• Shuffle
    Toggle On
    Toggle Off
  • Alphabetize
    Toggle On
    Toggle Off
  • Front First
    Toggle On
    Toggle Off
  • Both Sides
    Toggle On
    Toggle Off
  • Read
    Toggle On
    Toggle Off
Reading...
Front

Card Range To Study

through

image

Play button

image

Play button

image

Progress

1/159

Click to flip

Use LEFT and RIGHT arrow keys to navigate between flashcards;

Use UP and DOWN arrow keys to flip the card;

H to show hint;

A reads text to speech;

159 Cards in this Set

  • Front
  • Back
What is epidemiology?
-The study of the distribution of disease and the determinants of disease
-Operationally: counting the cases of disease, ascertaining how much disease is in a population, determining the variables that are associated with disease, and identifying the factors that are causes of disease and are potentially modifiable
What are the 4 measures of occurrence?
1. Prevalence
2. Risk (Incidence proportion)
3. Incidence Rate
4. Odds
Define Prevalence
# cases of disease / # of people in population

Unit: at a specified point in time (usually point prevalence, not period prevalence)

Range: 0-1

It measures who have a disease

+: Measure of existing disease, especially useful with onset or risk is hard to obtain
-: not causal, cross sectional, could be due to incidence or duration, not good for diseases of short duration or high fatality
Define Risk (Incidence proportion)
# of new cases of disease / Total number of persons followed that are disease free initially (often follow up)

Units: Over a time period (***Risk is MEANINGLESS w.o a time interval)

Range: 0-1

The number of people who develop new or incident cases of disease during a certain period of time.

Applies to an individual and refers to the probability that a person will develop a given disease. Seldom measured at the individual level.

+: Easy to interpret and for lay people to understand, valid with short follow up and low competing risks
-: difficult to measure in reality due to LFU and death from CR.
Loss to follow up
occurs when people drop out of your study for any reason.

- didnt want to participate any more
-moved
-changed contact info
- etc
Competing risks
when people are removed from a study via death.

They diem so they do not live long enough to experience your outcome.

Makes it difficult to use risk as the only assessment of disease occurrence in a study.

Competing risks is a specific type of loss to follow up.
Define Incidence Rate
# of new cases / Total time a risk of persons followed (amount of person time at risk)

Unit: cases / person time OVER TIME PERIOD

Range: 0 - infinity

Refers to how rapidly new disease events are occurring in a population. Can be conceptualized as instantaneous or average rate.

Assumptions:
1. People at risk for a disease until: they get disease, die, LFU, or study ends
2. When people become ill, they become ill on the last possible part of the time period, so they are at risk for the time period when they became ill. Count the time period where they became ill in person time at risk.

+: Measures onset, accommodates LFU and CR, good when people contribute varying amounts of time.
-: Difficult for lay people to understand.
Factors leading to increased prevalence
-Increased incidence rate
-Increased duration of disease, prolonged survival
-More case finding
-Lower mortality
Define Odds
p / (1-p)

Can be either prevalent or incident odds. Prevalent odds is more common.

If odds are 75%, this means a horse has 75% chance of losing a race.

+: Approximates incidence x duration in a steady state
-: less common measure, largely cross sectional
Annual mortality rate
# of deaths in 1 year / # of people in population at mid-year

X 1000
Case fatality rate
total # of deaths / # of persons with disease

x 1000

*death for ANY reason
Prevalence Ratio
Prevalence exposed / Prevalence unexposed
Risk Ratio
R exposed / R unexposed

Ranges from 0 - infinity
No units
Risk Difference
R exposed - R unexposed

Ranges from -1 to 1
No units
Incidence Rate Ratio
IR exposed / IR unexposed

Ranges from 0 to infinity
No units
Incidence Rate Difference
IR exposed - IR unexposed

Ranges from - infinity to + infinity
Unit: time -1 (cases per unit time)
Odds Ratio
odds exposed / odds unexposed

Can be either prevalence or incidence
No units

OR ad/bc
Absolute vs relative scales
absolute: the absolute increase or decrease in effect (helpful for understanding the population impact)

relative: relative increase or decrease in effect (helpful for etiologic association)

