EPIDEMIOLOGY –John M Last Def 2021

“The study of the distribution and determinants of health-related states or events in specified
populations, and the application of this study to control of health problems”

Making observations at the group level

• Defining characteristic/s of interest

• Measuring them in individuals

• Summarizing them for the group –we can express our findings as mathematical quantities of
disease/event frequency

• Numbers need to be related to other numbers to have meaning

Epidemiology is a population science Moving from an individual to a group perspective

Making observations at the group level

• Defining characteristic/s of interest

• Measuring them in individuals

• Summarizing them for the group –we can express our findings as mathematical quantities of
disease/event frequency

• Numbers need to be related to other numbers to have meaning

In epidemiology, random errors refer to variations in data that arise by chance, leading to deviations
from the true value. These errors can result in incorrect conclusions in research studies.
Understanding Type I (alpha) and Type II (beta) errors is crucial when interpreting the results of
epidemiological studies.

ERRORS

1. Type I Error (Alpha Error):

- Definition: A Type I error occurs when a researcher rejects the null hypothesis (which states there
is no effect or association) when it is actually true. In other words, it is a "false positive" result.

- Implication: The researcher concludes that there is an effect or association when there really isn’t
one.

- Example: In a study investigating whether a new drug reduces the incidence of a disease, a Type I
error would occur if the study concludes that the drug is effective when it actually has no effect.

- Significance Level (Alpha): The probability of making a Type I error is denoted by the alpha level
(α), which is often set at 0.05. This means there is a 5% risk of concluding that there is an effect
when there is none.

2. Type II Error (Beta Error):

- Definition: A Type II error occurs when a researcher fails to reject the null hypothesis when it is
actually false. This is known as a "false negative" result.
 - Implication: The researcher concludes that there is no effect or association when there actually is
one.

- Example: In the same drug study, a Type II error would occur if the study concludes that the drug
does not reduce disease incidence when it actually does.

- Power (1 - Beta): The probability of avoiding a Type II error is called the power of the study,
denoted by (1 - β). A study with high power (commonly 0.8 or 80%) is more likely to detect a true
effect if one exists.

Relationship Between Type I and Type II Errors:

- Balancing Errors: There is often a trade-off between Type I and Type II errors. Lowering the alpha
level (reducing the risk of Type I error) usually increases the risk of a Type II error, and vice versa. For
example, if a study uses a very stringent alpha level (e.g., 0.01), it reduces the chance of a false
positive but might increase the chance of missing a true effect.

- Contextual Considerations: The consequences of making these errors should guide the choice of
alpha and beta levels. In clinical settings, a Type I error might lead to the adoption of an ineffective
treatment, while a Type II error might result in missing out on a potentially beneficial intervention.

Practical Example:

Imagine you are studying the effect of a new vaccine on preventing a disease.

- Type I Error: You conclude that the vaccine is effective in preventing the disease, but in reality, it
has no effect.

- Type II Error: You conclude that the vaccine does not prevent the disease, but in reality, it does.

Understanding and minimizing these errors is crucial in designing reliable and valid epidemiological
studies.

In epidemiology, understanding confounding and interaction (also known as effect modification) is

critical for interpreting study results accurately. These concepts address how different variables may
influence the relationship between an exposure and an outcome.

Confounding, Collider bias, Effect modifier

• Confounder – a variable that is associated with the exposure and the outcome but is not part of
the causal pathway of the exposure and the outcome

Eg; Age, SES, Gender

• Collider – a variable that is causally influenced by both the exposure and the outcome but is not
part of the causal pathway of the exposure and the outcome

• Effect modifier - a variable that is associated with the exposure and the outcome and is part of the
causal pathway of the exposure and the outcome

1. Confounding:
 - Definition: Confounding occurs when a third variable (confounder) is related to both the exposure
and the outcome but is not on the causal pathway. The presence of a confounder can distort the
apparent relationship between the exposure and the outcome, either exaggerating or
underestimating the true effect.

- Example: Suppose a study is investigating the relationship between coffee drinking (exposure)
and heart disease (outcome). If age is not controlled for, it could confound the results because:

- Age is associated with coffee drinking (older people might drink more or less coffee).

- Age is also associated with heart disease (older people are at higher risk).

If older age leads to both higher coffee consumption and higher heart disease risk, it might falsely
suggest that coffee drinking is associated with heart disease, even if no direct relationship exists.

- Addressing Confounding: Confounding can be controlled for during study design (e.g., through
randomization, matching, or restriction) or analysis (e.g., using stratification or multivariable
adjustment methods like regression).

2. Interaction / Effect Modification:

- Definition: Interaction or effect modification occurs when the effect of the main exposure on the
outcome differs depending on the level of another variable. In other words, the relationship
between the exposure and the outcome changes based on the presence or level of a third factor
(modifier).

- Example: Consider a study examining the effect of a new drug (exposure) on reducing blood
pressure (outcome). If the drug’s effectiveness varies by sex (i.e., it works better in men than in
women), sex is an effect modifier. This means that the effect of the drug on blood pressure is
different in men compared to women.

- Detecting Interaction: Interaction is usually explored by stratifying the analysis by levels of the
potential effect modifier or by including interaction terms in statistical models. If an interaction is
present, the association between exposure and outcome should be reported separately for each
level of the modifier.

Differences Between Confounding and Interaction:

- Confounding is a bias that needs to be controlled to reveal the true relationship between
exposure and outcome. It is a problem that distorts the apparent effect.

