Lecture 1
What is a variable:
- A variable has a value, and varies
- Epidemiological variables aid in the depiction, analysis and interpretation of difference in disease
patterns
- “The first question for the epidemiologist, in any investigation, is the nature and validity of the
definition of the disease or other problem under investigation.”
- A good epidemiological variable should:
- Have an impact on health in individuals and populations
- Be measurable accurately
- Differentiate populations in their experience of disease or health
- Differentiate populations in some underlying characteristic relevant to health e.g.,
income, childhood circumstance, hormonal status, genetic inheritance, or
behavior relevant to health
- Generate testable aetiological hypotheses, and/ or
- Help to develop health policy, and/ or
- Help to plan and deliver health care, and/ or
- Help to prevent and control disease
- E.g., would hair color be a good epidemiological variable? ⇒ most of the time
no, but depending on the context, people with red hair are more prone to sun
burn/ skin cancer
Why study disease frequency?
- To help control an abrupt rise in disease incidence
- To understand the factors which influence disease frequency
- To develop health policy and plans
- “The key strategy in epidemiology is to seek out and quantify disease variation, and then develop
and test hypotheses to understand the causal mechanisms which led to it”.
What is true for the population may not be true for the individual (and the other way around)
- How to unravel the interplay between
- Environment
- Genes
- (Bad) luck
- Nature vs nurture
- People from within countries are more similar than people between countries
- Living in certain towns might affect you differently
- People within families are more similar
- Even if you want to focus on country differences, it might be important to realize the other factors
that play a role
,Demographic changes in locations and time
- Characteristics of people alter (individual changes throughout time)
- Genetic inheritance
- Behavior
- Lifestyle
- Characteristics of the environment alters
- Location
- Quality of living environment
- Health care
- Major events (war, poverty, disasters)
- Shifts due to age (e.g., longer lives, lower birthrates) or migration
- Cultural changes (e.g., men vs women, wealth distribution, cultural beliefs or habits)
- Migration and integration
,Is variation real, or artefact?
- No variation means no association
- If you want to test something with age and you only have young people, you’re just going to find
whatever association that is present
- You first want to see if there is an association ⇒ exploratory data analysis
- If no, you can still model, but probably won’t be valuable
- If yes, you can make hypothesis
- Then try to find out if the association is real, if so, look at the underpinnings
Step 1: Demonstrate a (real) variance or association
- Reasons why variations can be illusions:
- Chance: the numbers of cases are randomly fluctuating over time
- Errors of observation: biased techniques are the most common reason for making
erroneous observations
- Changes in the size and structure of the population from which the cases arose
- The likelihood of people seeking health care and hence being diagnosed and eventually
counted in statistics
- This varies with their level of knowledge
- Expectations
- Accessibility and acceptability of health care
- Reasons why variations can be illusions:
- The likelihood of the correct diagnosis being reached
- Dependent on availability and use of medical care
- The level of skill of the doctor
- The quality of the diagnostic facilities
- Changes:
- In the clinical approach to diagnosis
- E.g., dependent on changing medical trends
- In data collection methods
- In the way diseases are diagnostically coded
, - In the way data are analyzed and presented
- E.g., altering the “standard” population used in adjusting disease rates
for differences in age and sex can spuriously (false/ inauthentic) alter
disease incidence
- Different populations, different SDs/ mean
Error, bias, and confounders
Random error:
- What is an error?
- Random error
- Given that the measure is valid and accurate and with proper sensitivity and validity
- Can be due to (changes/ differences in)
- Human factors:
- Techniques
- Labs
- Skillsets
- Effort
- Coding errors
- Typos (e.g., in statistics)
- Bad days?
- Machine/ technical problems
- Error is not tied to anything in particular ⇒ probably won’t impact your
association
- If there is any type of randomisation, they will unlikely be all in the same group
Bias:
- An error caused by systematically favoring some outcomes over others
- Affects populations or study subgroups unequally or
- Inappropriate generalization
- Best known from but not limited to a preference or an inclination;
- Especially one that inhibits impartial judgment or that leads to an unfair act or policy
- May stem from prejudice
What is a variable:
- A variable has a value, and varies
- Epidemiological variables aid in the depiction, analysis and interpretation of difference in disease
patterns
- “The first question for the epidemiologist, in any investigation, is the nature and validity of the
definition of the disease or other problem under investigation.”
