Summary GZW3024
Epidemiology: measures of disease frequency
Prevalence, incidence and mortality
1. Prevalence = number of cases within a population at a certain moment
a. Point prevalence = number of cases within a population at a certain point in time
𝑁 𝑒𝑥𝑖𝑠𝑡𝑖𝑛𝑔 𝑐𝑎𝑠𝑒𝑠
𝑁 𝑡𝑜𝑡𝑎𝑙 𝑟𝑖𝑠𝑘
b. Period prevalence = number of cases within a population within a time period
𝑁 𝑒𝑥𝑖𝑠𝑡𝑖𝑛𝑔 𝑐𝑎𝑠𝑒𝑠 𝛥𝑡𝑖𝑚𝑒
𝑁 𝑡𝑜𝑡𝑎𝑙 𝛥𝑡𝑖𝑚𝑒
c. Lifetime prevalence = part of the population that was diseased during life
𝑁 𝑒𝑣𝑒𝑟 ℎ𝑎𝑑 𝑡ℎ𝑒 𝑑𝑖𝑠𝑒𝑎𝑠𝑒
𝑁 𝑡𝑜𝑡𝑎𝑙
o Prevalence = incidence * average disease duration
2. Incidence = number of new cases within a population
a. Cumulative incidence (CI) = absolute risk of disease within a given population
▪ Only for closed population/cohort -> membership in cohort is of unlimited
duration
• Complete follow-up required
▪ Unit: %, proportion
▪ All members of population are at risk at T0 -> exclude all prevalent cases from
population at risk at T0
𝑁 𝑛𝑒𝑤 𝑐𝑎𝑠𝑒𝑠
▪
𝑁 𝑑𝑖𝑠𝑒𝑎𝑠𝑒 𝑓𝑟𝑒𝑒 𝑎𝑡 𝑠𝑡𝑎𝑟𝑡 𝑜𝑓 𝑓𝑜𝑙𝑙𝑜𝑤 𝑢𝑝
b. Incidence rate/density (ID) = the pace at which new cases occur within a population
▪ For both open/dynamic and closed populations -> membership in open population
is of variable duration and ends as soon as someone is no longer in a certain
condition
• No complete follow-up required
▪ Unit: cases per year
▪ Person-time = sum of disease-free time of each individual within the population
at risk
𝑁 𝑛𝑒𝑤 𝑐𝑎𝑠𝑒𝑠
▪
𝛴 𝑃𝑒𝑟𝑠𝑜𝑛 𝑡𝑖𝑚𝑒
3. Mortality = number of deaths during a certain period in a certain population (incidence of death)
o Usually expressed as per 100.000
𝑁 𝑑𝑒𝑎𝑡ℎ𝑠
o
𝑁 𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛
4. Case fatality rate = proportion of incident patients with a certain condition who die because of
that condition
o Reflects severity of the disease and the effectiveness of the provided care
𝑁 𝑑𝑒𝑎𝑡ℎ𝑠 𝑑𝑢𝑒 𝑡𝑜 𝑐𝑒𝑟𝑡𝑎𝑖𝑛 𝑐𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛
o
𝑁 𝑖𝑛𝑐𝑖𝑑𝑒𝑛𝑡 𝑝𝑎𝑡𝑖𝑒𝑛𝑡𝑠 𝑤𝑖𝑡ℎ 𝑐𝑒𝑟𝑡𝑎𝑖𝑛 𝑐𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛
Other measures of disease frequency
1. R0 = basic reproduction number = number of secondary
infections produced by each infected individual in a completely
susceptible population
2. Attack rate = % chance a given contact will get the disease
o Attack rate of 60% means that a person will infect 3
people in a population of 5
▪ R0 = 3
• R0 > 1 -> disease spreads exponentially
• R0 < 1 -> disease extinguishes exponentially
3. R1 = average number of people who will catch the disease from a single infected person given
acquired immunity
o Determinants of Rt: BcD
▪ B = attack rate
▪ C = constant rate (number/unit of time)
▪ D = duration of infectivity (time)
, o Rt > 1 -> exponential growth
o Rt < 1 -> exponential shrinkage
4. Infection fatality ratio (IFR) = proportion of deaths due to infection among all infected
individuals
Epidemiology: observational study designs
Types of epidemiology
1. Descriptive epidemiology: who is diseased, is the disease occurrence stable or changing?
2. Analytical epidemiology: examines cause-effect relationships (observational) and what
determines disease occurrence?
