Intro to Epidemiology Flashcards
Epidemiology
The study of the distribution and determinants of health-related states or events (including disease), and the application of this study to the control of diseases and other health problems (WHO, 2024).
The study of the distribution and determinants of disease in occurrence in human populations.
- outcome of interest
-exposure of interest
-biostatistical methods
Study Base
-Reference population
-Source of the study population
-Population giving rise to the cases
-Defined before cases appear by a geographical area or some other entity like a cohort study
Person time
-Estimate of the actual time-at-risk in years, months, or days that all persons contributed to a study
-Time will differ by people depending on endpoint: disease, death, end of follow-up
Exposure
An exposure, risk factor, or other characteristic being observed or measured that is hypothesised to influence an event or manifestation.
Outcome
-Disease
-Disease progression
-Death
-Comorbidity
-Questionnaire data
-Biological endpoints – expression levels
- The result, or effect, of an action, situation, or event.
- Typical outcomes investigated in medical research include:
1.Mortality (how long did the patient survive after the onset of disease.)*
2.Progression (did the disease progress, and how long after onset did this progression occur).*
3.Morbidity (did the disease result in the occurrence of further illness or disease).*
Prevalence
Proportion of a population found to have a condition at a specific point in time.
Incidence
There were about 357,000 new cases of cancer in the UK on average each year.
Risk
Probability of disease developing in an individual in a specified time interval.
Measurement of effect
Relative: exposed versus unexposed
Absolute: i.e. incidence, prevalence
Thing to consider when designing a research study
- research question (what is the research question?)
-study design (how will you undertake the study?)
-subjects (who are the subjects? how will they be selected?)
-data (what data needs to be collected? what kind of measures?)
-analysis: what analyses do you need to undertake?
-interpretation: what do the results mean? how valid are they?
Quantitative study
:Uses numbers; exposures and outcomes are measurable
-How many?
-Who is at risk?
-What causes this disease?
-Is there an improvement?
Qualitative
Uses words; stories, experiences, observations
-Why do people do ……?
-How do they feel about…?
-What is their experience of…?
Observational vs Experimental
Does the researcher control the exposure?
yes: experimental
no: observational
Do you want to determine if something is “causal”?
yes: experimental
no: observational
Types of observational studies
Descriptive: used to formulate a certain hypothesis.
Examples: case-studies; cross-sectional studies, ecological studies
Analytical: used to test hypotheses:
Examples: case-control, cohort
How do you choose between observational methods?
Depends on:
- how rare the outcome is
- what data exists for the population of interest
- whether the temporal relationship is important
i.e. Exposure –> Outcome (cause –> effect)
- how quickly you want the answer
money / resources
Ecological studies
Studies that investigate risk factors of health outcomes in which the unit of analysis is at the group level rather than the individual.
Group measures (exposure and or outcome) can include:
- summary measures of a group (mean, average rate)
- environmental factors (air pollution, hours of sun-light, fast-food shops)
i.e. something that is not measured at the individual level
Examples:
Time trends, geographic comparisons
What are the advantages and disadvantages of ecological studies?
Advantages:
- Easy to do
- No individual data necessary
- Good to generate ideas about potential associations
Disadvantages:
- No information on the individual level
- Not able to account for other factors that might explain the association –> ecological fallacy
Ecological fallacy
The ecological fallacy occurs where an analysis of group data is used to draw conclusions about the individual.
Example:
The average salary is higher in countries that sell more reading glasses.
Therefore if you wear reading glasses you are likely to have a higher salary.
Likely to be due to other factors that are not taken into account (confounders)
Cross-sectional study design process and example
Select a sample (representing the population of interest)
Measure exposure and outcome variables at the same time
Determine prevalence
Data is collected at a single point in time from a population or a representative sample. It provides a snapshot of the prevalence of certain outcomes and exposures at a specific moment.
Participants are selected based on certain characteristics (e.g., age, gender, location) and data is collected through surveys, interviews, or examinations at one specific time point.
What is the prevalence of smoking among college students?
Study Design: Researchers survey 500 college students from various universities about their smoking habits during a single time period. They collect data on whether the students smoke, how many cigarettes they smoke per day, and other relevant factors.
Outcome: The researchers analyse the survey responses to determine the proportion of college students who smoke and explore any associations between smoking and other variables, such as gender, age, or academic performance. However, since this is a cross-sectional study, they cannot determine whether smoking causes certain outcomes or vice versa; it only provides a snapshot of smoking prevalence among college students at that particular time.
Cross-sectional studies strengths and weaknesses
Strengths
- fast and inexpensive
- immediate answers – no follow-up time
- no loss to follow-up (but can have non-responders)
Weaknesses
- can’t determine temporal relationship
- not good for rare exposures or outcomes
- bias can be a problem - measurement bias, survivor bias
Cohort study design process and example
- Start with the POPULATION of interest
- Identify or assemble a cohort
- Measure risk factor(s) and potential confounders
- Measure the outcome over the follow-up period
A group of individuals, known as a cohort, is followed over a period of time to observe and analyse outcomes related to exposure to certain factors.