May be interested in both
Absolute Null
0
Relative Null
1
Attributable fraction among the exposed
(attributable risk among the exposed)
AF exposed = (R exposed - R unexposed)
/
R exposed

(OR RR - 1 over RR)

the proportion of disease cases among the exposed that would be removed if there were no longer any exposure
Population Attributable Fraction
AF pop = (R pop - R exp) / R pop

the proportion of disease cases in the population that would be prevented if there were no longer any exposure
What is counterfactual thinking about cause?
A person is exposed to a certain exposure, and develops a disease. In an alternate universe, the same person is not exposed to that exposure, and doesn't develop the disease.

Then that exposure is a cause of that disease.
Hill criteria for cause
1. Strength of association
2. replication of findings
3. Specificity of association
4. Temporality******
5. Dose-response
6. Biological plausibility
7. Coherence with established facts
8. Experimental evidence (cessation with exposure)
9. Analogy
Sufficient cause
a set of minimal conditions or events that INEVITABLY produce disease. (causal pie is full)
Necessary cause
A component cause that is a member of ever sufficient cause. Must be in every disease pie.
Limits of sufficient cause model
-omits discussion if origins of cause
-focuses on proximal cause
-shows components but not links between them
-does not consider factors that control distribution of risk factors
-ignores dynamic, non-linear relations
Can cause be proven?
No. can be only supported or refuted. Strongest evidence is from strongest study designs.
What is a cohort?
1. A population that is surveyed at a given moment in time
2. A group of persons sharing a particular statistical or demographic characteristic
What is a cohort study?
A study that enrolls people and follows them over time (longitudinal) to measure the incidence of some disease or health outcome.

People are selected by:
1. membership in defined cohort
2. exposure status

Disease status is NOT known at enrollment.
Fixed cohort
fixed membership. once group is defined and follow up begins, no one can be added
Open Cohort (dynamic cohort)
can take on new members during a study period and people can leave
External validity
the generalizability of the study with regard to the rest of the population in terms of time, place, and types of people
What measures of association can you calculate from a cohort study?
-risk
-incidence rate
-odds
-prevalence
-risk ratio
-risk difference
-odds ratio
-incidence rate ratio
-rate difference
-AF exp and pop

All of them
In a cohort study, if a disease is rare...
RR ~ OR
Case control study
an observational study design that enrolls people in a study based on their disease outcome status.

the investigator does not know the subject's exposure status before the study begins. (opposite of a cohort study).
How are controls selected in a case control study?
Controls are selected that are representative of the population at risk that gave rise to the study cases.

Must belong to the same population of origin of the cases, or the same source population.
Where do cases come from in a case control study?
Should preferably be new.

1. population based cases
2. hospital based cases
3. other sources
Source population
The underlying population that gave rise to the cases in the study.
What are the steps for selecting cases and controls in a case control study?
1. select cases
2. determine source population
3. sample controls that represent the underlying distribution of exposure
4. ascertain exposure status of cases and controls
5. calculate measure of association.
What measure of association can we calculate for case control studies?
ONLY odds ratio.
In case control studies, if disease is rare, then...
IRR ~ RR ~ OR
What are the study designs, from strongest to weakest?
1. experimental
2. cohort
3. case control
4. cross sectional
What is an experimental study?
Also RCT. Are a specific type of prospective cohort study in which exposure is randomly assigned by the investigator.
Randomization
procedure by which participants are randomly assigned to treatment groups to receive different exposure levels
Why are experimental studies (RCT) the gold standard? Why are they so strong?
Because with the use of randomization, we control for confounding.

eg, exposure and control groups are exchangeable. Exactly the same except for exposure, which we give.
Confounding by indication
bias that results when different types of patients receive different types of treatment because of the severity of their sickness.
Random sampling
Sampling strategy. a way to get people into the study, to select them as representative of the population. NOT necessary for RCT, as is randomization.
What are 4 key features necessary for a RCT?
1. randomization
2. Assigned exposure
3. Equipoise/ Do no harm
4. ITT analysis
Blinding
single: either patient or investigator knows the treatment assignment, but the other does not.

double: both groups dont know.

