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;
96 Cards in this Set
- Front
- Back
What is a p-value?
|
The probability of obtaining results as or more extreme, by chance alone, given that H-null is true.
|
|
What is the interpretation of a confidence interval?
|
"The interval was produced by a method that contains the true value 95% of the time in the absence of bias."
|
|
What is the source population?
|
The population from which the study population and base is selected by use of exclusions.
(the pool from which we're pulling people) |
|
What is the study population?
|
The population whose experience constitutes the study base.
(people we select) |
|
What is the study base?
|
The population experience captured in a study.
(experiences we select to study) |
|
What is a confounder?
|
An extraneous determinant of the outcome that biases the association.
|
|
What is a study?
|
A project designed to yield evidence for the advancement of knowledge.
|
|
What are the two overarching categories of epidemiologic study designs?
|
Experimental and observational
|
|
What is an experimental study design?
|
One in which the determinant/exposure is intentionally assigned solely for the goals of the study. (randomized controlled trial)
|
|
What is an observational study design?
|
One in which the determinant/exposure is determined by some other mechanism (non-assigned).
-cohort -case-control -cross-sectional -Ecologic |
|
What is the study base of a cross-sectional study?
|
A population cross-section (as opposed to a population experience over time)
|
|
When is the outcome assessed in a cross-sectional study?
|
At a SINGLE point in time. (A study of PREVALENT cases)
Determinant usually assessed at a single point in time. |
|
What can be uncertain about cross-sectional studies?
|
Temporality of determinant and outcome.
(Which came first - determinant or outcome?) |
|
What does a cross-sectional study allow for the estimation of?
|
The prevalence ratio and prevalence difference.
|
|
Is there follow-up in a cross-sectional study?
|
No.
|
|
Which measures of association can be estimated in a cross-sectional study?
|
Prevalence rate or prevalence difference.
|
|
What are the strengths of a cross-sectional study? (there are 2)
|
-Relatively quick and inexpensive
- Good for early stages of knowledge |
|
What are the limitations of a cross-sectional study? (there are 2)
|
-Can have unclear temporality of determinant and outcome
-Since cross-sectional studies use prevalent cases, the duration of illness can bias estimated association if the determinant affects the duration of illness. |
|
What is the key feature of an ecologic study?
|
It is a study of the determinant and the outcome at the population level, and NOT an individual level.
|
|
What type of bias is an ecologic study subject to?
|
Ecologic fallacy!
|
|
How would the study below be changed to an ecologic study?
Does smoking cause death from coronary disease? |
Do states with higher smoking levels have higher coronary disease mortality rates than states with lower smoking levels?
|
|
What is population level information?
|
Each observation represents a group (e.g. countries, cities, year) rather than an individual.
Ex: Particulate concentration in air OR asthma hospitalization rate |
|
What would the determinant and outcome be for an ecologic study asking:
Do states with higher smoking levels have higher coronary disease mortality rates than states with lower smoking levels? |
Determinant: Per capita cigarette sales
Outcome: heart disease mortality rates |
|
What is ecologic fallacy bias?
|
Estimates of an effect in an ecologic study does not reflect effects at the individual level.
-People with outcome may not be those with the determinant of interest. (Interest in individual-level effects but measuring with group-level data) |
|
What are the strengths of ecologic studies? (there are 4)
|
-They are inexpensive
-Good for early stages of knowledge -May allow for a wider range of exposures -Can help study ecologic relationships (e.g. effect of seat belt laws on motor vehicle fatalities) |
|
What measures of association are used for ecologic studies?
|
linear regression to estimate the slope of the regression line beta.
(How much Y changes for a one-unit change in X) |
|
What are the limitations of an ecologic study?
|
-Ecoligic bias (group data to study individual-level associations
-Little to no control of confounding |
|
What is the study base of a cohort study?
|
The population experience over time
|
|
What are the defining features of a cohort study?
|
A cohort study involves follow-up (documentation) of cohort members over time.
-It's a study of incidence (e.g. diagnoses) or change (e.g. progression or improvement of illness) |
|
What is estimated for a cohort study?
|
risk and rates
|
|
What are the key features of a randomized controlled trial?
|
-prognistic study about the future course of health or illness
-study of INCIDENCE (or change) -An experiment; the determinant/exposure is intentionally assigned for the goals of the study |
|
In an RCT, what forms the study population and study base?
|
Enrollment into the cohort.