3. Experimental epidemiology: examines cause-effect relationships (experimental)
Study designs
Study design = a specific plan or protocol for conducting a
scientific investigation, which enables the researcher to:
1. Translate the conceptual hypothesis into an operational
one
2. Statistically test the formulated hypothesis
➔ Deduction = from theory (general) to hypothesis (specific)
➔ Induction = developing theories/generalizations based on
specific observations/data
Experimental designs Non-experimental/observational designs
1. Randomized 1. Population level:
control trial 1. Ecological study: correlation between exposure and disease
(RCT) on population level
▪ Hypothesis generating
▪ Advantages: easy to perform, low cost
▪ Disadvantage: uncertain cause-effect relationship,
regional differences in diagnostics and exposure
measurements, no additional information on
confounders, ecological fallacy (data on population
level cannot be translated into individual level)
2. Individual level:
1. Case report/series looks at cases of a specific disease
▪ No measures of association
2. Cross-sectional study: examining exposure and disease
status at the same time
▪ Advantage: relatively low cost, easy to perform (only
1 measurement moment), ideal for diagnostic studies,
data on individual level
▪ Disadvantages: no cause-effect relationship
3. Case-control study: selecting a specific group of patients and
a control group without the disease from a valid source.
Measure exposure in both groups and compare them to see
whether there is an association between exposure and
disease:
▪ Advantages: data on individual level, suitable for
studying rare diseases, efficient compared to cohort,
easier to study multiple exposures (hypothesis
generating)
▪ Disadvantages: difficult to select control group from
population at risk to develop disease (solution: select
controls from base-population that produced the cases
e.g. hospital or friend-control group), not suitable to
study rare exposures, sensitive to bias (recall and
representation bias), odds ratios are more difficult to
, interpret, only 1 outcome can be studied, biomarkers
difficult to use
4. Cohort study
i. Prospective: cohort at certain moment in
time -> first/baseline measurement ->
follow cohort during certain period ->
determine who has been exposed and who
not and who has developed a certain
disease or not.
ii. Retrospective: looking at retrospective
data within a specific cohort
▪ Advantages: data on individual level, cause-effect
relationship, study multiple diseases in relation to
exposure(s)
▪ Disadvantages: costs and organization, rare disease
needs a huge cohort (feasible?)