Purpose is to reove systematic error, often as information bias or differential care.
Internal validity
The ability to reach the correct conclusion among those actually studied.

The absence of bias.
When is RCT appropriate?
WHen the exposure is modiafiable.

When there is equipoise.

When a particular exposure may have an influence on multiple outcomes of importance.
What measures of association can you calculate from an experiment?
RR, IRR, RD, IRD, OR

Odds ratio is NOT preferred, although possible.
Contamination
Where the experimental group influences the control group by coming in contact and trading information.
What are 3 types of RCTs?
1. Clinical trials
2. Field trials
3. Group randomized trials.
Validity
Concerns whether or not there are imperfections in the study design, the methods of data collection, or methods of data analysis that might distort conclusions made about an exposure-disease relationship.

If there are no imperfections, the study is valid.
Bias
Systematic error in the design (wrong design, wrong sampling strategy), conduct (problems in enrollment of cases or controls, loss to fup, poor collection of data), or analysis of a study (wrong modeling assumptions, miscategorization of variables) that results in a mistaken estimate of an exposure's effect on disease.
Threat to validity
When something has the potential to bias the study.
The fundamental issue in statistical inference is quantifying our certainty about
how well the findings in our study population reflect the source population given that there is always a role of chance.
Two main approaches to statistical inference in epi:
1. hypothesis testing and p values
2. confidence interval estimation
Selection Bias
Systematic error in selecting subjects into one or more of the study groups, such as cases and controls, or exposed and unexposed.

Where the study population does not mirror the source population.

Can occur in any type of study design, but commonly occurs in case controls studies: how cases or controls are selected into the study, and how well this reflects exposure history.
In cohort studies and RCTs, primary sources of selection bias are
1. loss to fup
2. withdrawal
3. non response

*The health outcomes does not influence how people get into the study, but if the health outcomes is not ascertained among everyone, the cohort may no longer represent the source population.
Information bias
Systematic error arising because or incorrect information obtained on one or more variables measured in the study.

Also called measurement error or misclassification.

EG: bias in recall, in collecting data, in interview, or in reporting.
Construct v. Measure
construct: the theoretical thing you are trying to measure, in terms of what it is, when it occurred, etc.

Measure: actually how it was measured.
Misclassification
If we mismeasure and have binary categories of exposure and disease, it won't reflect the true 2x2 table.

The effect estimate may be biased.
Misclassifying disease status can occur via
1. Incorrect dx. Due to limited knowledge, complex diagnostic process, inadequate access to technology, lab error, subclinical disease, detection bias.

2. Self report. Incorrect recall or illness in the past, social desirability bias, reluctance to admit a disease to oneself.

3. Patient records incorrectly coded in database.
Misclassifying exposure status can occur via
1. imprecise measurement (random error). Due to poorly constructed questionnaire, faulty measuring device.

2. Subject self-report. Due to reluctance, or failure to recall.

3. Interviewer bias. May more thoroughly probe about exposure among cases than for controls.

4. Incorrect coding of exposures in database.
Two types of information bias
Nondifferential misclassification (biased towards the null) and differential misclassification (biased towards or away from the null)
Nondifferential misclassification
Type of information bias where either exposure or disease is misclassified. BUT misclassification of disease is the same among non/exposed, or misclassification of exposure is same among non/diseased.

Measurement is equally good/poor for disease and non disease, or equally good/poor for exposed and non exposed.

Association is always biased towards the null.
Differential misclassification
Information bias where misclassification of exposure is greater among non/diseased, OR misclassification of disease is greater among non/exposed.

Mismeasurement is different.

Association can be biased towards or away from the null.
Misclassifying both exposure and disease
Can occur when subjects self report exposure and disease.

When there is misclassification of both exposure and disease, the association will be biased towards the null.
Non respondent bias
bias resulting when persons who do not participate in a study may be different from those who due. a special type of bias.
Recall bias
When cases and controls differentially recall their past/prior exposure. Cases often recall better. This is a problem when measuring exposure retrospectively. A special type of bias
Healthy worker bias
If using workers in a study instead of general population, workers dont represent the general population. Those who are employed are healthier than the general population. A special type of bias.
Volunteer bias
Bias resulting from volunteers. Volunteers are often very different in motivations from non volunteers. A special type of bias.
What are ways to prevent selection bias?
1. case control: make sure controls represent the source population, try to use prospective studies, ensure equal participation.

