HNH-31006 Study Design and Interpretation in Epidemiology and Public Health
Refresh knowledge
Cohort study
- Group of people followed over time until disease develops
- Purpose:
o To evaluate the occurrence (incidence) of disease in a carefully defined group of people
(monitoring)
o To investigate the causes of disease and to establish links between risk factors and health
outcomes (etiology)
- Measurements of exposure and outcome
o At baseline
▪ Exposure status
▪ Covariates: variables that are possibly related
with the exposure/outcome
o During follow-up
▪ Disease outcome
▪ Update information on exposure and covariates
▪ You count the number of incidence cases
Occurrence
- Count as incidence proportion or incidence rate
What is the exposure-disease association?
- Compare the occurrence of disease in people with and without exposure (etiology)
- We can compare incidence proportion or incidence rate in the exposed versus the unexposed
Example
- Calculate the Relative Risk (RR) of alcohol consumption on
cardiovascular disease
- RR = (10/(10+100))/(15/15+85)) = 0.61
Historical cohort study
- In a prospective cohort study the researcher starts with collecting the baseline data at the beginning
- In a retrospective cohort the researcher starts at the end
o Identify a population and observe events as they occur, and goes back to archives to find back
the previous exposure (determine exposure status from historical records)
o → requires good records of past exposure, such as birth weight
o → avoid long follow-up period, it gets more difficult to have the data present
,Strengths and limitations of a Cohort study
Strengths:
- Time sequence can be determined → causality
- Multiple outcomes /risk factors → sub analysis
- Study rare exposures
Limitations:
- Costs
- Not good for rare diseases
- Not good for diseases that take a long time to develop
- Ensuring people who started the study stay till the end
Case-control study design
Case-control study is very suitable:
- for rare conditions
- When disease slowly develop/have a long latency period
→ focus is usually on a single health outcome
→ much effort can be put on measuring the exposure of interest and potential confounders and on more
detailed measurements (e.g. interview, biomarkers (urine))
- Other advantages (though not always…)
o Less burden to participants
o Fast and cheap
The proper case group?
- Incident cases
o Newly diagnosed with the diseases
- Prevalent cases
o Both old and new occurrences
o Be aware of ‘reverse causation’ (similarly to cross-sectional studies)
- Be aware of the ‘supervisor effect’ (→ selection bias)
o Example: in a study of myocardial infarction, severe cases with highest exposure may die
before they reach the hospital and not enter the study
The proper control group?
- The control group should be representative for the (source) population of the cases
- Controls can be selected from general population, hospital, relatives, colleagues, friends
- Population controls
o Persons from the population where the cases came from who would also be identified as a
case when developing the disease
o Population registries
o Random digit dialling
o GP registrations etc
o Postal codes (neighbourhood controls)
Selection of controls
- Hospital controls = Patients admitted to the same hospital as the cases, but for other reasons (same
environment)
→ drawback: their risk factor distributions may not reflect that of the source population from which the cases
emerged
→ advantage: efficiency
Cross-sectional studies
- Measure all kind of exposures and all kind of diseases
- You can divide the population in people who are exposed to a certain factor and those who have the
disease where you are interested in
- Measures exposure and disease at the same moment in time
, Experimental study/trial
- The investigator assigns the exposure (randomly)
Epidemiological effect measures
Relationship between incidence, prevalence and disease duration
- Incidence = new cases
- Prevalence varies directly with both incidence and duration
o If incidence is low, and duration is long (chronic), prevalence will be large in relation to
incidence
o If the duration is short (due to recovery, migration or death),
prevalence will be small in relation to incidence
- Duration
Use of incidence and prevalence
1. Incidence is generally used for acutely acquired disease, prevalence is used for more permanent
states, conditions or attributes of ill-health
2. Incidence is more important when thinking of etiology of the disorder, prevalence when thinking of
societal burden of the disorder including the costs and resources consumed as a result of the disorder
Special types of incidence and prevalence measures
- Ways to express incidences: morbidity rate, mortality
rate, case-fatality rate, attack rate
- Ways to express prevalences: disease rate at autopsy,
birth defect rate
Incidence
- Counting new cases (e.g. 1 million)
- Measures how fast people are ‘catching’ the disease over a period in time
o E.g. in one month/year
- Expressed as:
o Proportion of people who develop the disease (e.g. 0.05 or 5%)
o Rate at which new cases of a disease have occurred (e.g. 5000 new cases per 100000 persons
per year)
Refresh knowledge
Cohort study
- Group of people followed over time until disease develops
- Purpose:
o To evaluate the occurrence (incidence) of disease in a carefully defined group of people
(monitoring)
o To investigate the causes of disease and to establish links between risk factors and health
outcomes (etiology)
- Measurements of exposure and outcome
o At baseline
▪ Exposure status
▪ Covariates: variables that are possibly related
with the exposure/outcome
o During follow-up
▪ Disease outcome
▪ Update information on exposure and covariates
▪ You count the number of incidence cases
Occurrence
- Count as incidence proportion or incidence rate
What is the exposure-disease association?
