question

Epidemiological Studies

answer

-Observational -Experimental

question

Randomized Trial

answer

A planned experiment where participants are randomly assigned to a treatment or control group.

question

Types of Randomized Trials

answer

-Natural Experiment -Community Trials -Cluster Randomized Trials -Individually Randomized

question

Natural Experiment

answer

Unplanned events that exhibit a planned trial. Levels of exposure to a causative agent differs within the population and thus affects people differently. Ex: John Snow - Cholera and the Broad Street Pump

question

Community Trials

answer

An experimental study where one group of people or section of the community receives an intervention while the other group/section does not. Ex: The intervention to determine if fluoride in the water benefits community A in a way that community B, with no fluoride in the water, is not benefited.

question

Cluster Randomized Trials

answer

Clusters (i.e. individuals in communities, hospitals, or other aggregates of individuals) are randomized rather than individuals; Less prone to individual-level selection biases Useful when:

question

When to Use Cluster Randomized Trials

answer

- Nature of the intervention (e.g., mass media campaign, population-level interventions) - Acceptability and reduced stigma (everyone gets the same treatment within a cluster) - Can get data on many nested population subgroups within clusters - Allows assessment of population-level effects (Population Attributable Fraction)

question

Individually Randomized

answer

- Conventional medical RCTs (clinical trials) or behavioral interventions - Only eligible individuals are randomized - Self-selection: volunteers more likely to enroll

question

FDA/WHO classification of RCTs (Phase 1-4)

answer

- Phase I: Initial studies to determine metabolism, pharmacologic actions, and safety of drug in humans - Phase II: Controlled clinical studies evaluating preliminary efficacy of drug in patients with disease , determine common short-term side effects - Phase III: Expanded trials for gathering additional information on overall benefit-risk relationship of the drug [Needed for FDA approval] - Phase IV: post-marketing trials in general population

question

Key features of RCTs

answer

- Randomization - Controlled - Blinding

question

Randomization

answer

A method of assigning participants to different arms of a trial. Ensures that intervention and control groups "look alike" with respect to all other factors except for the treatment. Done to avoid any bias in assignment to control/experiment arm that may be introduced by the investigator.

question

Whens it's Unethical to Randomize

answer

- An effective treatment already exists: can't give a placebo - Interventions are very different (i.e. all the different forms of birth control) - can only randomize Pill A vs Pill B - Risks of new treatment likely to exceed risks of existing treatment

question

Controlled

answer

The control condition provides a comparison arm by which the investigator gauges the effect of the treatment - Pre-specified hypotheses - Established inclusion/exclusion criteria - Primary and secondary endpoints (e.g., behavioral change, HIV/HCV incidence) to evaluate hypotheses - Carefully detailed study protocols (enrollment, treatment delivery and follow up) - Rigorous monitoring of treatment and outcomes - Known analysis plans and stopping rules

question

Blinding

answer

Randomization to the intervention and control trial is unknown during the study. - avoids bias in enrollment - avoids bias during the trial - avoids bias during follow-up

question

Three Levels of Blinding

answer

- Single blind: participants are blinded but investigators are aware of who is in the intervention and control arm. - Double blind: neither participants nor investigators know who is receiving the intervention. - Triple blind: participants, investigators, and data analyst don't know intervention assignment.

question

Ways to Randomize

answer

- Random number table - Toss a coin for each subject: heads=A, tails=B - Computer generated programs - Sealed envelopes with randomization info

question

Ways to Assure Compliance

answer

- Call participants the day before clinical visits - Provide reimbursements - Monitor adherence to the intervention protocol:

question

Reasons Loss to Follow-Up Must Be Reduced

answer

- Losses are often selective (e.g., high risk persons drop out of trials) and this introduces bias - Losses to follow up should be comparable in the intervention and control arms to avoid biased comparisons - Losses to follow up reduce study power by reducing the person-time of observation

question

Ethical Considerations

answer

- Equipoise - Informed Consent - Stopping Rules - Intention to Treat

question

Equipoise

answer

There must be genuine doubt about efficacy of treatment yet sufficient belief that it may work; must honestly not know whether or not the treatment will be better

question

Informed Consent

answer

an ethical principle requiring that research participants be told enough to enable them to choose whether or not they wish to participate int he study

question

Stopping Rules

answer

What if it becomes apparent, before the trial is over, that the new treatment is beneficial (and should not be withheld from the placebo group) or is harmful (and treatment should be withdrawn)? Data Safety and Monitoring Boards (DSMB)

