Epi Class 4 Flashcards Preview

Epi > Epi Class 4 > Flashcards

Flashcards in Epi Class 4 Deck (46):


Most research studies in public health can be expressed in terms of: [exposure] and [disease/outcome] in [population]

Exposures include: SES, health-related behaviors, health status, environmental exposures, etc.

Diseases/Outcomes include: injuries, communicable/infectious diseases, noncommunicable/chronic diseases, neuropsychiatric disorders

Population- try to be specific: ex. Australian children under 5, teachers with at least 10 years of classroom experience


Reviews / Meta-Analyses

A review article or meta-analysis carefully gathers all prior publications on a specific topic and summarizes them to provide a big-picture analysis.

1. An extensive search of the literature
2. Extraction of key information from relevant articles
3. Clear and concise presentation of this information
• The goal of a meta-analysis is to combine the
results of several high-quality articles that used
similar methods to collect and analyze data
into one summary statistic.

• Meta-analysis usually begins with a
comprehensive systematic review of the
literature to identify every single possibly
relevant article.


Ecological Studies (Correlational Studies, Aggregate Studies)

Uses population-level data to examine the relationship between exposure rates (E) and disease rates (D) in different populations (P).

No individuals – just average rates for a population.
• Use country-level data from 191 nations to explore
whether there is an association between fat intake
(mean dietary fat consumption) and breast cancer
(population incidence rate).

• Use city-level data from 40 U.S. cities to explore
whether there is an association between air pollution
(mean parts per million) and lung disease (rate of
hospitalization for chronic lung disease).

The exposure is often (but not always) “environmental.”


Strengths of ecological studies

• Low cost and convenient

• Measures environmental exposures (like air pollution) that are experienced by large groups of people

• Simplicity


Ecological Fallacy

The incorrect attribution of population-level associations to individuals.

Correlational studies compare groups rather than individuals.

No individual-level data are included in the analysis, only population-level data.
• Studies of group (not individual) characteristics
do not tell us whether these characteristics are
linked at the individual level.


Cross-Sectional Surveys (Prevalence Studies)

“Snapshot” in time – quick population survey

The goal is to measure the proportion of a population with a particular exposure or disease (or both) at one point in time based on a representative sample of a population.

Cross-sectional surveys are among the most popular study approaches in the health sciences because they allow for the relatively rapid collection of new data.
• Uses: describe communities, assess
population needs, evaluate programs,
establish baseline data prior to the initiation of
longitudinal studies

Simple study design: ask a few hundred people to complete a short questionnaire and then analyze the data.

The participants must be reasonably representative of some larger population.

It is usually NOT acceptable to use a convenience population

The sampled population must be as diverse as the source/target populations.


Cross-Sectional Surveys (Prevalence Studies): Analysis

Prevalence: the proportion of the population with a given trait at the time of the survey.

Comparative statistics: prevalence rate ratios (PRRs) can be used to compare the prevalence of a characteristic in two population subgroups by taking a ratio of their prevalence rates; odds ratios (ORs) can also be used.

An exposure can be said to be "associated" or "related" to a disease, but a cross-sectional survey cannot show that an exposure caused a disease



• A study population should be representative of a well-defined source population.

• Study populations should not be too big (wastes resources) or too small (wastes resources)… but bigger is usually better! (Cross-sectional studies usually need to have >300 people.)

• If only some members of the source population will be sampled, an appropriate sampling technique must be used. Each unique sample drawn from a larger population will yield a slightly different study population.

• Bigger sample sizes = narrower confidence intervals = more likely to have a “statistically significant” result

• The goal is to have a sample size that is large enough to capture TRUE differences in the population



The ability of a statistical test to detect significance in a population when differences really do exist (power = 1-β).

