Epidemiology Methods I exam #1 Study guide

Descriptive Epidemiology

Measures how disease frequency and other population health indicators vary with age, geographic location, race/ethnicity, and other characteristics of person, place, and time.

Example: Number of adult women in the United States of America who received a diagnosis of type 2 diabetes in 2022.

Analytic Epidemiology

Assesses the effect of exposures, which include possible causes, and the occurence of disease.

Exposures

-A personal characteristic (assigned sex)

-environmental characteristic (air pollution)

-A behavioral characteristic (smoking)

-A socio-economic characteristic (family income)

-An aggregate level characteristic (urban setting)

-Modifiable (Stress)

-Non-Modifiable (genetics)

Prevention outcome of interest

absence of disease

Prevention research

We intervene on the causes to prevent disease.

History of epidemiology

Epidemiology is rooted in early attempts to undersrtand and improve public health.

As people started living more densly, people were more susceptible to epidemic diseases.

To prevent epidemic diseases, people used observations to generate hypotheses on causes of disease.

First smallpox vaccine (1796)

Characterized as skin leisons affected societies around the world for centuries.

It was widely known that people who had smallpox became immune.

Initially inoculation was performed using matter from smallpox leisons. Based on local beliefs that cowpox protected against smallpox Edward Jenner inoculated people with cowpox matter. ‘

In 1796, Jenner demonstrated that inoculation with cowpox protected against smallpox.

He called the procedure vaccination coming from the latin word for cowpox vaccinia.

Hand hygiene and childbed fever

Ignaz Philip Semmelweis studied why women in a clinic staffed by physicians were twice as likely to die from childbed fever compared to women in a clinic staffed by midwives.

The women performed autopsies prior to providing clinical care while midwives did not.

One of the physicians died after performing an autopsy and Semmelweis noticed that this physician had childbed fever symptoms as well.

Therefore he hypothesized that cadaverous particles caused childbed fever.

The handwashing intervention resulted in a dramatic decrease of childbed fever incidence.

Cholera outbreak in London

From August-Semtember 1854 a cholera epidemic struck london. John snow observed that cholera cases were clustered around the broad street pump.

He also observed that individuals who got their water from other sources were less likely to have cholera.

Snow hypothesized that the water from the broad street pump was contaminated and was a source of cholera.

After removing the handle of the broad street pump the cholera epidemic disappeared within a few days.

Epidemiologic research questions

Well defined research questions are important for two reasons

1. To ensure researchers understand the scope of the study.

2. To guide the study design and statistical analyses.

they include PICOT elements

PICO(T) Format

Population: the population in which the study is conducted

Intervention: intervention (or exposure of interest)

Comparison: comparison group (or control group)

Outcome: outcome of interest (disease)\

Time: time period during which the outcome is measured.

Causal effect

comparison of two counterfactuals or potential outcomes.

For example includes outcome if a person had chosen A at the moment T. And outcome if a person had chosen B at moment T.

Difference between the two potential outcomes

No effect: when disease status is the same across the two potential outcomes.

Effect: when disease status differs between the two potential outcomes.

Casual effect example

Riley smoked tobacco between the ages of 20 and 60 and was diagnosed with thrombosis at age 60.

Riley’s thrombosis status at age 60 if Riley did not smoke tobacco between the ages of 20 and 60.

Potential outcomes notation

Y represents the outcome

X represents the exposure, with x and x* representing the compared exposure values.

Values x and x* are commonly referred to as the casual contrast.

y(x): Outcome when individual A is exposed during time period T.

y (x*): Outcome when individual A is not exposed during time period T.

Fundamental problem of causal interference

It is impossible to observe both potential outcomes for the same individual over the same time period.

Scale of outcomes Causal effect

Binary Scale: outcomes is either present or absent (disease)

Continuous scale: outcome can take on many values (blood pressure)

Average potential outcomes binary outcome

Percentage of individuals who develop the outcome during time period T among exposed individuals

Percentage of individuals who develop the outcome during time period T among unexposed individuals.