Involves follow-up of cohort members. |
|
What is the source population in an RCT?
|
individuals with indication and without contraindication for intervention enrolled in the cohort
|
|
What is the study population in an RCT?
|
Formed by enrollment into the cohort
|
|
What is the study base in an RCT?
|
The follow-up experience of individuals in the cohort
|
|
What are two different types of RCTs?
|
Treatment allocated to INDIVIDUALS
Treatment allocated to ENTIRE COMMUNITIES (number of communities, not individuals, constitutes the study size |
|
What are the two ways to minimize systematic study error (bias) in a RCT?
|
-Use of a placebo (allow study to be "blinded", make groups as comparable as possible)
-Maintaining high compliance (run-in period, compensation, frequent contact) |
|
what is a run-in period for an RCT?
|
Certain number of weeks on treatment to find compliant subjects. Only use compliant subjects (randomize after suitable amount of time)
|
|
What measures of association are used for RCTs?
|
-Risk ratio:
(often presented as excess risk: (1-RR)*100% or risk difference -also can estimate rate ratio and rate difference |
|
What is intent to treat analysis?
|
Includes all participants based on their treatment assignment; non-compliers are considered having complied with assigned treatment
|
|
what are the advantages of intent-to-treat analysis?
|
Maintains randomization, groups are balanced
|
|
What are the disadvantages of intent-to-treat analysis?
|
May obscure real therapeutic effect among those who actually had the intervention
|
|
What is Efficacy analysis (aka as-treated analysis)?
|
Includes only participants who received the treatment as assigned (only compliers)
|
|
what are the advantages of efficacy analysis?
|
Tests effect of treatment in those who were actually treated
|
|
What are the disadvantages of efficacy analysis?
|
Lose randomization, so groups may not be balanced
|
|
What are the 2 main strengths of randomized controlled trials?
|
1. Can minimize bias
2. They are optimal to detect statistically small to moderate (but clinically worthwhile) effects |
|
What are the 3 main limitations of randomized controlled trials?
|
1. Acceptability (??)
2. Feasibility: time, cost 3. Equipoise: (humanitarian viewpoint) have to have enough belief to give treatment, enough doubt to withhold treatment |
|
When one initiates a cohort, what does it involve?
|
Follow-up of people, data collection, etc. for the occurrence of illness (or change)
|
|
For a cohort study, how is membership defined?
|
Membership into the cohort defined on the basis of some event
Person is forever after a cohort member |
|
What types of studies can be conducted in a cohort?
|
Cohort study designs
cross-sectional study designs case-control study designs |
|
What are the different types of cohort studies?
|
-non-experimental prognostic study
-etiologic study -descriptive study |
|
What is studied/estimated in a cohort study?
|
-Incidence (or change) studied
-Risks or rates estimated -involves follow-up of cohort members |
|
What is the key feature of a non-experimental cohort study as opposed to an RCT?
|
The determinant/exposure is NOT assigned by the investigator
|
|
How is the outcome assessed in a cohort study?
|
Many different ways (examples below):
-Medical or occupational records -Routine examinations at cohort visits (lab tests, physical measurements) -Questionnaires |
|
How is the determinant assessed in a cohort study?
|
Lots of different ways, just like outcome:
-Medical or occupational records -Routine examinations at cohort visits (lab tests, physical measurements) -Questionnaires Determinant may be fixed (blood type) or may change over time (smoking) |
|
What is the outcome of a cohort study?
|
Incidence (e.g. diagnosis)
OR change (e.g. progression or improvement of illness) |
|
What are the two ways a reference group can be chosen for a cohort study?
|
Internal or External reference group
|
|
What is an internal reference group for a cohort study?
|
Members of the same cohort WITHOUT the characteristic of interest
|
|
What is an external reference group for a cohort study?
|
Members of another cohort WITHOUT the characteristic of interest
OR subjects from the general population |
|
What are the strengths of using an internal reference group?
|
It's the best reference group when feasible
Comes closest to the "counterfactual ideal" because it is most comparable to the index (exposed) group |
|
Why would an external reference of another cohort group be used?
|
-Often used in studies of occupational exposures --> Another occupational cohort could share some behavioral risk factors
-A working reference cohort eliminates the Healthy Worker Effect |
|
What is one caution about using an external reference group of another cohort?
|
-Need to be careful that the reference cohort doesn't have its own characteristics that may affect illness
|
|
Why would an external reference group of the general population be used?
|
-It's accessible and cheaper
|
|
What are some cautions about using an external reference group of the general population?
|
-limited on outcomes with available information
-Overall population includes individuals with and without the determinant of interest -Healthy worker effect -Limited data on confounding variables |
|
What measures of association are used for a cohort study?
|
-rate ratio and rate difference
(Can validly estimate risk ratio or risk difference when most individuals are followed for the entire risk period) |
|
What is non-differential loss to follow-up and what does it affect?
|
Non-differential = not related to outcome or determinant
Affects study size |
|
What is differential loss to follow-up and what does it affect?
|
Differential = related to both outcome and determinant
Affects study validity (selection bias) |
|
How can we maintain follow-up (study retention)?
|
-Make it easy for the subject to participate and update you
-Incentives -Maintain contact with the subject (study visits, questionnaires, study updates, cards, keep notes on contact, update addresses, etc) |
|
What is a prospective cohort study?
|
A study that starts prior to any illness occurrence in the entire cohort
|
|
What is a retrospective cohort study?
|
A study that starts after all illness in the entire cohort has occurred.
|
|
What are the positives of a retrospective study?
|
-Time/costs
-Better for diseases with long illness induction |
|
What are the positives of a prospective study?