Sample sizes must be especially large to have the power to detect differences between means that are close to one another or significance for relative risks and odds ratios that have a point estimate close to 1.
• Use a computer program to estimate sample size and power, like OpenEpi.com or the StatCalc function in Epi Info

• Estimates of sample size and power are based on estimates of the prevalence / incidence of disease / exposure in specific population groups and the estimated RR or OR

• This is an estimate not a calculation


Target Population

the general population that the study seeks to understand


Source Population

the specific individuals from which a representative sample will be drawn


Sample Population

individuals asked to participate


Study Population

eligible participants


Simple Random Sampling

each person has an equal chance of being selected


Systematic Sampling

after a random start point, every nth person is selected


Stratified Sampling

simple random samples selected from each of several strata


Cluster Sampling

an area is divided into geographic clusters and some clusters are selected for inclusion


Case Report

describes one patient

Case Report: (n = 1)

All participants have the disease (no controls).

No comparison group.


Case Series

describes two or more patients who have the same disease condition or who have undergone the same procedure.

Case Series: (n ≥ 2)

All participants have the disease (no controls).

No comparison group.


Uses of Case Report/Series

• Describing the characteristics of and similarities among a group of individuals with the same signs and/or symptoms of disease

• Identifying new syndromes and refining case definitions

• Clarifying typical disease progression (natural history of disease)

• Developing hypotheses for future research


Case-Control Study

• Individual participants in a case-control study are selected for inclusion in the study based on
their disease status
• Cases = participants with the disease of
• Controls = participants without the disease
• Both cases and controls are asked the same
set of questions about past exposures

• Use: A case-control study is often the best study approach when the disease is relatively uncommon and a study of the general population is unlikely to yield more than a few cases.

• Case-control studies are the best (observational, analytic) study design for RARE DISEASES.


Case-Control Study: Case Definition

All cases must have the same disease, disability, or other health related condition. A case definition should specify exactly what characteristics must be present or absent for a person to be deemed a case


Finding Controls for Case-Control Study

Controls must be reasonably similar to cases except for their disease status.
• The inclusion and exclusion criteria for cases that do not specifically relate to the disease should also apply to controls.
 For example, if cases must be males
between 25 and 39 years of age, controls
must also be men in this age group.

• Controls may be recruited from, among other sources: friends and relatives of cases, hospital or clinic patients without the disease of interest, the general population.



any systematic error that results in a mistaken estimate of an exposure’s effect on the risk of disease.


Misclassification Bias

errors are made in classifying either disease or exposure status

All participants must be asked questions that confirm whether each is a case, a control, or neither. Adhering to strict definitions for what constitutes a case and what constitutes a control minimizes the risk of misclassification bias.


Recall Bias

occurs when cases and controls systematically have different memories of the past. Cases may have more vivid memories than controls of participation or lack of participation in activities perceived to be risky or beneficial because they are searching for explanations for their illnesses


Nested case-control study

Select cases and controls from among the participants in a cohort study, then use records from the cohort study to assess the exposure histories of people classified as “cases” and “controls.” This is economical and minimizes recall bias.



There are three basic options for matching cases and controls:
1. no matching
2. frequency (group) matching
3. matched-pairs (individual) matching.

• Many case-control studies use no matching. They assume that similar inclusion and exclusion criteria for cases and controls will result in case and control populations that have similar distributions according to sex, age group, socioeconomic status, and other characteristics.


Frequency (Group) Matching

The goal of frequency (group) matching is to recruit a control population that is similar to the case population.
• Individual cases are not tied to individual
controls during analysis.
• Example: For each hospital case, a
researcher identifies select one control from
the hospital registration files who was
admitted the same week as the case, who is
the same sex as the case, and who is within
± 3 years of the age of the case.


Matched-Pairs (Individual) Matching

each case is personally linked to a particular individual control.
• This approach is fairly common in genetic studies, in which a case is linked to a genetic sibling or other close genetic relative for analysis.

• This kind of matched-pairs approach requires a special type of analysis


Limitations of Matching

The variables used as matching criteria cannot be considered as exposures during analysis.
• Example: If cases and controls are matched based age, then the cases and controls in the study will have the same mean age, even if in the general population cases tend to be older than non-cases.

It can be difficult to find controls who meet all of the matching criteria when there are many matching characteristics.