Participants are initially identified as either exposed or unexposed to a particular factor of interest. They are then followed prospectively to observe the development of outcomes.
Does regular exercise reduce the risk of heart disease?
Study Design: Researchers recruit a group of 1000 middle-aged adults who are free of heart disease. They divide them into two cohorts: one that engages in regular exercise (at least 30 minutes of moderate-intensity exercise five times a week) and one that does not. The participants are followed for 10 years, during which their incidence of heart disease is monitored.
Outcome: Researchers compare the incidence of heart disease between the two cohorts to determine if regular exercise is associated with a lower risk of heart disease.
Cohort study design can be:
Prospective: start with assembling a cohort, measure risk factors then follow over time to measure outcomes –> where data is collected forward in time.
OR
Retrospective (historical): identify a suitable cohort (from the past), collect risk factor data measured in the past, collect subsequent outcome data –> where data is collected from past records.
Cohort studies strengths (in general, then for each)
Cohort studies (in general):
- Can establish sequence of events
- Can assess risk of multiple outcomes at the same time
- Can estimate incidence (how many new events within a certain time)
- Able to directly calculate absolute and relative risk
Prospective cohort studies:
- Can control who is in the cohort
- Lower risk of bias (exposures measured before outcomes)
Retrospective cohort studies:
- More efficient: less time, less costly
Cohort studies weaknesses (in general, then for each)
Cohort studies (in general):
- not a controlled experiment so can’t claim ‘causation’
- difficult to control for all other confounding factors
- expensive – often require large sample
not good choice for rare outcomes
Prospective:
- not timely, long follow-up
- potential loss to follow-up (can lead to bias)
Retrospective:
- knowing the outcome might lead to bias
- limited to data already collected
- little control over who is in the cohort
Case-control study design process and example
Select a sample of ‘cases’ (i.e. people who have the condition/disease)
Select a sample of ‘controls’ (i.e. people without the disease but who have the same chance of having the disease)
Measure (past) exposure to risk factors of interest
Researchers start by identifying individuals with the outcome of interest (cases) and individuals without the outcome (controls). They then look back in time to compare the past exposure history of cases and controls to determine the association between exposure and outcome.
Is there an association between the consumption of red meat and the development of colorectal cancer?
Study Design:
Selection of Cases: Researchers identify 200 individuals diagnosed with colorectal cancer from hospital records or cancer registries. These individuals will be the cases in the study.
Selection of Controls: For each case, researchers select a control individual without colorectal cancer from the same population. Controls are typically matched to cases based on age, gender, and other relevant factors to minimize potential confounding variables.
Data Collection: Researchers conduct interviews or surveys with both cases and controls to gather information on their dietary habits, including the frequency and amount of red meat consumption over the past 10 years. Other potential confounding factors such as family history of colorectal cancer, smoking, and physical activity may also be collected.
Analysis: After collecting the data, researchers compare the frequency and amount of red meat consumption between cases and controls. They calculate odds ratios (ORs) or relative risks (RRs) to assess the strength of the association between red meat consumption and the risk of developing colorectal cancer. Statistical methods are used to control for potential confounding variables.
Outcome: The researchers find that individuals with colorectal cancer are more likely to have a higher intake of red meat compared to controls without colorectal cancer. After controlling for confounding factors, they determine that there is a statistically significant association between red meat consumption and the risk of developing colorectal cancer.
Case-control strengths and weaknesses
Strengths
- Useful for rare outcomes (eg specific cancers)
- Efficient/less costly than cohort study: smaller sample size; no follow-up required
Weaknesses
- Biases if cases and controls come from different populations
- Biases due to measuring exposure after the outcome
- Confounding due to other influential factors (not measured)
- Can only study one outcome
Summary of study design
- Research question determines the study design
- Source population & study population should reflect the target population
- Valid and reliable measures of exposure, outcome and confounders
- Hard to establish causation in observational studies due to bias/confounding/temporality
Internal Validity
- Validity of the inferences as they pertain to members of the source population AKA the extent to which the conclusions drawn from a study accurately represent the individuals in the original population studied
- Prerequisite for external validity AKA necessary for ensuring that the findings of a study can be generalised to other populations or contexts
External validity
Validity of the inferences as they pertain to people outside that population AKA the accuracy of the conclusions drawn from a study in relation to individuals who are not part of the original population studied.
Validity (in general)
Refers to the degree to which a study accurately measures what it claims to measure or accurately reflects the true relationship between variables.
In order for a study to be valid, it must be free from bias (validity= lack of bias)
Bias
Systematic error in the study design that results in an estimate of the association between exposure and outcome that is different from the causal association AKA An error that occurs in the way a study is designed, causing the estimated connection between an exposure and an outcome to differ from the true cause-and-effect relationship
Sources of bias
- Selection of participants
- Measurement aspects (e.g., exposure, outcome, confounders) on the selected participants
Three types of bias
- selection bias
- information bias
- confounding
Selection bias
Distortions that arise from the methods used to choose participants and from factors that affect their participation in the study.