- A good epidemiological variable should:
- Have an impact on health in individuals and populations
- Be measurable accurately
- Differentiate populations in their experience of disease or health
- Differentiate populations in some underlying characteristic relevant to health e.g.,
income, childhood circumstance, hormonal status, genetic inheritance, or
behavior relevant to health
- Generate testable aetiological hypotheses, and/ or
- Help to develop health policy, and/ or
- Help to plan and deliver health care, and/ or
- Help to prevent and control disease
- E.g., would hair color be a good epidemiological variable? ⇒ most of the time
no, but depending on the context, people with red hair are more prone to sun
burn/ skin cancer
Why study disease frequency?
- To help control an abrupt rise in disease incidence
- To understand the factors which influence disease frequency
- To develop health policy and plans
- “The key strategy in epidemiology is to seek out and quantify disease variation, and then develop
and test hypotheses to understand the causal mechanisms which led to it”.
What is true for the population may not be true for the individual (and the other way around)
- How to unravel the interplay between
- Environment
- Genes
- (Bad) luck
- Nature vs nurture
- People from within countries are more similar than people between countries
- Living in certain towns might affect you differently
- People within families are more similar
- Even if you want to focus on country differences, it might be important to realize the other factors
that play a role
,Demographic changes in locations and time
- Characteristics of people alter (individual changes throughout time)
- Genetic inheritance
- Behavior
- Lifestyle
- Characteristics of the environment alters
- Location
- Quality of living environment
- Health care
- Major events (war, poverty, disasters)
- Shifts due to age (e.g., longer lives, lower birthrates) or migration
- Cultural changes (e.g., men vs women, wealth distribution, cultural beliefs or habits)
- Migration and integration
,Is variation real, or artefact?
- No variation means no association
- If you want to test something with age and you only have young people, you’re just going to find
whatever association that is present
- You first want to see if there is an association ⇒ exploratory data analysis
- If no, you can still model, but probably won’t be valuable
- If yes, you can make hypothesis
- Then try to find out if the association is real, if so, look at the underpinnings
Step 1: Demonstrate a (real) variance or association
- Reasons why variations can be illusions:
- Chance: the numbers of cases are randomly fluctuating over time
- Errors of observation: biased techniques are the most common reason for making
erroneous observations
- Changes in the size and structure of the population from which the cases arose
- The likelihood of people seeking health care and hence being diagnosed and eventually
counted in statistics
- This varies with their level of knowledge
- Expectations
- Accessibility and acceptability of health care
- Reasons why variations can be illusions:
- The likelihood of the correct diagnosis being reached
- Dependent on availability and use of medical care
- The level of skill of the doctor
- The quality of the diagnostic facilities
- Changes:
- In the clinical approach to diagnosis
- E.g., dependent on changing medical trends
- In data collection methods
- In the way diseases are diagnostically coded
, - In the way data are analyzed and presented
- E.g., altering the “standard” population used in adjusting disease rates
for differences in age and sex can spuriously (false/ inauthentic) alter
disease incidence
- Different populations, different SDs/ mean
Error, bias, and confounders
Random error:
- What is an error?
- Random error
- Given that the measure is valid and accurate and with proper sensitivity and validity
- Can be due to (changes/ differences in)
- Human factors:
- Techniques
- Labs
- Skillsets
- Effort
- Coding errors
- Typos (e.g., in statistics)
- Bad days?
- Machine/ technical problems
- Error is not tied to anything in particular ⇒ probably won’t impact your
association
- If there is any type of randomisation, they will unlikely be all in the same group
Bias:
- An error caused by systematically favoring some outcomes over others
- Affects populations or study subgroups unequally or
- Inappropriate generalization
- Best known from but not limited to a preference or an inclination;
- Especially one that inhibits impartial judgment or that leads to an unfair act or policy
- May stem from prejudice