Epidemiology: measures of association
Cohort studies Case control studies
1. Relative risk (RR)/incidence density ratio ➔ Cases and controls are selected, and
𝐼𝑛𝑐𝑖𝑑𝑒𝑛𝑐𝑒 𝑒𝑥𝑝𝑜𝑠𝑒𝑑 cases have not occurred during the study
𝑅𝑅 =
𝐼𝑛𝑐𝑖𝑑𝑒𝑛𝑐𝑒 𝑛𝑜𝑛 𝑒𝑥𝑝𝑜𝑠𝑒𝑑 -> you cannot calculate cumulative
o RR > 1-> determinant is risk incidence or incidence density or calculate
factor RR
o RR = 1 -> determinant has no 1. Odds ratio (OR) = an estimate of RR
effect (𝐶𝑎𝑠𝑒𝑠 𝑒𝑥𝑝 ∗ 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑠 𝑛𝑜𝑛 𝑒𝑥𝑝 )
𝑂𝑅 =
o RR < 1 = determinant is protective (𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠 𝑒𝑥𝑝 ∗ 𝑐𝑎𝑠𝑒𝑠 𝑛𝑜𝑛 𝑒𝑥𝑝 )
factor o OR > 1-> determinant is risk
2. Attributive risk/risk difference/excess risk factor
𝐴𝑅 = 𝑖𝑛𝑐𝑖𝑑𝑒𝑛𝑐𝑒 𝑒𝑥𝑝 − 𝑖𝑛𝑐𝑖𝑑𝑒𝑛𝑐𝑒 𝑛𝑜𝑛 𝑒𝑥𝑝 o OR = 1 -> determinant has no
3. Attributable proportion among exposed effect
𝐼𝑛𝑐𝑖𝑑𝑒𝑛𝑐𝑒 𝑒𝑥𝑝 − 𝑖𝑛𝑐𝑖𝑑𝑒𝑛𝑐𝑒 𝑛𝑜𝑛 𝑒𝑥𝑝 o OR < 1 = determinant is protective
𝐴𝑃𝑒 =
𝐼𝑛𝑐𝑖𝑑𝑒𝑛𝑐𝑒 𝑒𝑥𝑝𝑜𝑠𝑒𝑑 factor
4. Attributable proportion among total 2. Probability
population 𝑁 𝑐𝑎𝑠𝑒𝑠
𝐼𝑛𝑐𝑖𝑑𝑒𝑛𝑐𝑒 𝑡𝑜𝑡𝑎𝑙 − 𝑖𝑛𝑐𝑖𝑑𝑒𝑛𝑐𝑒 𝑛𝑜𝑛 𝑒𝑥𝑝 𝑃=
𝑁 𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛
𝐴𝑃𝑡 =
𝐼𝑛𝑐𝑖𝑑𝑒𝑛𝑐𝑒 𝑡𝑜𝑡𝑎𝑙
o Odds:
𝑝𝑟𝑜𝑏𝑎𝑏𝑖𝑙𝑖𝑡𝑦
𝑂𝑑𝑑𝑠 =
1 − 𝑝𝑟𝑜𝑏𝑎𝑏𝑖𝑙𝑖𝑡𝑦
➔ For cohort studies both RR and OR can be calculated, for case control studies only OR
Epidemiology: bias
Validity and bias
, Much scientific research aims to investigate cause-effect relations -> high quality research is needed
➔ High quality = research that is internally valid = research designed,
conducted and analyzed to minimize potential errors
o Internal validity = degree to which study findings represent
(unknown) truth in target population
▪ Is the observed association in sample equal to true
association in population?
Threats to validity
Errors in design, conduct and analysis threaten validity of study findings:
➔ Precision = repeated measures give approximately the same results
o Smaller sample -> more chance of variation between sample
results
o Random errors threaten precision, due to random differences
between study findings -> lack of precision = sampling error
➔ Validity = measure what you want to measure
o Systematic errors threaten accuracy due to systematic
deviation of study findings from truth -> lack of accuracy =
bias
Bias comes to expression in measures of association
➔ Bias = observed association deviates from true association
o Goes into certain direction and has certain magnitude
▪ Defined relative to null value of association measure:
• RR = 1 -> no association
• Regression coefficient = 0 -> no association
o Value of no association/effect = value
under H0
Three main sources of bias:
1. Selection bias = systematic errors during selection and/or follow-up of study population ->
results in the study population misrepresent target population
o Can be caused by researcher and/or participants
o Susceptibility for selection bias depends on study
design:
▪ Case-control and cross-sectional studies are
susceptible upon participant inclusion
▪ Experimental and cohort studies are susceptible
during follow-up
o Needs to be controlled during design and conduct of the
study
▪ Cannot be corrected anymore in data analysis,
but can be combatted through e.g. random
selection, maximizing response, minimizing
attrition, randomization