|
-Data quality on determinant
-Data quality on confounders |
|
What are the strengths of cohort studies?
|
-Efficient for rare exposures
-Collection of data in a cohort allows evaluation of determinant in relation to many illnesses -Temporality of determinant-illness usually clear -Determinant history documented before the illness occurs |
|
What are the limitations of cohort studies?
|
-Costly for rare outcomes or illnesses with long induction periods
-Expensive to follow individuals, collect data, conduct routine medical examinations and lab tests etc -can have loss to follow-up |
|
What are the key features of a case-control study?
|
-Includes cases in the study base and a SAMPLE of the population experience of the study base (controls)
-Can have a primary or secondary study base -Usually ETIOLOGIC study -Usually a study of incidence |
|
What is estimated in a case-control study?
|
Incidence odds ratio
|
|
Why conduct a case-control study?
|
-Efficient design to study an etiologic question, especially when:
1.Illness is rare 2. Exposure is difficult or expensive to obtain 3. Disease has a long time between exposure and disease 4. Difficult to follow and track individuals in a particular dynamic population |
|
In a cohort study, we have complete documentation of the experience of the entire study base (i.e. follow-up). What do we have in a case-control study?
|
Cases (individuals with the outcome) and a SAMPLE of the experience of the study base (controls)
We sample for the control |
|
What does a 2x2 table look like for a CASE-CONTROL study?
|
Columns = determinant (index on L, Reference on R)
Rows = Cases (top) and Controls (bottom) |
|
What is a primary study base?
|
-The source population that gave rise to the cases is KNOWN FIRST and THEN cases are identified
(Cases come from a population that is explicitly and directly defined) The study base is defined directly (and not indirectly through case identification) |
|
Where do the controls come from in a primary study base?
|
Examples:
-Cohort (in a case-control study conducted within a cohort/fixed population) -National identity registries -Town lists -Random digit-dialed telephone numbers |
|
How is a secondary study base defined?
|
-Cases come from a population that cannot be explicitly defined
(We identify cases first then try to identify the source population that gave rise to the cases) -The study base is INDIRECTLY defined; it is defined secondarily to the identification of the cases |
|
When do conditions arise for a secondary study base?
|
Examples:
-When cases are identified from hospitals whose catchment areas cannot be specified -From other medical practices serving patients with various places of residence, health insurance, referral patterns -Disease registry that does not cover a defined population |
|
Where do the controls come from ion a secondary study base?
|
Examples:
-Family, friends -Residents of the same neighborhood -Patients admitted to the hospital with other illness unrelated to the determinant of interest -Patients attending the clinic for other illness unrelated to determinant of interest |
|
What are the potential strengths of using a primary study base?
|
-It's clear that controls represent the study base (population experience) that produced the cases (same source population)
-In a case-control study conducted in a cohort: data and specimens often already drawn and stored (quick and inexpensive to conduct as long as specimen quality not an issue) |
|
What are some limitations to using a primary study base in a case-control study?
|
-Expensive to obtain controls: random digit dialing
-Controls tend to be healthy so: a. They can be less interested in participating and lower response rate (selection bias) b. Accuracy of determinant information may be different than the cases (misclassification bias) |
|
What are the potential strengths to using a secondary study base in a case-control study?
|
-Less expensive than a primary base (no random-digit dialing)
-Controls often have another illness (not healthy) and therefore more interested in participating and higher response rate -Accuracy of information on determinant from controls may be similar to cases |
|
What are the potential limitations of using secondary study bases?
|
-Difficult to know if controls represent the study base that produce the cases (come from the same source population)
-May be difficult to select control diseases unrelated to determinant of interest |
|
Odds Ratio
|
(Index cases / index controls) / (reference cases / reference controls) = odds ratio
|
|
Odds
|
index cases / index controls = odds of illness among index (exposed)
reference cases / reference controls = odds of illness among referent (unexposed) |
|
What is the purpose of controls in a case-control study?
|
-To provide information on the distribution of the determinant in the population experience that produced the cases. (proportion of people with and without determinant in study base)
-Provide information for denominators which is necessary for calculation of odds ratio |
|
What is the purpose of a placebo group in an RCT?
|
The placebo group provides an estimate of the frequency of the outcome in the reference group.
|
|
What are the two main criteria for valid control selection in a case-control study?
|
-Controls come from the source population that produced the cases; if they develop illness, would be a case in the study
- Controls are selected INDEPENDENT OF EXPOSURE |
|
AF for case control study?
PAF? |
AF = (OR-1)/OR
PAF = Pe*(OR-1)/OR Pe = # of exposed cases/total number of cases |
|
What are the potential strengths of case control studies?
|
-Efficient for rare diseases or for diseases with a long induction period
-Efficient when exposure information is expensive -Can evaluate multiple exposures for a single illness |
|
What are the potential limitations of case-control studies?
|
-Inefficient for rare exposures (??)
-Can be susceptible to selection bias (controls not representative of study base) -Can be susceptible to information bias (if determinant information inaccurately obtained) -Can't calculate cumulative incidence and incidence rate |