question

Intention to Treat

answer

- All participants allocated to a given arm of the trial are analyzed together as representing that trial arm irrespective of whether they received or completed the prescribed treatment - If the analysis is not based on original assignment, you are breaking the randomization and the groups may not be comparable anymore - Results can, otherwise, be invalidated

question

Assessing Associations in RCTs

answer

- Outcomes at fixed points in time -Proportion with outcome at each follow up -Logistic regression (Odds Ratio) - Events in person time -Rate of outcomes per 100 person years -Poisson regression, incidence rate ratio (IRR) - Time-to-event -Time from enrollment until outcome -Cox proportional hazard regression, hazards ratio (HR) -Kaplan-Meier survival analyses, log rank test

question

RCTs Weaknesses

answer

- Often more costly and more timely - Many research questions are not suitable for experimental designs b/c of ethical barriers & b/c of rare outcomes - Many research questions are not suitable for blinding - Standardized interventions may be different from common practice - May have limited generalizability due to the use of volunteers, eligibility criteria, and loss to follow-up

question

RCTs Strengths

answer

- Demonstrate cause-effect relationships - May produce a faster and cheaper answer than observational studies (e.g. compared to cohort studies) - Only appropriate approach for some research questions - Allow investigators to control the exposure levels as needed

question

External Validity

answer

the ability to make sound generalizations from the study to the target population note: the study sample must be representative of the target population

question

Internal Validity

answer

the ability to draw sound conclusions about the relationship between exposure(s) and outcome observed in a study; did the study do what it was supposed to?

question

Types of Error

answer

1. Random 2. Systematic

question

Random Error

answer

Fluctuation around the true value of a parameter due to: -Sampling variability - sample does not represent the target population -Poor precision - exposure and/or risk factors are not measure 'sharply' or accurately -Measurement variability - assessment tool yields unstable results

question

Systematic Error

answer

aka BIAS stems from the design, conduct or analysis of a study that results in a mistaken estimate of an exposure's effect on disease note: can not be reduced by increased sample size

question

Direction of Bias

answer

Positive: observed value is higher than the stratum Negative: crude value is lower than the stratum

question

The Three Types of Bias

answer

- Selection Bias - Information Bias -Confounding

question

Selection Bias

answer

Flawed selection in study participants Ex: -Selection procedure; differential application of eligibility criteria to cases and controls -Differential unavailability due to death ("selective survival"), illness, migration, or refusal to participate (nonresponse bias) -Loss to follow-up / Attrition

question

Differential Loss to Follow-Up

answer

Lose more people in one group than in the other

question

Non-Response Bias

answer

bias resulting when individuals selected to be in a survey either cannot be contacted or refuse to answer survey questions.

question

Selective Survival

answer

bias of concentrating on the people or things that "survived" some process and inadvertently overlooking those that didn't because of their lack of visibility. This can lead to false conclusions and limits external validity

question

Preventing Selection Bias

answer

- Have clear definition of study population - Use explicit case and control definitions - Derive cases and controls from same population

question

Information Bias

answer

Flaws or inaccuracies in measurement or classification of relevant information Differences in accuracy -Of exposure data between cases and controls -Of outcome data between different exposure groups Study subjects are classified in the wrong category

question

Sources of Information Bias

answer

- Respondent: inability to understand, recall, articulate; unwillingness to disclose or social desirability - Data collector: unclear or ambiguous questions, lack of a neutral demeanor, insufficiently conscientious, inaccurate transcription, fraud - Data managers: inaccurate transcription, misreading, miscoding, programming errors - Data analysts: variable coding and programming errors - Study investigator: inadequate appreciation of the characteristics of the measure or of the relations being studied

question

Types of Information Bias

answer

1. Surveillance/observer bias 2. Reporting bias 3. Misclassification: are error processes different for the

question

Surveillance/observer bias

answer

Disease ascertainment in monitored pop. better than in general pop.

question

Reporting bias

answer

- Recall bias: differential recall of exposure between cases and controls - Socially desirable responding

question

Misclassification

answer

Are error processes different for the groups (cases/controls or E/~E) being compared - Yes=differential misclassification of exposure/outcome - No=non-differential misclassification of exposure/ outcome

question

Differential v Non-Differential Misclassification

answer

Non-Diff: misclassification is similar in both cases and controls; Usually biases estimate of association towards 1 (the null) Diff: misclassification is unequal in cases and controls

question

Confounding

answer

Inappropriate comparison between groups; association may not truly exist Note: confounders are not biases, rather the act of neglecting to control for them is