o Target population is defined by inclusion and exclusion
criteria and is the population on which research question is focused
▪ Study population = sample from target population -> must be representative of
target population
Epidemiology: measures of disease frequency
Prevalence, incidence and mortality
1. Prevalence = number of cases within a population at a certain moment
a. Point prevalence = number of cases within a population at a certain point in time
𝑁 𝑒𝑥𝑖𝑠𝑡𝑖𝑛𝑔 𝑐𝑎𝑠𝑒𝑠
𝑁 𝑡𝑜𝑡𝑎𝑙 𝑟𝑖𝑠𝑘
b. Period prevalence = number of cases within a population within a time period
𝑁 𝑒𝑥𝑖𝑠𝑡𝑖𝑛𝑔 𝑐𝑎𝑠𝑒𝑠 𝛥𝑡𝑖𝑚𝑒
𝑁 𝑡𝑜𝑡𝑎𝑙 𝛥𝑡𝑖𝑚𝑒
c. Lifetime prevalence = part of the population that was diseased during life
𝑁 𝑒𝑣𝑒𝑟 ℎ𝑎𝑑 𝑡ℎ𝑒 𝑑𝑖𝑠𝑒𝑎𝑠𝑒
𝑁 𝑡𝑜𝑡𝑎𝑙
o Prevalence = incidence * average disease duration
2. Incidence = number of new cases within a population
a. Cumulative incidence (CI) = absolute risk of disease within a given population
▪ Only for closed population/cohort -> membership in cohort is of unlimited
duration
• Complete follow-up required
▪ Unit: %, proportion
▪ All members of population are at risk at T0 -> exclude all prevalent cases from
population at risk at T0
𝑁 𝑛𝑒𝑤 𝑐𝑎𝑠𝑒𝑠
▪
𝑁 𝑑𝑖𝑠𝑒𝑎𝑠𝑒 𝑓𝑟𝑒𝑒 𝑎𝑡 𝑠𝑡𝑎𝑟𝑡 𝑜𝑓 𝑓𝑜𝑙𝑙𝑜𝑤 𝑢𝑝
b. Incidence rate/density (ID) = the pace at which new cases occur within a population
▪ For both open/dynamic and closed populations -> membership in open population
is of variable duration and ends as soon as someone is no longer in a certain
condition
• No complete follow-up required
▪ Unit: cases per year
▪ Person-time = sum of disease-free time of each individual within the population
at risk
𝑁 𝑛𝑒𝑤 𝑐𝑎𝑠𝑒𝑠
▪
𝛴 𝑃𝑒𝑟𝑠𝑜𝑛 𝑡𝑖𝑚𝑒
3. Mortality = number of deaths during a certain period in a certain population (incidence of death)
o Usually expressed as per 100.000
𝑁 𝑑𝑒𝑎𝑡ℎ𝑠
o
𝑁 𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛
4. Case fatality rate = proportion of incident patients with a certain condition who die because of
that condition
o Reflects severity of the disease and the effectiveness of the provided care
𝑁 𝑑𝑒𝑎𝑡ℎ𝑠 𝑑𝑢𝑒 𝑡𝑜 𝑐𝑒𝑟𝑡𝑎𝑖𝑛 𝑐𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛
o
𝑁 𝑖𝑛𝑐𝑖𝑑𝑒𝑛𝑡 𝑝𝑎𝑡𝑖𝑒𝑛𝑡𝑠 𝑤𝑖𝑡ℎ 𝑐𝑒𝑟𝑡𝑎𝑖𝑛 𝑐𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛
Other measures of disease frequency
1. R0 = basic reproduction number = number of secondary
infections produced by each infected individual in a completely
susceptible population
2. Attack rate = % chance a given contact will get the disease
o Attack rate of 60% means that a person will infect 3
people in a population of 5
▪ R0 = 3
• R0 > 1 -> disease spreads exponentially
• R0 < 1 -> disease extinguishes exponentially
3. R1 = average number of people who will catch the disease from a single infected person given
acquired immunity
o Determinants of Rt: BcD
▪ B = attack rate
▪ C = constant rate (number/unit of time)
▪ D = duration of infectivity (time)
, o Rt > 1 -> exponential growth
o Rt < 1 -> exponential shrinkage
4. Infection fatality ratio (IFR) = proportion of deaths due to infection among all infected
individuals
Epidemiology: observational study designs
Types of epidemiology
1. Descriptive epidemiology: who is diseased, is the disease occurrence stable or changing?
2. Analytical epidemiology: examines cause-effect relationships (observational) and what
determines disease occurrence?