Average potential outcomes continuous outcome

Mean outcome value after time period T among exposed individuals

Mean outcome value after time period T among unexposed individuals.

General steps of epidemiological studies

Selection of a group of individuals who are a representative of the population of interest.

Ascertainment of exposure and outcome

Comparison of outcome across exposure groups.

Two types of study designs

Experimental studies: epidemiologist controls exposure status of the participants by randomization to an intervention.

Nonexperimental studies:epidemiologist observes exposure status that participants “self select” (no intervention)

Randomized control trials

Experimental design: epidemiologist controls exposure status.

individuals are randomized to the intervention condition or the control condition at study entry.

Follow-up over time to determine outcome status.

Steps in a randomized control trials

1. Selection of individuals from the population of interest

2. Randomize individuals to the intervention group or the control group.

3. Measure each individual’s outcome status at the end of the follow-up.

4. Compare outcome between intervention and control group.

Randomization (RCT)

intervention is determined with a randomization scheme. Randomization scheme can be developed using psuedo random number generator.

Goal of randomization

even distribution of characteristics (cofounders) across the intervention group and control group.

The success is differences in the outcome between groups is due to the intervention and not due to the other confounding factors.

Participants characteristics are evenly distributed across the two groups in expectation so not necessarily in a single study.

Blinding in randomization

ideally for anyone who is involved in the study so they are unaware of intervention assignment.

goal is to limit biases that could affect intervention delivery, compliance, measurement of variables, analyses, and study conclusions.

Randomized controlled trials limitations

It is not ethical to randomize individuals to harmful exposures (eg smoking)

It is also unethical to include vulnerable populations in RCTS (eg children and pregnant women) due to potential unknown adverse side effects of the intervention.

Loss to follow-up can affect the validity of the results.

Perspective cohort studies

observational design: epidemiologist observes self-selected exposure status. Exposure status is measured at study entry.

Follow-up over time to determine outcome status.

Typically individuals with the outcome at study initiation are excluded.

Perspective follow-up in real time.

Perspective cohort study

1. Select individuals from the population of interest using current records.

2. Determine each individuals exposure at the time of selection.

3. Determine each individual’s outcome at the end of follow-up.

4. compare outcome between exposure groups.

Perspective studies limitations

Observed differences between groups may be due to confounding factors other than the exposure because of self selection of exposure status.

Not efficient for the study of rare outcomes as you need many participants to observe enough participants with the outcome.

Can be very costly and time consuming

Loss to follow-up.

Retrospective cohort studies

Observational design: epidemiolgist observes self selected exposure status.

exposure status is measured at the study entry (which is in the past)

Follow-up over time to determine outcome status

Typically individuals with the outcome at the time of selection are excluded.

The follow-up period is in the past.

Steps in a retrospective cohort study

1. Randomly select individuals from the population of interest using records from the past.

2. Determine each individuals exposure at the time of selection (ie the past)

3. Determine each individuals outcome at the end of the follow-up.

4. Compare outcome between exposure groups.

Retrospective cohort studies selection

Two general strategies for selection of study participants.

Random selection of individuals from the target population

Selection based on exposure status when the exposure is rare.

limitations of retrospective cohort studies

observed differences between groups may be due to confounding factors other than the exposure, because of self selection of exposure status.

Not efficient enough for the study of rare outcomes.

Loss to follow-up

Cannot be used for new exposures.

Historical records may not be completely accurate and may not contain information on all relevant variables.

Cross sectional studies

Observational design: epidemiologist observes self-selected exposure status.

Exposure and outcome are measured at the same time (no follow-up)

Also called a prevalence study as it is used to determine disease prevalence.

Steps in a cross sectional study

1. Randomly select individuals from the population of interest.

2. Determine each individual’s exposure and outcome of the time of selection.

3. Compare outcome between exposure groups.

cross sectional studies limitations

High risk of causality timing of exposure and outcome often remains unclear.