question

Confounder Criteria

answer

1. Risk factor for the disease 2. Associated with the exposure 3. Not on the causal pathway

question

Confounder Diagram

answer

DRAW IT USING: speed, car crashes, and age

question

Direction of Confounding Try It Crude OR.... 1.20 Stratum1 OR... 1.75 Stratum OR....... 1.80

answer

Negative Confounding: bias toward the null (crude association smaller than the adj. association)

question

Direction of Confounding Try it Crude OR.... 2.57 Stratum1 OR... 1.02 Stratum2 OR....... 1.03

answer

Positive Confounding: bias away from the null (crude association greater than the adj. association)

question

Qualitative Confounding

answer

When the crude and adjusted values are on the opposite sides of "1"

question

Methods to Addressing Confounding (Design Phase)

answer

1. Randomization: controls for both know and unknown factors; eliminates relationship btwn confounders & exposure 2. Matching (only in case-control): case and control matched on a 3rd variable 3. Restricting: limit subject population to a specific category

question

Disadvantages of Matching in Design Phase

answer

-Logistically difficult -Over Matching -Cannot later study the matched factors

question

Methods to Addressing Confounding (Analysis Phase)

answer

1. Stratification: hold confounder constant and examine relationship within stratum (analyzes one confounder at a time) 2. Adjustment: use a statistical test to estimate the association IF the confounder was not present (analyzes multiple confounders at once)

question

Mediator

answer

when the third variable being controlled for is on the causal pathway; the extraneous factor is result of the exposure

question

Effect Modification

answer

EM occurs when the relationship between two factors (exposure and outcome) is different across different levels of a third factor

question

Interaction

answer

when the compared stratum are very different from one another (as opposed to the stratum being alike with confounders) crude estimates are not used to evaluate interaction

question

Confounding vs Effect Modification

answer

- Either, both or neither may exist - Confounder: a) Depends on distribution of the confounder among strata of the exposure of interest b) Obscures the true relationship between an exposure and an outcome c) Thus, is a nuisance effect to be adjusted for - Effect Modifier: a) An inherent feature of the strata to be described Alters effect in size/direction among strata

question

Independent Risk Factor vs. Confounding vs. Mediation vs. Effect Modifier

answer

-Independent Risk factor: there is no 3rd variable; direct relationship (E causes O) -Confounding: 3rd variable obscures the true relationship (E causes O but X also leads to O, and the relationship between E and X may make the relationship between E and O look stronger or weaker than it truly is) -Mediation: the 3rd variable appears in the middle the relationship (E causes X then X causes O) -Effect Modifier: 3rd variable alters the strength of the association (E causes O but X creates an even stronger association when present)

question

Draw the DAG for Independent Risk

answer

should look like a direct line from Exposure to Outcome

question

Draw the DAG for Confounding

answer

should look like a triangle in which: 1. the third variable is on top 2. the exposure is on the bottom left 3. the outcome is on the bottom right 4. there is a one way solid line from exposure to outcome 5. there is a one way solid line from third variable to outcome 6. there is a two way dotted line from third variable to exposure

question

Draw the DAG for Mediation

answer

should look like a horizontal line from exposure to third variable and from third variable to outcome

question

Draw the DAG for Effect Modifier

answer

should look like a direct line from exposure to outcome then there should be a line pointing down to the middle of the line from third variable

question

Henle-Koch Postulates (Causality)

answer

1. The parasite must be present in all who have the disease 2. The parasite can never occur in healthy persons 3. The parasite can be isolated, cultured and capable of passing the disease to others

question

Bradford Hill Casual Inference (Causality)

answer

1. Temporality 2. Strength of association 3. Dose-response 4. Consistency 5. Biological plausibility 6. Consideration of alternate explanations 7. Cessation 8. Coherence 9. Specificity

question

Temporality (Bradford Hill)

answer

If a relation is causal, exposure must precede outcome

question

Strength of Association (Bradford Hill)

answer

If a relation is causal, the larger the association, the more likely it is that the exposure is causing the disease; strong associations are less likely to be caused by chance

question

Dose-Response (Bradford Hill)

answer

If a relation is causal, changes in exposure are related to trend in risk of disease; typically, the greater the exposure, the greater the probability for outcome

question

Consistency (Bradford Hill)

answer

If a relation is causal, multiple studies can demonstrate this same relationship

question

Biological Plausibility (Bradford Hill)

answer

If a relation is causal, the proposed mechanism should be etiologically plausible; can we explain the relationship on the basis of existing biology?

