding bias. This can pose a
threat to the internal validity of a study
- Classification of observational research in epidemiology:
1. Descriptive epidemiological studies
2. Analytical epidemiological studies: has 2 further classifications:
a. Correlation studies conducted on the level of populations
(ecological studies, geographical comparison, time trend
studies)
b. Studies conducted on the level of the individual: further
classified into:
- Cross-sectional studies
- Longitudinal studies (cohort, studies, case control
studies)
c. Hybrid, nested designs
- Experimental studies → the researchers intervened by assigning individuals to an
intervention group or a control group and subsequently measuring the effects of
the intervention group or a control group.
➔ Assessment of the Course:
1. Test 80% of final grade
2. Written assignment 20%
, Introduction on measures of
frequency and association
Lecture 1
➔ Epidemiology: Objectives and Uses:
- Epidemiology: is the study of how disease is distributed in the population and of
the factors that influence or determine this distribution.
- Why does a disease develop in some people and not in others?
- Why does a treatment work in certain patients and not in others?
➔ Uses:
- Determination of burden of disease on the population:
- Priorities for research and policy
- Goals for prevention
- Studying the prognosis of disease and the effectiveness of therapies
- Studying the value of screening programmes (Diagnosis)
- Studying the causes/risk factors of disease (Etiology)
➔ Types of Epidemiological study designs:
1. Experimental research
2. Observational research:
a. Descriptive epidemiology:
b. Analytical epidemiology:
i. Population Level:
- Correlation studies (geographical, time trend)
ii. Individual Level:
- Case Series (no control group)
- Cross-Sectional research
- Longitudinal research:
- Case control studies
- Cohort studies (follow
up)
- Hybrid designs (nesting)
, ➔ Notations and some algebra:
➔ Examples:
1. Descriptive epidemiology:
Example: Lifetime risk of being diagnosed with cancer in the US
2. Analytical epidemiology:
- Analytical epidemiology can be on the population level
Example: Relation of national per capita fat intake with risk of breast
cancer mortality.
- Such analyses are not adjusted to any confounding factors
- Statistical analysis is based on aggregated data. Data of the entire
population there is NO distinction between males and females.
Such figures can be biased (overestimation of the effect can
happen = Ecological Fallacy)
➔ Advantages and disadvantages ecological studies:
Advantages:
1. Generate hypothesis about exposure and disease relations
2. Quick and inexpensive
Disadvantages.
1. The statistical analysis is based on aggregated (e.g. population or community
level variables) rather than individual data
2. Regional differences in assessment of exposure and outcome measures
3. No information on confounders
4. Bias is likely → Ecological Fallacy
Observational Research
EPI4921
LECTURES
➔Notes from Syllabus
- Observational research → Aims at studying the occurrence of phenomena that are
“naturally” present in the population or society, and characteristics that are
associated with these phenomena.
- As experimental research is not feasible for many relevant health issues and also
sometimes unethical, observational research is often needed to answer these
questions.
- The following topics will be dealt with:
1. Observational research designs and their advantages and disadvantages:
cross-sectional research, case-control studies, prospective and historical
cohort studies, nested designs
2. Applications of measures of disease frequency and association in various
designs
3. Exposure measurement in observational research and misclassification
4. Criteria for evaluating the quality of research
5. Sources of bias
6. Ways to deal with bias
7. Effect modification and procedures to detect effect modification
8. Simple statistical analysis techniques in observational research
9. Application of these analysis techniques in different observational
research designs
10.Principles of causality, causal reasoning and causal diagrams
11.Reporting of observational research.
➔ Introduction:
- Epidemiology research can be divided into observational and experimental
research.
- Observational research → information is being collected about certain
characteristics of one or more groups of individuals, but the researcher does not
manipulate the determinants.
- The design and execution of observational research (that aims to simulate
the results of experimental research), is aimed at achieving comparability
of the groups of individuals that are being studied. If this comparability is
not achieved, several types of bias can occur, roughly divided into
, selection bias, information bias and confounding bias. This can pose a
threat to the internal validity of a study
- Classification of observational research in epidemiology:
1. Descriptive epidemiological studies
2. Analytical epidemiological studies: has 2 further classifications:
a. Correlation studies conducted on the level of populations
(ecological studies, geographical comparison, time trend
studies)
b. Studies conducted on the level of the individual: further
classified into:
- Cross-sectional studies
- Longitudinal studies (cohort, studies, case control
studies)
c. Hybrid, nested designs
- Experimental studies → the researchers intervened by assigning individuals to an
intervention group or a control group and subsequently measuring the effects of
the intervention group or a control group.
➔ Assessment of the Course:
1. Test 80% of final grade
2. Written assignment 20%
, Introduction on measures of
frequency and association
Lecture 1
➔ Epidemiology: Objectives and Uses:
- Epidemiology: is the study of how disease is distributed in the population and of
the factors that influence or determine this distribution.
- Why does a disease develop in some people and not in others?
- Why does a treatment work in certain patients and not in others?
➔ Uses:
- Determination of burden of disease on the population:
- Priorities for research and policy
- Goals for prevention
- Studying the prognosis of disease and the effectiveness of therapies
- Studying the value of screening programmes (Diagnosis)
- Studying the causes/risk factors of disease (Etiology)
➔ Types of Epidemiological study designs:
1. Experimental research
2. Observational research:
a. Descriptive epidemiology:
b. Analytical epidemiology:
i. Population Level:
- Correlation studies (geographical, time trend)
ii. Individual Level:
- Case Series (no control group)
- Cross-Sectional research
- Longitudinal research:
- Case control studies
- Cohort studies (follow
up)
- Hybrid designs (nesting)
, ➔ Notations and some algebra:
➔ Examples:
1. Descriptive epidemiology:
Example: Lifetime risk of being diagnosed with cancer in the US
2. Analytical epidemiology:
- Analytical epidemiology can be on the population level
Example: Relation of national per capita fat intake with risk of breast
cancer mortality.
- Such analyses are not adjusted to any confounding factors
- Statistical analysis is based on aggregated data. Data of the entire
population there is NO distinction between males and females.
Such figures can be biased (overestimation of the effect can
happen = Ecological Fallacy)
➔ Advantages and disadvantages ecological studies:
Advantages:
1. Generate hypothesis about exposure and disease relations
2. Quick and inexpensive
Disadvantages.
1. The statistical analysis is based on aggregated (e.g. population or community
level variables) rather than individual data
2. Regional differences in assessment of exposure and outcome measures
3. No information on confounders
4. Bias is likely → Ecological Fallacy