Not possible to study incidence new number of cases due to absence to follow-up.

Observed differences between groups due to confounding factors other than the exposure, because of self selection of exposure status.

Case control studies

observational design: epidemiologist observes self selected exposure status.

Selected based on the outcome (rather than the exposure)

Exposure is retrospectively assessed.

Sometimes referred to as a trohoc study.

Case control studies steps

1. Select individuals with the outcome (cases) and without the outcome (controls) from the target population.

2. Determine each individual’s past exposure status.

3. Compare exposure status between cases and controls.

Strengths of Case control studies

Well suited for studying rare outcomes because of selection on the outcome.

Ecological studies

Based on aggregate data (i.e populations) rather than individual data.

Comparison of the average exposure values and average outcome values across populations.

Steps in an ecological study

1. Select populations of individuals

2. Determine the average exposure status and average outcome status at a single point in time.

3. Plot average exposure status against average outcome status for all selected populations.

Ecological fallacy

Associations at the aggregate level differ from associations at the individual level.

Limitations of ecological studies

Limited evidence of causality because of ecological fallacy.

Frequencies strengths and limitations

Frequencies are useful for distribution of health resources.

However frequencies can be misleading when comparing outcomes across populations.

To compare groups of individuals we use relative measures that take into account population size.

Relative measures of disease ocurence

Are used to determine what proportion of individuals in a specific population:

Have the disease at the specific point in time.

Develop the disease over a specific period in time.

Point prevalence

The proportion of people in the population who have the outcome at a specific point in time.

is based on existing cases

applicable study design: cross-sectional studies.

Estimation of the point prevalence

Total population: the proportion of people in the total population who have the outcome at a specific point in time.

Exposed subpopulation: the proportion of exposed people who have the outcome at the specific point in time.

Unexposed subpopulation: The proportion of unexposed people who have the outcome at a specific point in time.

Incidence proportion (risk)

The proportion of people in a subpopulation who develop the outcome over a specific time period.

Applicable study designs: randomized controlled trials, prospective cohort studies, and retrospective cohort studies.

Period prevalence

The proportion of the subpopulation who either have the outcome at the study entry or develop the outcome over a specific time period.

Applicable study designs: prospective cohort studies, and retrospective cohort studies.

Risk difference (RD)

incidence in exposed-incidence in exposed group.

incidence in unexposed group reflects the background risk.

Risk ratio

incidence in the unexposed group reflects the background risk.

Risk difference versus risk ratio

Risk difference: is a measure of association on the additive scale. Represents the absolute difference in risks between two groups. Indicates how often a disease occurs.

E[Y(x)]-E[Y(x*)]

Risk ratio: is a measure of association on a multiplicative scale. Represents the risk in one group relative to the risk group in another group. Does not indicate how disease occurs.

The risk ratio is an estimate of

E[Y(x)]/E[Y(x*)]

Prevalence difference

=prevalence in the exposed group-prevalence in the unexposed group.

The prevalence based equivalent of the risk difference.

Prevalence ratio

prevalence of the exposed group/prevalence of the unexposed group

Incidence/prevalence versus outcome odds

Incidence/Prevalence: percentage of people with the outcome in the subpopulation.

Outcome odds: the odds of having the outcome versus not having the outcome in a subpopulation.

Outcome odds ratio

outcome odds in the exposed group/outcome odds in the unexposed group

Direction of odds ratio

Odds ratio=1 no difference in the odds of the outcome between the exposed and unexposed groups.

Odds ratio>1 odds of the outcome is higher in the exposed group compared to the unexposed group.

Odds ratio<1: Odds of the outcome is lower in the exposed group compared to the unexposed group.

Reference group

is the comparison group, as defined by C in the PICOT.

The order matters in the 2×2 and should be the 2nd row in the table.