2. Occupational cohort studies: be sure that unexposed group is equally as healthy as exposed group.

3. Cohort studies: maximize participant follow up.
What are ways to prevent information bias?
1. standardization of measurement and instrument

2. Staff training and quality control

3. Blinding of investigator and participant
Confounding
When a non-causal association between a given exposure and an outcome is observed as a result of the influence of a third variable, that third variable is usually designated as a confounder.

A variable that produces a distortion in the true effect estimate or measure of association.

We can have small and large confounding, not just yes and no.
Most common potential confounders:
Age and gender
Is nondifferential or differential misclassification exposure preferred?
Non differential is preferred because we can predict the direction of the bias (will always be towards the null). For non differential, we don't know whether the ratio will be biased towards or away from the null.
How does confounding relate to counterfactual thinking?
Confounding is an exchange of viability. The presence of some other variable or variables make the counterfactual scenarios non-comparable.

When confounding occurs, the apparent effect of exposure on outcome is distorted because the effect of an extraneous factor is mistaken for an actual exposure effect.
A variable is a potential confounder if:
1. it is a risk factor for (or is associated with) the outcome

2 It is associated with the exposure

3. It is NOT on the causal pathway between exposure and outcome (NOT a mediator). (this can't be derived with data).
What are some ways to control for confounding?
Study design:
-randomization
-restriction
-matching

Analysis:
-restriction
-stratification (can calculate by hand)
- Adjustment
Randomization
Randomly assigning subjects to different treatment groups to balance all potential confounders between treatment groups.

Only appropriate in experimental studies.

This will break rules for confounding: the confounder is NOT associated with exposure.
Restriction
Selecting subjects for a study who have the same value for a variable that might be a confounder. Or selecting people who are the same based on one value of a potential confounder. Include only one level of the confounding variable in the study.
Matching
Process of choosing someone with a similar confounding factor to participate in the study, based on exposure (in cohort) or disease (in case control).
Stratification
process of examining the effect of one variable on disease within categories of a third variable. Analyze groups separately.
Adjustment
Statistical technique to estimate what the association would be IF the confounder was not associated with the exposure. Use either regression or standardization.
What are some ways to control for confounding?
Study design:
-randomization
-restriction
-matching

Analysis:
-restriction
-stratification (can calculate by hand)
- Adjustment
What are the two methods of adjustment?
1. Regression (allows us to model the associateion between an exposure with an outcome, controlling for multiple covariates simultaneously.

2. Standardization (calculates weights against another population to make different rates comparable between 2 or more groups)
Steps to determining if a variable is a confounder via stratification
1. Calculate the CRUDE measure of association between exposure and disease (ignore 3rd variable)

2. calculate a stratum-specific measure of association between exposure and disease within strata of the confounder.

Compare the results of the stratum specific estimates with each other and with the crude estimates.

IF stratum specific measures are the same, and these values differ from the crude measure, then there is confounding. Report on adjusted measure.
Randomization
Randomly assigning subjects to different treatment groups to balance all potential confounders between treatment groups.

Only appropriate in experimental studies.

This will break rules for confounding: the confounder is NOT associated with exposure.
Effect Measure Modification
Situation where the size of a measure of effect/association changes (or is different) over the value of some 3rd variable.

The association between exposure and disease is different for different groups, on a third variable.
Restriction
Selecting subjects for a study who have the same value for a variable that might be a confounder. Or selecting people who are the same based on one value of a potential confounder. Include only one level of the confounding variable in the study.
What are the steps to determine effect measure modification?
1. Identify exposure, disease, and potential modifier.