- Interaction (Effect Modification) is a real phenomenon that should be explored and reported. It
reveals how the relationship between exposure and outcome changes across different levels of
another variable.

Practical Example Involving Both Concepts:

Suppose a study is examining the relationship between physical activity (exposure) and the risk of
developing diabetes (outcome).
- Confounding: If age is a confounder (older people are less active and more likely to develop
diabetes), not adjusting for age could falsely suggest that physical activity has a stronger or weaker
effect on diabetes risk than it actually does.

- Interaction: If the effect of physical activity on diabetes risk is stronger in individuals with a family
history of diabetes (family history being an effect modifier), this interaction should be reported
separately to provide a clearer understanding of the risk for different groups.

Conclusion:

- Confounding is an error that needs to be corrected to understand the true association between
exposure and outcome.

- Interaction (Effect Modification) is a genuine effect that, when present, offers valuable insights into
how different groups are affected differently by the exposure.

Properly addressing both confounding and interaction is essential for accurate epidemiological
research and effective public health interventions.

In epidemiology and other fields like psychology, mediators and moderators are important concepts
used to understand and analyze the relationships between variables. Though they might seem
similar, they serve different purposes in research.

1. Mediators:

- Definition: A mediator is a variable that explains the mechanism through which an independent
variable (exposure) influences a dependent variable (outcome). It acts as a pathway or a link in the
causal chain between the exposure and the outcome.

- Example: Suppose you are studying the relationship between education level (independent
variable) and health outcomes (dependent variable). Income could be a mediator if higher education
leads to better job opportunities and income, which in turn leads to better health. Here, income
explains part of the reason why higher education might lead to better health.

- Role of Mediator:

- Pathway: The mediator explains how or why an effect occurs.

- Causal Chain: The sequence is typically: Independent Variable → Mediator → Dependent

Variable.

- Analyzing Mediators: To test for mediation, researchers often use statistical methods like
mediation analysis, which can decompose the total effect of the independent variable on the
dependent variable into direct and indirect effects (the latter being the effect mediated by the
mediator).

- Diagram:

```
 Education → Income (Mediator) → Health

```

2. Moderators:

- Definition: A moderator is a variable that influences the strength or direction of the relationship
between an independent variable and a dependent variable. In other words, it affects how strong
the relationship is or whether it even exists in different circumstances or groups.

- Example: Consider a study on the effect of stress (independent variable) on mental health
(dependent variable). Social support could be a moderator if the impact of stress on mental health is
weaker for people with strong social support compared to those with weak social support. This
means the relationship between stress and mental health depends on the level of social support.

- Role of Moderator:

- Interaction: The moderator changes the effect of the independent variable on the dependent
variable.

- Context-Dependent: The effect of the independent variable varies depending on the level of the
moderator.

- Analyzing Moderators: Researchers use interaction terms in regression models or stratified

analysis to test for moderation. A significant interaction indicates that the relationship between the
independent variable and the dependent variable changes based on the level of the moderator.

- Diagram:

```

Stress → Mental Health

Social Support (Moderator)

```

Key Differences Between Mediators and Moderators:

- Purpose:

- Mediators explain the process or mechanism through which one variable affects another.

- Moderators indicate under what conditions or for whom the relationship between two variables
changes.

- Role in Analysis:
 - Mediators are part of the causal pathway, helping to understand "how" or "why" an effect occurs.

- Moderators are not part of the causal pathway but rather influence the relationship's strength or
direction, helping to understand "when" or "for whom" the effect occurs.

- Causal Sequence:

- Mediator: Independent Variable → Mediator → Dependent Variable.

- Moderator: Independent Variable → Dependent Variable, with the effect modified by the
Moderator.

Practical Example Involving Both:

Suppose a study investigates the impact of physical activity (independent variable) on body weight
(dependent variable).

- Mediator: Diet could be a mediator if increased physical activity leads to healthier eating habits,
which in turn affects body weight. Here, diet explains part of how physical activity influences weight.

- Moderator: Age could be a moderator if the effect of physical activity on body weight is stronger in
younger individuals compared to older individuals. This means the relationship between physical
activity and weight depends on the person's age.

Conclusion:

- Mediators help to uncover the mechanisms underlying a relationship.

- Moderators help to identify the conditions under which a relationship is stronger, weaker, or
different.

Both concepts are crucial for developing a deeper and more nuanced understanding of complex
relationships in research.

These terms refer to different types of study designs commonly used in epidemiology and clinical
research. Each has its specific methodology, strengths, and limitations. Let's break them down:

TRIALS
1. Uncontrolled Trials:

- Definition: An uncontrolled trial is a type of study where all participants receive the intervention,
and there is no separate control group for comparison. Researchers observe the outcomes after the
intervention without comparing them to a non-intervention group.

- Example: A study where all participants receive a new drug, and the outcomes are measured
without comparing them to a group that did not receive the drug.

- Strengths: Simpler to conduct, useful for early-phase research, or when the intervention is
expected to have a dramatic effect.

- Limitations: Lack of a control group makes it difficult to determine whether the observed
outcomes are due to the intervention or other factors (e.g., natural disease progression, placebo
effect).
2. Non-Randomized Controlled Trials:

- Definition: In a non-randomized controlled trial, participants are assigned to either the

intervention group or the control group, but the assignment is not random. Instead, allocation might
be based on certain criteria or convenience.

- Example: A study where patients in one clinic receive a new treatment, while patients in another
clinic continue with standard treatment.