Cohort Study

A cohort study follows participants through time to calculate the rate at which new (incident) disease occurs and to identify risk factors for the disease.

All cohort studies have at least two measurement times:
• An initial survey that determines the baseline exposure and disease status of all participants

• One or more follow-up assessments that determine how many participants have developed a new (incident) disease since the initial examination

• All participants must complete the same assessments of exposure and disease at baseline and follow-up to prevent the information bias that might result when exposed participants are more thoroughly examined for disease than unexposed participants.


Three main types of Cohort Studies

1. Retrospective
2. Prospective
3. Longitudinal


Retrospective Cohort Study

Recruit participants based on their exposure status:
• one group of participants is recruited because they are known to have had a particular exposure
• a second group is recruited because they are known not to have been exposed.

Use baseline information collected at some point in the past and follow the cohort to another point IN THE PAST or to the present (or future).

• A good study design for RARE EXPOSURES.

• The members of the two (or more) comparison groups for both prospective and retrospective of studies should be similar except for their exposure status.

• Because the goal of cohort studies is to examine incident disease, retrospective cohort studies must be able to demonstrate that the outcome of interest was not present in any members of the cohort at baseline.


Prospective Cohort Study

Recruit participants based on their exposure status:
• one group of participants is recruited because they are known to have had a particular exposure
• second group is recruited because they are known not to have been exposed.

Collect baseline data about exposures and outcomes IN THE PRESENT and follow the cohort to some point in the future.

• A good study design for RARE EXPOSURES.

• The members of the two (or more) comparison groups for both prospective and retrospective of studies should be similar except for their exposure status.

Because the goal of cohort studies is to examine incident disease, prospective cohort studies must be able to demonstrate that the outcome of interest was not present in any members of the cohort at baseline.
• Examples:
 Recruit industrial workers exposed to a certain chemical and similar workers in a plant that does not use the chemical

 Recruit children with high blood lead levels (indicating environmental exposure to lead) and low blood lead levels who attend the same elementary school


Longitudinal Cohort Study

Longitudinal cohort studies usually last for years or even decades.

Individual participants are assessed at baseline for several exposures and diseases. Then they are followed forward in time to determine the incidence rate for one or more outcomes of interest.

Examples of populations for a longitudinal study:
• All the residents of one town
• A representative sample of members of one professional organization
• A representative group of students recruited from the same university


Longitudinal Cohort Study: Population Dynamics

• In a longitudinal study with a fixed population, all participants start the study at the same time and no one is allowed to join later.

• In a study with a dynamic population, participants are recruited using rolling admission and replacement of dropouts.
• For dynamic populations, the time to follow up is usually based on
individual participants’ dates of enrollment rather than on a fixed
calendar date.


Longitudinal Cohort Study: Retention

For prospective and longitudinal studies, loss of participants to follow-up before the end of the study period is a major concern.

Researchers must develop strategies that minimize the burden of participation and that maximize interest in continuing to participate.


Longitudinal Cohort Study: Strengths

• Can examine multiple effects of a single exposure

• Can elucidate temporal relationship between exposure and disease

• If prospective, minimizes bias in the ascertainment of exposure

• Allows direct measurements of incidence of disease in the exposed and unexposed groups


Descriptive observational studies

Case Report / Case Series: studies without comparison groups [not really an epidemiological design]

Ecological Surveys: compares group characteristics (no individuals) [not really an epidemiological design]

Cross-Sectional Surveys: exposure and outcome measured at same time


Analytic observational Studies

Case-Control Study: participants are selected based on disease status, then past exposures are assessed.

Cohort Study: participants are selected based on exposure status or population membership, then future incidence of outcomes/diseases assessed.


Analytic experimental (interventional) studies

The exposure is assigned

Randomized Controlled Trials (RCTs): especially double- or triple-blinded studies, are the ‘gold standard’ for causal research, but there are often ethical concerns.


Types of Study Designs

Primary, Secondary and Tertiary


Primary Study Design

Collect and analyze new data


Secondary Study Design

Analyze existing data


Tertiary Study Design

Review and synthesize the literature