2. Choose scale to measure the association.

3. Calculate the measure of association between exposure and outcome, stratified by the potential effect modifier.

4. Compare: How similar are the associations between the two strata?
- If they are similar, effect modification is NOT present
- if they are different, effect modification IS present.
Matching
Process of choosing someone with a similar confounding factor to participate in the study, based on exposure (in cohort) or disease (in case control).
Stratification
process of examining the effect of one variable on disease within categories of a third variable. Analyze groups separately.
Adjustment
Statistical technique to estimate what the association would be IF the confounder was not associated with the exposure. Use either regression or standardization.
What are the two methods of adjustment?
1. Regression (allows us to model the associateion between an exposure with an outcome, controlling for multiple covariates simultaneously.

2. Standardization (calculates weights against another population to make different rates comparable between 2 or more groups)
Steps to determining if a variable is a confounder via stratification
1. Calculate the CRUDE measure of association between exposure and disease (ignore 3rd variable)

2. calculate a stratum-specific measure of association between exposure and disease within strata of the confounder.

Compare the results of the stratum specific estimates with each other and with the crude estimates.

IF stratum specific measures are the same, and these values differ from the crude measure, then there is confounding. Report on adjusted measure.
Effect Measure Modification
Situation where the size of a measure of effect/association changes (or is different) over the value of some 3rd variable.

The association between exposure and disease is different for different groups, on a third variable.
What are some ways to control for confounding?
Study design:
-randomization
-restriction
-matching

Analysis:
-restriction
-stratification (can calculate by hand)
- Adjustment
What are the steps to determine effect measure modification?
1. Identify exposure, disease, and potential modifier.

2. Choose scale to measure the association.

3. Calculate the measure of association between exposure and outcome, stratified by the potential effect modifier.

4. Compare: How similar are the associations between the two strata?
- If they are similar, effect modification is NOT present
- if they are different, effect modification IS present.
Randomization
Randomly assigning subjects to different treatment groups to balance all potential confounders between treatment groups.

Only appropriate in experimental studies.

This will break rules for confounding: the confounder is NOT associated with exposure.
Restriction
Selecting subjects for a study who have the same value for a variable that might be a confounder. Or selecting people who are the same based on one value of a potential confounder. Include only one level of the confounding variable in the study.
Matching
Process of choosing someone with a similar confounding factor to participate in the study, based on exposure (in cohort) or disease (in case control).
Stratification
process of examining the effect of one variable on disease within categories of a third variable. Analyze groups separately.
Adjustment
Statistical technique to estimate what the association would be IF the confounder was not associated with the exposure. Use either regression or standardization.
What are the two methods of adjustment?
1. Regression (allows us to model the associateion between an exposure with an outcome, controlling for multiple covariates simultaneously.

2. Standardization (calculates weights against another population to make different rates comparable between 2 or more groups)
Steps to determining if a variable is a confounder via stratification
1. Calculate the CRUDE measure of association between exposure and disease (ignore 3rd variable)

2. calculate a stratum-specific measure of association between exposure and disease within strata of the confounder.

Compare the results of the stratum specific estimates with each other and with the crude estimates.

IF stratum specific measures are the same, and these values differ from the crude measure, then there is confounding. Report on adjusted measure.
Effect Measure Modification
Situation where the size of a measure of effect/association changes (or is different) over the value of some 3rd variable.

The association between exposure and disease is different for different groups, on a third variable.
What are the steps to determine effect measure modification?
1. Identify exposure, disease, and potential modifier.

2. Choose scale to measure the association.

3. Calculate the measure of association between exposure and outcome, stratified by the potential effect modifier.

4. Compare: How similar are the associations between the two strata?
- If they are similar, effect modification is NOT present
- if they are different, effect modification IS present.
What is it called when there is a difference in effect modification between scales?
A statistical interaction, which refers to deviation from underlying model form. The relationship has a mathematical function.
What is the difference between effect modification and interaction?
Effect modification emphasizes the effect of exposure on a health outcome, which is often non-quantitative, clinical, or a biological attribute of a population.

Interaction is a statistical term that emphasizes that exposure variable and the control (3rd) vatiable are interacting within a mathematical way in determining the health outcome. Often data specific, quantitative. Depends on the scale of measurement.
What is the difference between effect modification and confounding?
Confounding indicates a distortion of the association between exposure and outcome, where effect modification indicates a real difference between exposure and outcome between two groups.
On what scales should you calculate confounding and effect modification?
On BOTH the absolute and relative scale.