- Strengths: Easier to implement than randomized trials, especially when randomization is

impractical or unethical.

- Limitations: The lack of randomization can introduce selection bias, as the groups may differ
systematically in ways other than the intervention (e.g., different demographics or baseline
characteristics).

3. Randomized Controlled Trials (RCTs):

- Definition: An RCT is a study where participants are randomly assigned to either the intervention
group or the control group. This randomization aims to ensure that the groups are comparable in all
respects except for the intervention being studied.

- Example: A clinical trial where participants are randomly assigned to receive either a new drug or
a placebo, with outcomes compared between the two groups.

- Strengths: Randomization minimizes bias, making RCTs the gold standard for assessing causality
and the effectiveness of interventions.

- Limitations: Can be complex, expensive, and time-consuming. Sometimes ethical or practical

concerns may limit the use of randomization.

4. Cluster Randomized Controlled Trials:

- Definition: In a cluster randomized controlled trial, groups or clusters (e.g., schools, communities,
hospitals) are randomized to either the intervention or control group, rather than individual
participants.

- Example: A trial where entire schools are randomized to receive either a new educational
program (intervention) or the standard curriculum (control), and student outcomes are compared.

- Strengths: Useful when individual randomization is not feasible or when interventions are
naturally administered at the group level.

- Limitations: More complex analysis is required due to potential intra-cluster correlation

(individuals within a cluster might be more similar to each other than to individuals in other
clusters).

5. Intervention Clusters and Comparison Clusters:

 - Definition: These terms are used in cluster randomized trials. Intervention clusters refer to the
groups or clusters that receive the experimental intervention, while comparison clusters (or control
clusters) are those that receive the standard treatment or no intervention.

- Example: In a study of a new public health campaign, some communities (intervention clusters)
might receive the campaign, while others (comparison clusters) do not, and outcomes like smoking
rates are compared between the two groups.

- Strengths/Limitations: These are not distinct study designs but refer to how groups are managed
within cluster randomized trials.

6. Community Intervention Trials:

- Definition: Community intervention trials, also known as community trials or field trials, are
studies where interventions are applied at the community level rather than to individuals. Entire
communities are typically assigned to intervention or control conditions, and outcomes are
measured at the population level.

- Example: A study where one town receives a water fluoridation program to prevent dental
cavities, while a similar town does not, and dental health outcomes are compared.

Eg: Helmet laws

- Strengths: These trials are valuable for studying the effectiveness of public health interventions
that are designed to influence behaviors or outcomes at the community level.

- Limitations: Controlling for confounders can be challenging, and external factors can influence the
results. There may also be ethical considerations if withholding an intervention could harm a
community.

Summary of Study Designs:

- Uncontrolled Trials: No control group; all participants receive the intervention.

- Non-Randomized Controlled Trials: Comparison between intervention and control groups without
randomization.

- Randomized Controlled Trials (RCTs): Participants are randomly assigned to intervention or control
groups, minimizing bias.

- Cluster Randomized Controlled Trials: Clusters, not individuals, are randomized; used for group-
level interventions.

- Intervention Clusters and Comparison Clusters: Terms used in cluster trials to differentiate between
those receiving the intervention and those not.

- Community Intervention Trials: Entire communities are assigned to intervention or control groups
to study public health interventions.

These designs are chosen based on the research question, ethical considerations, and logistical
constraints. Each has its strengths and weaknesses, and the choice of design can significantly impact
the validity and generalizability of the study’s findings.
RANDOM ALLOCATION & RANDOM SELECTION

Random allocation and random selection are two important concepts in research, particularly in the
design of experiments and studies. They are distinct processes that serve different purposes:

1. Random Selection:

- Definition: Random selection refers to the process of selecting individuals from a larger
population to be included in a study. This process ensures that every individual in the population has
an equal chance of being selected for the study, which helps make the sample representative of the
population.

- Purpose: The main goal of random selection is to obtain a sample that accurately reflects the
broader population. This enhances the external validity or generalizability of the study’s findings,
meaning the results can be more confidently applied to the larger population.

- Example: Suppose a researcher wants to study the eating habits of high school students in a city.
If they randomly select students from all the high schools in the city, every student has an equal
chance of being chosen, making the sample likely to represent the overall population of high school
students.

- Key Point: Random selection is about how participants are chosen from the population to be
included in the study.

2. Random Allocation:

- Definition: Random allocation (also known as random assignment) refers to the process of
assigning the selected participants to different groups within the study, such as an intervention
group or a control group. This process is random, meaning each participant has an equal chance of
being assigned to any group.

- Purpose: The goal of random allocation is to ensure that the groups being compared in the study
are similar in all respects except for the intervention. This minimizes selection bias and confounding
variables, enhancing the internal validity of the study. It allows researchers to attribute differences in
outcomes between groups to the intervention itself rather than to other factors.

- Example: In a clinical trial testing a new medication, participants who have been selected to
participate are then randomly allocated to either the group receiving the new medication or the
group receiving a placebo. This random assignment ensures that any differences observed between
the two groups can be attributed to the medication rather than other factors.

- Key Point: Random allocation is about how participants are distributed into different groups
within the study after they have been selected.

Summary of Differences:

- Random Selection:

- Focus: Selection of participants from the broader population.

 - Purpose: To ensure the sample is representative of the population, improving external validity.

- Stage in Research: Occurs before participants are divided into groups.

- Random Allocation:

- Focus: Assignment of participants to groups within the study (e.g., intervention vs. control).