3. Experimental epidemiology: examines cause-effect relationships (experimental)
Study designs
Study design = a specific plan or protocol for conducting a
scientific investigation, which enables the researcher to:
1. Translate the conceptual hypothesis into an operational
one
2. Statistically test the formulated hypothesis
➔ Deduction = from theory (general) to hypothesis (specific)
➔ Induction = developing theories/generalizations based on
specific observations/data
Experimental designs Non-experimental/observational designs
1. Randomized 1. Population level:
control trial 1. Ecological study: correlation between exposure and disease
(RCT) on population level
▪ Hypothesis generating
▪ Advantages: easy to perform, low cost
▪ Disadvantage: uncertain cause-effect relationship,
regional differences in diagnostics and exposure
measurements, no additional information on
confounders, ecological fallacy (data on population
level cannot be translated into individual level)
2. Individual level:
1. Case report/series looks at cases of a specific disease
▪ No measures of association
2. Cross-sectional study: examining exposure and disease
status at the same time
▪ Advantage: relatively low cost, easy to perform (only
1 measurement moment), ideal for diagnostic studies,
data on individual level
▪ Disadvantages: no cause-effect relationship
3. Case-control study: selecting a specific group of patients and
a control group without the disease from a valid source.
Measure exposure in both groups and compare them to see
whether there is an association between exposure and
disease:
▪ Advantages: data on individual level, suitable for
studying rare diseases, efficient compared to cohort,
easier to study multiple exposures (hypothesis
generating)
▪ Disadvantages: difficult to select control group from
population at risk to develop disease (solution: select
controls from base-population that produced the cases
e.g. hospital or friend-control group), not suitable to
study rare exposures, sensitive to bias (recall and
representation bias), odds ratios are more difficult to
, interpret, only 1 outcome can be studied, biomarkers
difficult to use
4. Cohort study
i. Prospective: cohort at certain moment in
time -> first/baseline measurement ->
follow cohort during certain period ->
determine who has been exposed and who
not and who has developed a certain
disease or not.
ii. Retrospective: looking at retrospective
data within a specific cohort
▪ Advantages: data on individual level, cause-effect
relationship, study multiple diseases in relation to
exposure(s)
▪ Disadvantages: costs and organization, rare disease
needs a huge cohort (feasible?)