The relation between exposure and disease differs between individuals who take part in the study and those who should theoretically qualify for the study (including those who don’t participate).
When is selection bias a problem?
- in case-control studies because the selection of cases and controls (which takes place after outcome has occurred) may be related to exposure
- can also be a problem in retrospective cohort studies
What are the two types of selection bias?
Self-selection bias:
Self-referral is considered a threat to validity since the reasons for self-referral may be associated with the outcome.
“Healthy-worker effect”
Healthy people more likely to be working whereas those who remain unemployed, retired, or are disabled, are as a group less healthy.
Diagnostic bias:
Can occur if outcome in individuals is more likely to be ascertained as a consequence of a particular exposure AKA occurs when the likelihood of diagnosing a particular outcome is influenced by the presence of a certain exposure, leading to an overestimation or underestimation of the association between exposure and outcome
Example: OC use and venous thromboembolism
- Physicians were aware of a possible relationship
- Proportion of the women in the study had been hospitalised for evaluation of this disease because they were currently taking OCs
Information bias
Arises when errors occur in the measurement of data that are to be compared between groups within a study, potentially leading to biased conclusions.
Two types of information bias
Differential misclassification
Non- differential misclassification
Differential misclassification (in general)
- Refers to errors in classification that are influenced by the values of other variables.
- More of a problem for case-control studies because exposure classification typically happens after the occurrence of the disease.
- Bias can exaggerate or underestimate an effect
Miss-classification outcome= loss to follow up
Loss to follow up:
Subjects who are sick (and thus more likely to die) may tend to drop out of the study.
If they are also persons who are less active, we would observe a dilution of the true effect of physical activity on mortality since we would observe a falsely low death rate among inactive subjects.
Misclassification of exposure= reverse causation
Reverse causation: When sick persons, who are more likely to die, decrease their activity levels because of their ill health. The observed association between low physical activity level and high mortality rates may not be due to the former causing the latter, but rather reflecting that sick persons, who are more likely to die, are less physically active.
Can be handled through appropriate selection of study population or in statistical analysis/
Differential misclassification of exposure= recall bias
Recall bias can occur in case-control studies when cases recall their past exposures differently than controls.
Amount of time lapsed between the exposure and the recall is an important indicator of the accuracy of recall – therefore, cases and controls should have same recall period.
Non-differential misclassification (in general)
Classification error that does NOT depend
on values of other variables, i.e.,
Exposure nondiff. misclassification: the proportion of subjects misclassified on exposure does not depend on disease status
Disease nondiff. misclassification: the proportion of subjects misclassified on disease does not depend on exposure
Confounding
Confounding can be considered as a confusion of effects.
It occurs when the observed impact of the main exposure under study is altered because the influence of an unrelated factor (known as a confounder) is mistaken for or combined with the true effect of the exposure, potentially resulting in biased conclusions.
Example of confounding
In men baldness is associated with the risk of myocardial infarction …
Age increases both the likelihood of baldness and the risk of myocardial infarction; thus age confounds the association between baldness and myocardial infarction
Criteria for a confounding factor
- A confounding factor must be a risk factor for the disease.
- A confounding factor must be associated with the exposure under study in the source population.
- A confounding factor must not be affected by the exposure or the disease. Cannot be an intermediate factor.
Control of confounding
Randomisation
Restriction
- Select subjects with similar values for confounders, e.g., select only old, male
- Limitation: generalisability
Matching
Stratification
- Match on confounding variable to “force” comparison between cases and controls
Modelling
Confounding by indication:
Most common type of confounding
E.g., a drug might look ineffective or even harmful because of poor outcomes among those taking that drug; this, however, this could be merely because the drug was given to highly selected individuals who needed the drug because of their poor prognosis
Also known as channeling bias or indication bias
BUT
Nevertheless, many studies report that the association of exposure and outcome persist after adjustment for the measured confounders. The association may still be the result of residual confounding as most studies do not measure all potential confounders or do not measure these confounders perfectly or do not determine whether they change across time.
For example, other diseases that may affect both physical activity and mortality, such as diabetes or depression, are often not taken into consideration.
Review of confounding
Associated with the outcome of interest independent of its relation to the exposure of interest.
Associated with the exposure of interest in the study base that produced the cases.
Cannot be an intermediate variable on the causal chain leading from the exposure of interest to the onset of disease.
Confounding can be altered at the study design stage:
Matching
Restriction
Randomisation
Effect modification
The strength of the association between the exposure and the outcome varies by levels of a third variable.
Effect modifier = Interaction Term
Refers to a situation in which two or more risk factors modify one another’s effects on an outcome.
The effect of the physical activity categories sedentary or active on the presence or absence of chronic heart disease might differ depending on some third factor, such as sex; physical activity might reduce the risk of disease among men but not in women.
Sex would be the effect modifier.
Generalisability
Can results coming from a specified population be generalized to other populations?
Is there any reason to believe that the findings would not apply to other populations?
Answers involves a judgment based on an understanding of the research hypothesis/mechanism