- Purpose: To ensure groups are comparable, improving internal validity by reducing bias.

- Stage in Research: Occurs after participants have been selected.

Both processes are critical for different aspects of study validity. Random selection helps ensure that
the study's findings can be generalized to a larger population, while random allocation ensures that
the study's results are due to the intervention rather than other variables.

Why random allocation?

• Random allocation is expected to balance both known and unknown confounders and effect
modifiers in both groups.

• Thus we achieve a situation close to the counterfactual ideal. Ideally, a placebo in the control
group achieves this. In many situations, a placebo is not possible due to the nature of the
intervention or ethical issues when conventional treatment exists for the condition under study.

• Blinding – for equipoise – single, double, triple, quadruple

In the context of clinical trials and other experimental studies, "intervention arms" refer to the
different groups or segments of participants that receive specific treatments, interventions, or
exposures being studied. Each arm of the trial represents a different intervention strategy or
condition.

Intervention Arms:
1. Purpose:

- The purpose of having multiple intervention arms is to compare the effects of different
treatments or interventions on the participants. By observing the outcomes in each arm, researchers
can determine which intervention is most effective or whether there are differences in outcomes
between the interventions.

2. Types of Arms:

- Treatment Arm: This group receives the experimental treatment or intervention that is being
tested. For example, in a drug trial, one arm might receive the new medication.

- Control Arm: This group does not receive the experimental treatment but instead receives a
placebo, standard treatment, or no treatment at all. The control arm serves as a benchmark to
compare the effects of the experimental treatment.

- Placebo Arm: A specific type of control arm where participants receive a placebo (a substance
with no therapeutic effect) instead of an active treatment. This helps to assess the true effect of the
treatment by eliminating the placebo effect.
 - Comparative Arm: In some trials, there might be more than one treatment arm, where
participants receive different interventions for comparison. For example, one arm might receive the
new drug, another might receive an existing drug, and a third might receive a placebo.

3. Randomization:

- In randomized controlled trials (RCTs), participants are typically randomly assigned to one of the
intervention arms. This randomization helps ensure that the groups are comparable and that any
differences in outcomes can be attributed to the intervention rather than to pre-existing
differences between the groups.

4. Outcome Measurement:

- The outcomes for each intervention arm are measured and compared at the end of the study.
This comparison helps researchers determine the effectiveness and safety of the intervention(s).

Example of Intervention Arms:

Imagine a clinical trial designed to test the effectiveness of a new diabetes medication:

- Arm 1 (Treatment Arm): Participants in this arm receive the new diabetes medication.

- Arm 2 (Control Arm): Participants in this arm receive the standard treatment for diabetes (e.g., an
existing medication).

- Arm 3 (Placebo Arm): Participants in this arm receive a placebo.

The researchers would compare outcomes such as blood sugar levels, side effects, and overall health
between these arms to determine if the new medication is more effective or safer than the standard
treatment or placebo.

Importance of Intervention Arms:

- Comparative Analysis: By having different arms, researchers can conduct a comparative analysis to
see how each intervention performs relative to others.

- Bias Reduction: Randomization into different arms helps reduce bias, making the findings more
reliable.

- Safety and Efficacy: Multiple arms allow researchers to assess not only the effectiveness of the
intervention but also any potential side effects or risks associated with it.

Intervention arms are a crucial component of experimental study designs, particularly in clinical
trials, where the goal is to determine the efficacy and safety of new treatments or interventions.

The intervention arms you've listed (PQR, PQ0, PR0, QR0, P00, Q00, R00, 000) represent different
combinations of treatments or interventions in a study. This type of design is often used in factorial
trials, where multiple interventions are tested in various combinations to assess their individual and
combined effects.
Breakdown of the Intervention Arms:

Let's assume that in this study, there are three different interventions being tested, labeled as P, Q,
and R. The "0" represents the absence of that intervention. The different arms could then be
interpreted as follows:

1. PQR:

- Participants in this arm receive all three interventions (P, Q, and R).

2. PQ0:

- Participants in this arm receive interventions P and Q, but not R.

3. PR0:

- Participants in this arm receive interventions P and R, but not Q.

4. QR0:

- Participants in this arm receive interventions Q and R, but not P.

5. P00:

- Participants in this arm receive only intervention P, with neither Q nor R.

6. Q00:

- Participants in this arm receive only intervention Q, with neither P nor R

7. R00:

- Participants in this arm receive only intervention R, with neither P nor Q.

8. 000:

- Participants in this arm do not receive any of the interventions (P, Q, or R). This arm serves as the
control group.

Purpose of Such a Design:

- Factorial Design: This type of setup is often used in a factorial trial design where researchers want
to investigate not only the effects of each intervention individually but also the potential interactions
between them.

- Efficiency: By testing multiple interventions and their combinations within the same study, this
design allows for an efficient use of resources and can provide insights into how the interventions
work together.

- Interaction Effects: The study can reveal whether the combined effects of P, Q, and R are greater
(synergistic) or less (antagonistic) than the sum of their individual effects.

Example Scenario:

Imagine a trial testing three lifestyle interventions to reduce blood pressure:

- P: Physical exercise.

- Q: Dietary changes.

- R: Stress management.