Epidemiology: measures of association
Cohort studies Case control studies
1. Relative risk (RR)/incidence density ratio ➔ Cases and controls are selected, and
𝐼𝑛𝑐𝑖𝑑𝑒𝑛𝑐𝑒 𝑒𝑥𝑝𝑜𝑠𝑒𝑑 cases have not occurred during the study
𝑅𝑅 =
𝐼𝑛𝑐𝑖𝑑𝑒𝑛𝑐𝑒 𝑛𝑜𝑛 𝑒𝑥𝑝𝑜𝑠𝑒𝑑 -> you cannot calculate cumulative
o RR > 1-> determinant is risk incidence or incidence density or calculate
factor RR
o RR = 1 -> determinant has no 1. Odds ratio (OR) = an estimate of RR
effect (𝐶𝑎𝑠𝑒𝑠 𝑒𝑥𝑝 ∗ 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑠 𝑛𝑜𝑛 𝑒𝑥𝑝 )
𝑂𝑅 =
o RR < 1 = determinant is protective (𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠 𝑒𝑥𝑝 ∗ 𝑐𝑎𝑠𝑒𝑠 𝑛𝑜𝑛 𝑒𝑥𝑝 )
factor o OR > 1-> determinant is risk
2. Attributive risk/risk difference/excess risk factor
𝐴𝑅 = 𝑖𝑛𝑐𝑖𝑑𝑒𝑛𝑐𝑒 𝑒𝑥𝑝 − 𝑖𝑛𝑐𝑖𝑑𝑒𝑛𝑐𝑒 𝑛𝑜𝑛 𝑒𝑥𝑝 o OR = 1 -> determinant has no
3. Attributable proportion among exposed effect
𝐼𝑛𝑐𝑖𝑑𝑒𝑛𝑐𝑒 𝑒𝑥𝑝 − 𝑖𝑛𝑐𝑖𝑑𝑒𝑛𝑐𝑒 𝑛𝑜𝑛 𝑒𝑥𝑝 o OR < 1 = determinant is protective
𝐴𝑃𝑒 =
𝐼𝑛𝑐𝑖𝑑𝑒𝑛𝑐𝑒 𝑒𝑥𝑝𝑜𝑠𝑒𝑑 factor
4. Attributable proportion among total 2. Probability
population 𝑁 𝑐𝑎𝑠𝑒𝑠
𝐼𝑛𝑐𝑖𝑑𝑒𝑛𝑐𝑒 𝑡𝑜𝑡𝑎𝑙 − 𝑖𝑛𝑐𝑖𝑑𝑒𝑛𝑐𝑒 𝑛𝑜𝑛 𝑒𝑥𝑝 𝑃=
𝑁 𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛
𝐴𝑃𝑡 =
𝐼𝑛𝑐𝑖𝑑𝑒𝑛𝑐𝑒 𝑡𝑜𝑡𝑎𝑙
o Odds:
𝑝𝑟𝑜𝑏𝑎𝑏𝑖𝑙𝑖𝑡𝑦
𝑂𝑑𝑑𝑠 =
1 − 𝑝𝑟𝑜𝑏𝑎𝑏𝑖𝑙𝑖𝑡𝑦
➔ For cohort studies both RR and OR can be calculated, for case control studies only OR
Epidemiology: bias
Validity and bias
, Much scientific research aims to investigate cause-effect relations -> high quality research is needed
➔ High quality = research that is internally valid = research designed,
conducted and analyzed to minimize potential errors
o Internal validity = degree to which study findings represent
(unknown) truth in target population
▪ Is the observed association in sample equal to true
association in population?
Threats to validity
Errors in design, conduct and analysis threaten validity of study findings:
➔ Precision = repeated measures give approximately the same results
o Smaller sample -> more chance of variation between sample
results
o Random errors threaten precision, due to random differences
between study findings -> lack of precision = sampling error
➔ Validity = measure what you want to measure
o Systematic errors threaten accuracy due to systematic
deviation of study findings from truth -> lack of accuracy =
bias
Bias comes to expression in measures of association
➔ Bias = observed association deviates from true association
o Goes into certain direction and has certain magnitude
▪ Defined relative to null value of association measure:
• RR = 1 -> no association
• Regression coefficient = 0 -> no association
o Value of no association/effect = value
under H0
Three main sources of bias:
1. Selection bias = systematic errors during selection and/or follow-up of study population ->
results in the study population misrepresent target population
o Can be caused by researcher and/or participants
o Susceptibility for selection bias depends on study
design:
▪ Case-control and cross-sectional studies are
susceptible upon participant inclusion
▪ Experimental and cohort studies are susceptible
during follow-up
o Needs to be controlled during design and conduct of the
study
▪ Cannot be corrected anymore in data analysis,
but can be combatted through e.g. random
selection, maximizing response, minimizing
attrition, randomization
o Target population is defined by inclusion and exclusion
criteria and is the population on which research question is focused
▪ Study population = sample from target population -> must be representative of
target population