**Unless a case control: you can ONLY calculate the odds ratio and therefore only use the relative scale.
What bias does screening relate to?
Measurement - information bias.
What are the pros and cons to screening?
PRO:
-find people with disease who don't know they have it.
-treat
-cure
-prevent spread
-slow down natural history
-find disease early
-reduce cost
-money, fame, glory

CONS:
-may be no prognostic benefit, we can't do anything about it.
-Detectable phase of preclinical disease may be too short
-costs may outweigh benefits, including psychological costs.
-Emotional distress
-Risks of invasive procedures.
What is the gold standard?
The best available way to measure something, including the best way to diagnose someone.

Measurement with a gold standard is necessary to determine how good another screening is (its sensitivity and specificity)
Sensitivity (Sn)
Probability of a positive test if disease is present.

Sn = TP / (TP + FN)

Sn = true positives / everyone with disease.

This tells us how good a test is for finding people with disease.
Specificity (Sp)
Probability of a negative test if disease is not present.

Sp = TN / (TN + FP)

Sp = True negative / Everyone without disease.

Tells us how good a test is for confirming people do not have a disease.
Interpret:

Sn = 0.8667
Sp = 0.96
Sn = 0.8667
87% of those who are truly pregnant are classified as pregnant with this new test.

Sp = 0.96
96% of those who are truly not pregnant are classified as not pregnant with this test.
Infectious disease
An illness due to a specific infectious agent or its toxic products that arises through transmission of that agent or its products from an infected person, animal, or inanimate reservoir to susceptible host.
Epidemiologic Triad
Human disease results from interaction between the host, agent, and the environment. A vector may also be involved in the transmission.

Host susceptibility to the agent is determined by a variety of factors including genetic background, nutritional status, vaccination, prior exposure, context/environment.
What is the difference between a vector and a vehicle?
Vector is alive (animal, arthropod) and vehicle is non living (water, food, blood)
What are the two modes of transmission?
Direct - disease is transmitted by direct contact with the agent. (touching, biting, sneezing, contact w fecal matter)

Indirect transmission - disease transmitted by indirect contact with the agent (through inanimate objects, food, airborne, vector-bourne)
Incubation period
interval from receipt of infection to the time of onset of clinical illness (signs and symptoms)
What accounts for the incubation period?
time needed for the pathogen to replicate to the critical mass necessary for clinical disease

Dose of the infectious agent received at the time of the infection
Endemic v epidemic v pandemic
Endemic: the habitual presence of a disease within a given geographic area

Epidemic: the occurrence of a disease clearly in excess of normal expectancy, and generated from a common or propagated source

Pandemic: a worldwide epidemic affecting an exceptionally high proportion of the global population.
Common source outbreak v propagated source outbreak
Common: From a common exposure - a single exposure. EG: contaminated food or water, bioterrorist attack

Propagated: passed from one person to another. EG: std, measles, infectious diseases.
Steps of an outbreak investigation
1. Establish the existence of an outbreak

2. Define and identify cases (clinical info, who are affected, location, time)

3. Describe and orient the data in terms of time, place, and person

4. Develop hypothesis

5. Evaluate hypothesis and conduct analytic studies

6. Implement control and prevention measures

7. Communicate findings
Notifiable disease
Disease for which regular, frequent, and timely information regarding individual cases is considered necessary for the prevention and control of the disease
What are the three classifications of cases that are used to reflect the certainty regarding diagnosis?
1. Confirmed
2. Probably
3. Possible.

Use over laboratory verification and clinical features.
What info should be collected from every affected person during an outbreak?
1. identifying into (name, phone, address)

2. Demo information (age, sex, race, occupation)

3. Risk factor info (ask about known risk factors for the disease)

4. Clinical info (verify the info has been met for every date. Include date of onset of clinical symptoms.
Passive v active surveillance
passive: relies on routine notifications by healthcare personnel

active: regular outreach to potential reporters (physician offices, hospitals, schools)
Descriptive epidemiology
characterizing an outbreak by time, place, and person
What is the difference in the shape of an outbreak curve for common source v propagated epidemics?
Common source: one peak, tail off

Propagated: progressively taller peaks, separated by the incubation period
When developing a hypothesis regarding an outbreak, what should be included?
Source of the agent

Mode of transmission

Exposures associated with the disease should be proposed in a way that can be tested.
Attack Rate
the proportion of a well-defined population that develops illness over a limited period of time, such as during an epidemic or outbreak.