The intervention arms would be:

1. PQR: Participants do physical exercise, follow dietary changes, and engage in stress management.

2. PQ0: Participants do physical exercise and follow dietary changes, but do not engage in stress
management.

3. PR0: Participants do physical exercise and engage in stress management, but do not follow dietary
changes.

4. QR0: Participants follow dietary changes and engage in stress management, but do not do
physical exercise.

5. P00: Participants only do physical exercise.

6. Q00: Participants only follow dietary changes.

7. R00: Participants only engage in stress management.

8. 000: Participants do not receive any intervention.

Outcomes:

By comparing the outcomes across these arms, researchers can determine:

- The effect of each intervention individually (e.g., comparing P00 to 000).

- The combined effects of two interventions (e.g., comparing PQ0 to 000).

- The overall effect of all three interventions together (e.g., comparing PQR to 000).

- Potential interactions between interventions (e.g., if the combination of P and Q is more effective
than either one alone).

This approach provides a comprehensive understanding of how different interventions contribute to

the outcome, both individually and in combination.

STABLE POPULATION

“A population that has constant fertility and mortality rates, no migration, and consequently a
fixed age distribution and constant growth rate; a population with stable structure.”

Is the growth rate of stable population is nil?

The statement "Growth rate of a stable population is nil" is false.

Explanation:

A stable population is characterized by a situation where the age distribution remains constant over
time, meaning that the population structure does not change from one generation to the next. This
stability is achieved when the birth and death rates are balanced in such a way that the population
size remains constant.

Growth rate is constant but not necessarily nil – growth can be positive or
negative
However, the growth rate of a stable population can be zero, but not necessarily nil in all cases.
Specifically:

- Growth Rate = 0: A stable population can have a zero growth rate if the number of births equals the
number of deaths, resulting in no net increase in population size. In this case, the population size
remains constant, but the growth rate is zero, not nil.

- Population Dynamics: If a population is stable, it means the age-specific birth and death rates are
such that the age distribution does not change, but this does not imply that the growth rate is
necessarily nil. It is typically zero in a stable population if the fertility and mortality rates are such
that the population size remains constant over time.

Conclusion:

While a stable population implies a constant age distribution and a zero growth rate (in the sense
that the population size does not change), it does not mean that the growth rate is nil. Instead, it
means that the population growth rate is zero, reflecting balance in the rates of birth and death.

STATIONARY POPULATION:
“A stable population that has a zero growth rate with constant numbers of births and deaths each
year.

DYNAMIC POPULATION
• “A population that gains and loses members. All natural populations are dynamic—a fact
recognized by the term population dynamics, which is used by demographers to denote changing
composition.

OPEN AND CLOSED POPULATIONS

• Open population – one that can gain (or lose) members

• Closed population – one that cannot gain members – it keeps losing members due to death

GENERAL POPULATION
• “All members of a human population, defined essentially on the basis of geographical location, as
in a country, region, city, etc.

All inhabitants of some given area. Everyone in the POPULATION being studied, irrespective of race,
ethnicity, or professional status.”

In epidemiological parlance – “Individuals admitted to hospitals, other health care facilities, and
prisons are usually considered not to be part of the general population. The term is often used to
underline the different results that studies tend to obtain in the general population and in specific
populations, subgroups, or settings (e.g., in a working population, a hospitalized population).
• CLOSED COHORT-“A population in which membership begins at a defined time or with a
defined event and ends only through occurrence of the study outcome or the end of eligibility for
membership. An example is a population of women in labor being studied to determine the vital
status of their offspring (i.e., whether live or stillborn)”

• FIXED COHORT-“A cohort in which no additional membership is allowed—that is, it is fixed

by being present at some defining event (“zero time”); an example is the cohort comprising survivors
of the atomic bomb exploded at Hiroshima.”

Sometimes these terms are used synonymously –but there is a subtle difference –the latter is usually
used when the event is natural and not created for the purpose of the study. A fixed cohort is
unchangeable, while a closed cohort can continue to recruit those with the specific exposure as per
the selection definition.

• Open cohort –members can leave or new members can join

“Fixed” in populations and cohorts is theoretically impossible

• If nothing else, people will age • Thus, over time, fixed cohorts become increasingly
unrepresentative of the rest of the population

“But people themselves alter so much, that there is something new to be observed in them for
ever.”Jane Austen

 Source population/ Base population –“The group from which a study group is selected.”

• Reference population –“The standard against which a population that is being studied can be
compared.”

• Standard Population-“A population in which the age and sex composition is known precisely as a
result of a census or by an arbitrary means

(e.g., an imaginary population, the “standard million,” in which age/sex composition is arbitrary).

A standard population is used as a comparison group in the actuarial procedure of standardization of

mortality rates. / A population used as the reference in STANDARDIZATION”

An activity that shapes a particular epistemic entity, assumed to be “measurable”

Value – Certain body types are bad for heart health. If this issue is addressed early on: – Lot of heart
disease can be prevented. – Longer years of healthy life lived – Limited economic problems •
Concept – Overweight • Measurement – Body mass index, body fat, waist circumference, waist-hip
ratio • Element of interest – men and women in the age group 18 to 30 years

Epistemic Entity:

In this context, an epistemic entity refers to something related to knowledge or understanding—like

a concept, a measure, or a piece of information we are studying.

Activity:
The activity here is an action or process that influences or changes this entity.

Measurable:

Measurable means that we can quantify or assess it using some kind of measurement tool or
method.

Simplified Explanation:

Think of it this way: Suppose you are trying to understand how much people enjoy a new type of
exercise class.

- Epistemic Entity: This could be "enjoyment of the exercise class."