question

Alternate Explanations (Bradford Hill)

answer

If a relation is causal, investigator have ruled out other possible explanations

question

Cessation (Bradford Hill)

answer

If a relation is causal, the risk of disease expected to decline when exposure to a cause is reduced or eliminated

question

Coherence (Bradford Hill)

answer

If a relation is causal, would expect observed findings to be consistent with other epidemiological and biologic knowledge

question

Specificity (Bradford Hill)

answer

Specific exposure associated with only one disease (idea held on to from Henle-Koch postulates; caveat... many exposures are linked to multiple diseases

question

Sufficient Cause

answer

a set of minimal conditions or events that inevitably produce disease (the entire pie)

question

Necessary Causes

answer

a component cause that is a member of every sufficient cause (the one factor that is a part of each and every pie for the specific outcome)

question

Component Causes

answer

any one of a set of conditions that contribute to the completion of a sufficient cause (the factors of the pie other than the necessary cause)

question

Screening

answer

-tests, exams, procedures to identify asymptomatic subjects -secondary prevention (find ppl early in the natural history to reduce morbidity) -not intended to diagnose (doesn't confirm whether or not disease is present

question

Reasons to Screen Rather than Just Diagnose

answer

-cheaper -faster -sensitive -able to reach a large setting -less invasive

question

Screening Requirements

answer

1. suitable disease 2. suitable test 3. suitable screening program

question

Suitable Disease (screening)

answer

-must be progressive -must have a lengthy detectable preclinical phase, -must be treatable -must have a high prevalence of pre-clinically detectable people (screening meant for large population)

question

Suitable Test (screening)

answer

-Inexpensive -Easy to administer -Cause minmal discomfort -High validity -High reliability

question

Validity

answer

Ability of a test to correctly classify people with pre-clinical disease as positive and people without pre-clinical disease as negative

question

Reliability

answer

Ability of a test to give consistent results when performed more than once on the same individual under the same conditions

question

Sensitivity (validity measurement)

answer

Probability of correctly classifying those WITH disease as cases (Cases Identified / All Cases) ...or ( A / A+C )

question

Specificity (validity measurement)

answer

Probability of correctly classifying those WITHOUT disease as non-cases (Non-Cases Identified / All Non-Cases) ...or ( D / B+D )

question

Positive Predictive Value

answer

Proportion of true cases among all positive test results (True Positive / All Positives) ...or ( A / A+B) Note: increase in prevalence of disease increases the PPV

question

Negative Predictive Value

answer

Proportion of true negatives among all negative test results (True Negatives / All Negatives) ...or ( D / C+D)

question

Screening Biases

answer

1. Compliance Bias 2. Lead-Time Bias 3. Length Bases Sampling Bias

question

Compliance Bias

answer

Volunteer bias where people who choose to participate in the screening program may be healthier or at higher risk of developing the disease than those that don't participate

question

Lead-Time Bias

answer

(Note: Lead-time is the amount of time by which the diagnosis was advanced due to screening_ Lead time bias means that survival may erroneously appear to be increased among screen- detected cases simply because the diagnosis was made earlier in the course of the disease.

question

Length-Biased Sampling

answer

less aggressive forms of a disease are more likely to be picked up in a screening program because of a longer detectable preclinical phase, so it appears as though screened disease have better survival

question

Individual-Level vs Population-Level Screening

answer

Ind: patient-physician level (i.e. annual check-ups) Pop:national-level policy decision (i.e. mandatory breast screening in women 40+)

question

Inter-observer Variation and Kappa

answer

The extent to which observers (or raters) agree or disagree is an important issue Kappa: overall precent agreement ( A+D / total number of readings) x 100

question

Outbreak Investigation Steps

answer

- Verify the diagnosis - Confirm the outbreak -Come up w/case definition - Descriptive epidemiology (what do cases look like) - Develop a hypothesis - Test the hypothesis - Refine hypothesis/Execute additional studies - Implement control and prevention measures - Communicate findings

question

Attack Rate (outbreak)

answer

(# ppl at risk who develop the disease) / (total people at risk)

question

Food-Specific Attack Rate (outbreak)

answer

(# ppl who ate a specific food and got sick) / (total # ppl who ate the food)

question

Secondary Attack Rate (outbreak)

answer

(total number of cases - initial cases) / (# of susceptible ppl in group - initial cases)

question

Appropriate Measure of Association (Case-Control v Cohort)

answer

Case-Control = Odds Ratio (AD/BC) Cohort = Risk Ratio (A/A+B)/(C/C+D)

Epidemiology NYU Sp13 Pt. 2 – Flashcards

Unlock all answers in this set

Haven't found what you were looking for?

Search for samples, answers to your questions and flashcards