Is really an incidence proportion, even though it is called a rate.

Often expressed as a percent.

Attack rate = number of new cases in a given time period / total number of people at risk
How do you determine which food was the source of an outbreak?
Item with high ratio of attack rates and high risk difference

AND

High attack rate among those exposed

AND

A low attack rate among those not exposed
Quarantine v isolation
quarantine: restriction of persons who are PRESUMED to have been EXPOSED to disease but NOT YET ILL.

Isolation: separation of ill persons with contagious disease.
Herd Immunity
the resistance of a group to an attack by a disease to which a large proportion of the members of the group are immune.

If a large group is immune, the entire population is likely to be protected.

Therefore, it may not be necessary to vaccinate 100% of the population to achieve population immunity.
What are potential sources of error in epidemiological studies?
1. bias - systematic errors. (if there are many, repetitions will not center the data around the true source population)

2. Confounding

3. Sampling variability (random error)
Precision
the relative lack of random error. If there is ONLY random error, many study repetitions, the estimates of effect WILL center around the truth in the source population.
Statistical inference
the use of statistical approaches to determine our confidence in our measurement.
Inference
a generalization made about a source population from a sample of that population.

Involves quantifying the role of random error of the estimate from the sample.

Does NOT give information on the accuracy (truth) of the estimate.

ASSUMES: sample is random without systematic selection bias.
Null vs. alternate hypothesis
Null: assumes there is no effect between exposure and disease

Alternate: directional (higher or lower) or non directional (association) between exposure and disease.
Steps in hypothesis testing
1. Specify null (H0) and alternate (H1) hypothesis.

2. Decide on statistical significance level (usually p< 0.05).

3. Calculate a test statistic with the probability value (called P value)

4. P value is compared with the significance level you chose before.
What does p = 0.05 mean?
In 5% of samples, one would find an association AT LEAST AS STRONG as the one you found by chance alone, when the truth is that the exposure and disease are not associated in the source population.

ASSUMES that null hypothesis is true.
How do you interpret the mix of the P values with the size of association?
p value - large p value = large sample size = precise

association (ratio or difference) = close to 1 no association, far = strong association.
Confidence Interval
CI is a range of possible values around the point estimate (using an upper and lower limit) within which the population parameter of interest actually lies with a stated level of clarity.

Does not mix size of association with precision of estimate.
How would you interpret a confidence interval of 95%?
The CI will include within it the correct value of the measure of association 95% of the time.

OR

Over unlimited repetitions, 95% of the repetitions would contain the true population measure of association.
How do you interpret significance of P value and CI?
P value: if greater than 0.05

CI: If null value of association is contained within the interval.
Why use prisoners as subjects?
definable
little loss to fup
captive population, you have control over them and their activities
What are the basic ethical principles as defined by the NIH/DHHS?
1. Respect for persons (means persons should be treated as autonomous agents AND that people with diminished autonomy are entitled to protection)

2. Beneficence (do no harm/ maximize benefits and minimize possible harms)

3. Justice (injustice when someone is denied what they are entitled to or has a burden imposed unduly)
Nuremberg Code
set of principles intended to prohibit human experimentation wo subjects' consent. Result of postwar Nuremberg trial of 20 Nazi Drs.
What are the ethical requirements for research?
1. informed consent

2. risk/benefit analysis

3. selection of subjects
Informed consent
respect for persons requires that subjects, to the degree capable, be given the opportunity to choose what shall or shall not happen o them. The opportunity is provided when adequate standards for informed consent are satisfied:

1. information

2. comprehension

3. voluntariness
What are forms of scientific misconduct?
Fabrication
Falsification
Plagarism