- Activity: You might conduct a survey where participants rate their enjoyment on a scale from 1 to
10.

- Measurable: The ratings you get are quantifiable, meaning you can measure and analyze them.

So, the activity of surveying people (or asking questions) shapes our understanding of "enjoyment of
the exercise class" and allows us to measure how much people enjoy it. This helps us gather
knowledge about the class's popularity and effectiveness.

In summary, it means that you’re using an action (like a survey) to understand and measure a
particular piece of knowledge or information.

TARGET POPULATION -“The collection of individuals, items, measurements, etc., about

which inferences are desired. The term is sometimes used to indicate the population or group from
which a sample or study population is drawn and sometimes to denote a reference population
about which inferences are desired.”

“ The group to which inference from the study is directed.”

• “ The group of persons for whom an intervention is planned.”

• Notions like representativeness and generalizability are based on this concept.

REPRESENTATIVENESS
• “The degree to which the characteristics of a study (notably, of study subjects and setting, but
sometimes also of exposures and outcomes) are similar to those of an external population that did
not participate in the study.

Representativeness is time-, place-, and context-specific. Although it is not necessary for studies with
etiologic aims, it has an important place in health surveys and descriptive studies. Evidence-based
health policies require population-representative information. It should neither be avoided nor
uncritically sought.”

REPRESENTATIVE SAMPLE
• “A sample that to a large extent resembles a population of interest.

The term representative as it is commonly used is largely undefined in the statistical or mathematical
sense. The use of probability sampling will not ensure that a sample will be representative of the
population in all relevant aspects. It is unwarranted to assume that if the sample resembles the
reference population on factors that have been checked, no differences exist in other relevant
factors”

GENERALIZABILITY
• “The degree to which results of a study may apply, be relevant, or be generalized to populations
or groups that did not participate in the study. In etiological research, such inferences are not
merely statistical in nature but must be based on theory, judgment, and evidence external to the
study (e.g., on available knowledge on biological, clinical, or social mechanisms linking an exposure
to an outcome); inferences may partly be subjective but not arbitrary.

• REPRESENTATIVENESS of the study sample may enhance generalizability in studies with strong
descriptive components; yet, knowledge of subject matter, context, and study conditions enables
generalization in studies of etiologic nature that are not based on a strictly representative sample.
Scientific and statistical inferences build on different types of logic.”

GROUP COMPARISONS
• “Comparisons that consist in contrasting what is observed in a group of people in the presence of
exposure to what would have occurred had the group of interest not been exposed to the
postulated cause. Differences in frequency of disease occurrence between groups can logically be
interpreted as being caused by the exposure. Integratedly (sic) with POPULATION THINKING, this is
the main mode of knowledge acquisition in epidemiology.”

Here's an updated table that includes survival analysis measures alongside the other key differences
between case-control, cohort, and cross-sectional studies:

| Aspect | Case-Control Study | Cohort Study | Cross-Sectional Study |

|---------------------------|--------------------------------------------------------|----------------------------------------------
----------|---------------------------------------------------------|

| | Retrospective; starts with outcome (cases vs. controls) | Prospective or retrospective;

starts with exposure | Observational; examines exposure and outcome at a single point in time |

| Main Purpose | Identify associations between exposures and outcomes | Assess incidence of
outcomes based on exposure | Determine the prevalence of outcomes and exposures |

| Timing | Outcome has already occurred; looks back in time | Follows participants over
time to observe outcomes | Snapshot at a single point in time |

| Data Collection | Collect data on past exposures from cases and controls | Collect data on
exposure status and follow up for outcomes | Collect data on exposure and outcome simultaneously
|

| Sample Size | Typically smaller; cases and controls are matched | Typically larger;
especially if exposure or outcome is rare | Can vary; usually smaller than cohort studies |

| Biases | Recall Bias: Participants may not accurately remember past exposures. <br>
Selection Bias: Control selection may not represent the general population. | Selection Bias: Loss to
follow-up can affect study validity. <br> Information Bias: Misclassification of exposure or outcome.
| Sampling Bias: The sample may not represent the population. <br> Measurement Bias: Errors in
data collection. |

| Epidemiological Measures | Odds Ratio (OR): Measures the odds of exposure among cases vs.
controls. | Relative Risk (RR): Measures the risk of developing the outcome in the exposed group
compared to the unexposed group. <br> Incidence Rate: Measures the rate at which new cases
occur in a population. | Prevalence Ratio (PR): Compares the prevalence of the outcome between
different groups. <br> Prevalence: The proportion of individuals with a particular outcome or
exposure at a specific time. |

| Survival Analysis Measures | Not typically used for survival analysis. | Hazard Ratio (HR): Measures
the effect of an exposure on the risk of an event occurring over time. <br> Kaplan-Meier Curve:
Estimates survival probabilities over time. <br> Cox Proportional Hazards Model: Analyzes the
relationship between survival time and one or more predictors. | Not applicable, as survival analysis
requires time-to-event data. |

| Strengths | Useful for studying rare outcomes; relatively quick and inexpensive | Can
establish temporal relationships; useful for studying rare exposures; can measure incidence | Simple
and quick; useful for assessing the prevalence of outcomes and exposures |

| Limitations | Cannot establish causality; prone to recall and selection bias | Expensive and
time-consuming; potential for loss to follow-up; not ideal for rare outcomes | Cannot establish
causality; limited to associations at one point in time |

| Example | Doll and Hill’s study on smoking and lung cancer | Framingham Heart Study
on cardiovascular disease | National Health and Nutrition Examination Survey (NHANES) |

Aspect
Study Design

This table now includes survival analysis measures, which are primarily relevant to cohort studies
due to their longitudinal nature and the ability to analyze time-to-event data.