- Compare the occurrence of disease in people with and without exposure (etiology)
- We can compare incidence proportion or incidence rate in the exposed versus the unexposed
Example
- Calculate the Relative Risk (RR) of alcohol consumption on
cardiovascular disease
- RR = (10/(10+100))/(15/15+85)) = 0.61
Historical cohort study
- In a prospective cohort study the researcher starts with collecting the baseline data at the beginning
- In a retrospective cohort the researcher starts at the end
o Identify a population and observe events as they occur, and goes back to archives to find back
the previous exposure (determine exposure status from historical records)
o → requires good records of past exposure, such as birth weight
o → avoid long follow-up period, it gets more difficult to have the data present
,Strengths and limitations of a Cohort study
Strengths:
- Time sequence can be determined → causality
- Multiple outcomes /risk factors → sub analysis
- Study rare exposures
Limitations:
- Costs
- Not good for rare diseases
- Not good for diseases that take a long time to develop
- Ensuring people who started the study stay till the end
Case-control study design
Case-control study is very suitable:
- for rare conditions
- When disease slowly develop/have a long latency period
→ focus is usually on a single health outcome
→ much effort can be put on measuring the exposure of interest and potential confounders and on more
detailed measurements (e.g. interview, biomarkers (urine))
- Other advantages (though not always…)
o Less burden to participants
o Fast and cheap
The proper case group?
- Incident cases
o Newly diagnosed with the diseases
- Prevalent cases
o Both old and new occurrences
o Be aware of ‘reverse causation’ (similarly to cross-sectional studies)
- Be aware of the ‘supervisor effect’ (→ selection bias)
o Example: in a study of myocardial infarction, severe cases with highest exposure may die
before they reach the hospital and not enter the study
The proper control group?
- The control group should be representative for the (source) population of the cases
- Controls can be selected from general population, hospital, relatives, colleagues, friends
- Population controls
o Persons from the population where the cases came from who would also be identified as a
case when developing the disease
o Population registries
o Random digit dialling
o GP registrations etc
o Postal codes (neighbourhood controls)
Selection of controls
- Hospital controls = Patients admitted to the same hospital as the cases, but for other reasons (same
environment)
→ drawback: their risk factor distributions may not reflect that of the source population from which the cases
emerged
→ advantage: efficiency
Cross-sectional studies
- Measure all kind of exposures and all kind of diseases
- You can divide the population in people who are exposed to a certain factor and those who have the
disease where you are interested in
- Measures exposure and disease at the same moment in time
, Experimental study/trial
- The investigator assigns the exposure (randomly)
Epidemiological effect measures
Relationship between incidence, prevalence and disease duration
- Incidence = new cases
- Prevalence varies directly with both incidence and duration
o If incidence is low, and duration is long (chronic), prevalence will be large in relation to
incidence
o If the duration is short (due to recovery, migration or death),
prevalence will be small in relation to incidence
- Duration
Use of incidence and prevalence
1. Incidence is generally used for acutely acquired disease, prevalence is used for more permanent
states, conditions or attributes of ill-health
2. Incidence is more important when thinking of etiology of the disorder, prevalence when thinking of
societal burden of the disorder including the costs and resources consumed as a result of the disorder
Special types of incidence and prevalence measures
- Ways to express incidences: morbidity rate, mortality
rate, case-fatality rate, attack rate
- Ways to express prevalences: disease rate at autopsy,
birth defect rate
Incidence
- Counting new cases (e.g. 1 million)
- Measures how fast people are ‘catching’ the disease over a period in time
o E.g. in one month/year
- Expressed as:
o Proportion of people who develop the disease (e.g. 0.05 or 5%)
o Rate at which new cases of a disease have occurred (e.g. 5000 new cases per 100000 persons
per year)