For outbreak investigations, the case-control study design is typically the most suitable . Here’s
why:

Case-Control Study for Outbreak Investigations:

- Efficiency: Case-control studies are particularly efficient for outbreak investigations because they
allow researchers to quickly compare individuals who have the disease (cases) with those who do
not (controls).

- Rapid Results: In an outbreak, time is critical. Case-control studies can be conducted relatively
quickly, enabling investigators to identify the likely source or cause of the outbreak.
- Small Sample Size: These studies often require a smaller sample size than cohort studies, which is
beneficial in situations where the number of cases is limited, such as in a localized outbreak.

- Focus on Rare Outcomes: Outbreaks typically involve a sudden increase in the incidence of a rare
event (e.g., a specific infectious disease), making the case-control design particularly appropriate.

Why Not Other Designs?

- Cohort Study:

- Less Efficient: While cohort studies can be useful in some outbreak scenarios, they are generally
less efficient because they require following a group over time to see who develops the disease,
which may not be feasible during an acute outbreak.

- Time-Consuming: Cohort studies take longer to yield results, which can delay the identification of
the source and control measures during an outbreak.

- Cross-Sectional Study:

- Prevalence, Not Causality: Cross-sectional studies are good for measuring the prevalence of an
illness at a single point in time but are not ideal for identifying the cause of an outbreak, as they do
not establish temporal relationships between exposure and disease.

Example:

In a foodborne illness outbreak, a case-control study might compare the food consumption histories
of those who became ill (cases) with those who did not (controls) to identify the specific food item
responsible for the outbreak.

Summary:

Case-control studies are the most appropriate and efficient design for investigating outbreaks, as
they enable rapid identification of potential causes or sources of the outbreak, which is crucial for
implementing control measures.

The STROBE (Strengthening the Reporting of Observational Studies in

Epidemiology) guidelines provide a checklist of essential items to include when
reporting observational studies. Here are the key points:

I. Title and Abstract

- Clearly state the study's purpose, design, and main findings

II. Introduction

- Provide context, research question, and objectives

- Explain the study's significance and relevance

III. Methods

- Study design (e.g., cohort, case-control, cross-sectional)

- Setting and participants (e.g., population, sampling)
- Variables and data sources
- Bias and confounding control
- Statistical analysis

IV. Results

- Participant flow and losses

- Descriptive statistics and outcome measures
- Main results, including estimates and precision

V. Discussion

- Interpretation of results
- Limitations and potential biases
- Comparison with existing literature
- Implications and future research directions

VI. Other

- Funding and conflicts of interest

- Ethical approval and informed consent

The STROBE guidelines aim to improve the transparency and quality of reporting in
observational studies, facilitating critical evaluation and replication.

Note: STROBE has extensions for specific study designs, such as STROBE-nut for
nutritional epidemiology and STROBE-AMS for animal studies.

Rothman's causal pies, also known as "causal diagrams" or "directed acyclic

graphs" (DAGs), are a visual representation of the relationships between
variables in a causal system. They were developed by epidemiologist Kenneth
Rothman to help identify and understand causal relationships.

A causal pie consists of:

1. _Nodes_: Representing variables or factors.

2. _Arrows_: Indicating direct causal relationships between nodes.
3. _Paths_: Showing the flow of causality between nodes.

Rothman's causal pies help to:

1. _Identify confounding variables_

2. _Clarify causal relationships_
3. _Detect bias_
4. _Guide data analysis and interpretation_

By using causal pies, researchers can:

1. _Visualize complex relationships_

2. _Communicate findings effectively_
3. _Make informed decisions about data analysis and study design_

Rothman's causal pies are a powerful tool for understanding causality and have
been widely adopted in epidemiology, statistics, and other fields.

BIASES
BERKSONIAN BIAS / HOSPITAL BIAS OR ADMISSION RATE BIAS

Occurs when the sampling frame is restricted to hospitalized individuals, leading to a

biased sample.

This bias arises because:

1. _Selection for hospitalization_: Patients admitted to hospitals may not be

representative of all cases in the population.
2. _Disease severity_: Hospitalized cases may be more severe or have different
characteristics than non-hospitalized cases.
3. _Referral patterns_: Referral patterns to hospitals can vary, leading to an
unrepresentative sample.

Berksonian bias can affect:

1. _Incidence rates_: Estimates of disease incidence may be inflated or deflated.

2. _Risk factor associations_: Associations between risk factors and outcomes may
be distorted.
3. _Disease severity_: Estimates of disease severity may be biased towards more
severe cases.

To minimize Berksonian bias:

1. _Use population-based sampling frames_: Include non-hospitalized cases in the

sample.
2. _Adjust for referral patterns_: Account for differences in referral patterns to
hospitals.
3. _Validate findings_: Verify results using alternative data sources or study designs.

Remember, Berksonian bias can lead to incorrect conclusions, so it's essential to

consider this bias when designing and interpreting studies!

RECALL BIAS

- Recall bias occurs when participants in a study inaccurately recall or report past
events, exposures, or experiences.

*Types of recall bias:*

- _Reconstruction bias_: Participants may reconstruct past events based on
current knowledge or beliefs.
- _Retrospective bias_: Participants may recall events differently due to changes in
perspective or experiences over time.

*Factors that contribute to recall bias:*

- _Time elapsed_: Longer periods between the event and recall increase the
likelihood of bias.
- _Age and cognitive ability_: Older adults or those with cognitive impairments may
experience more recall bias.
- _Emotional state_: Strong emotions can influence recall accuracy.
- _Social and cultural factors_: Cultural norms, social pressures, or expectations can
shape recall.

*Consequences of recall bias:*

- _Biased estimates_: Recall bias can lead to inaccurate or incomplete data,

affecting study conclusions.
- _Loss of validity_: Recall bias can compromise the validity of findings, particularly
in studies relying on self-reported data.

*Mitigating recall bias:*

- _Use multiple sources_: Combine self-reported data with objective measures or

records.
- _Use memory aids_: Employ techniques like calendars or diaries to assist recall.
- _Minimize time elapsed_: Collect data soon after the event or exposure.
- _Train interviewers_: Ensure interviewers are aware of recall bias and use
techniques to minimize it.

By understanding and addressing recall bias, researchers can strengthen the

accuracy and validity of their findings.

INFORMATION/ MISCLASSIFICATION BIAS

Occurs when there are errors or inaccuracies in the collection, recording, or

classification of data. This can lead to incorrect conclusions or estimates, and can
affect the validity and reliability of study findings.

Types of information bias:

1. *Measurement bias*: Errors in measuring or collecting data, such as faulty

equipment or incorrect survey questions.
2. *Recall bias*: Participants' memories of past events or exposures are inaccurate
or incomplete.
3. *Selection bias*: Errors in selecting participants or samples, leading to an
unrepresentative group.
4. *Observer bias*: Researchers' or observers' expectations or biases influence data
collection or interpretation.
5. *Confirmation bias*: Researchers selectively collect or interpret data to support
preconceived notions.
6. *Social desirability bias*: Participants provide answers that are socially
acceptable, rather than truthful.
7. *Non-response bias*: Missing data or non-responses lead to an incomplete or
inaccurate picture.

To minimize information bias:

1. Use validated data collection tools and methods.

2. Train researchers and observers to collect data accurately and consistently.
3. Use multiple sources to verify data.
4. Implement quality control measures.
5. Consider potential biases during study design and analysis.

Remember, information bias can have significant impacts on research findings, so

it's crucial to be aware of and address potential biases!

PECOT is a mnemonic device used in epidemiology to help researchers design and

evaluate studies.

PECOT stands for:

P - *Population*: Define the study population, including characteristics such as age,

sex, and disease status.

E - *Exposure*: Identify the exposure or risk factor being studied, including its
measurement and categorization.

C - *Comparison*: Determine the comparison group, which may be a group without

the exposure or a group with a different level of exposure.

O - *Outcome*: Specify the outcome or disease being studied, including its

measurement and diagnosis.

T - *Time*: Consider the time frame of the study, including the duration of exposure
and follow-up.

Eg:
P - *Population*: Adults aged 18-65 with a history of smoking

E - *Exposure*: Current smoking status (yes/no)

C - *Comparison*: Non-smokers

O - *Outcome*: Incidence of lung cancer

T - *Time*: 10-year follow-up period

By using PECOT, researchers can systematically evaluate and design studies,
ensuring that they address critical components and produce high-quality evidence.

PICOT

P - _Population_: Identify the specific population or group being studied.

I - _Intervention_ (or Exposure): Describe the intervention, treatment, or exposure

being investigated.

C - _Comparison_: Define the comparison group or control group.

O - _Outcome_: Specify the outcome or result being measured.

T - _Time_: Consider the time frame or duration of the study.

PICOT helps researchers create a *clear and concise research question, ensuring
that all essential elements* are considered. This framework is particularly useful for:

1. Formulating research questions

2. Developing study protocols
3. Evaluating study validity
4. Conducting systematic reviews

Here's an example of PICOT in action:

P - _Population_: Adults with type 2 diabetes

I - _Intervention_: Lifestyle modification program (diet and exercise)

C - _Comparison_: Standard care (no lifestyle modification)

O - _Outcome_: Change in HbA1c levels (blood sugar control)

T - _Time_: 6-month follow-up period

By using PICOT, researchers can create a well-defined research question, ensuring

that their study is focused, relevant, and impactful.

Hill's Criteria for Causality

Sir Austin Bradford Hill's criteria are a set of guidelines for evaluating the causal
relationship between a risk factor and a health outcome. Here are the 9 points:

1. *Strength of Association*: The strength of the association between the risk factor
and outcome should be significant.

2. *Consistency*: The association should be consistent across different studies,

populations, and settings.
3. *Specificity*: The risk factor should be associated with a specific outcome, rather
than multiple outcomes.

4. *Temporality*: The risk factor should precede the outcome in time.

5. *Biological Gradient* (Dose-Response): Increasing exposure to the risk factor

should lead to an increased risk of the outcome.

6. *Plausibility*: The association should be biologically plausible and supported by

existing knowledge.

7. *Coherence*: The association should align with existing knowledge and evidence.

8. *Experiment*: Experimental evidence, such as randomized controlled trials, can

provide strong evidence for causality.

9. *Analogy*: Similarity with other established causal relationships can support

causality.

These criteria are not a checklist, but rather a framework to evaluate the evidence for
causality. The more criteria met, the stronger the evidence for causality.

Note: Hill's criteria are not applicable to all types of research questions, but are
particularly useful for evaluating causal relationships in observational studies.