EPI Review Flashcards
non differential misclassification biases the data toward or away from the null?
Non-differential misclassification biases the estimate toward the null; the “true” RR isfurther from the null than the reported RR.
What happens with a basic reproductive rate <1
Disease will disappear
What happens with a basic reproductive rate =1
Disease becomes endemic
What happens with a basic reproductive rate >1
Epidemic
How to calculate the basic reproductive rate (R0)?
What are the attributes of a surveillance system?
Sensitivity
* Timeliness
* Representativeness
* Positive Predictive Value
* Simplicity
* Acceptability
* Flexibility
What are the types of surveillance?
Active Surveillance
* Passive Surveillance
* Syndromic Surveillance
* Sentinel Surveillance
* Population-based Surveys
What is specificity?
ability of test to correctly identify all
those without disease
What is sensitivity?
ability of test to correctly identify all
those with disease
What is the sensitivity formula?
True positives/all diseased
What is the specificity formula?
True negatives/all non-diseased
What happens to a test if the cut-off is lowered?
More sensitive, less specific
What happens with multiple parallel screenings?
– Usually will increase sensitivity.
– Usually will decrease specificity.
What happens with multiple screenings in series?
– Some loss in sensitivity.
– Will tend to increase specificity.
Hot to calculate net sensitivity?
How to calculate net specificity?
What happens to net sensitivity in two stage screening?
Loss
What happens to net specificity in two stage screening?
Gain
What is PPV?
probability
person who tests positive actually has disease
What is NPV
probability
person who tests negative does not have disease
What are the measures of association in RCT?
Risk ratio, risk difference, rate ratio, numbers needed to treat
What does risk and rate difference measure?
Risk or rate difference measures clinical and public
health importance of the causal relationship
What does risk and rate ratios measure?
Risk or rate ratio assesses strength of the
association
What is selection bias?
What is information bias?
Systematic difference in the way exposure or
outcome is measured between comparison groups
in what type of study is selection bias more common?
case control
what is the direction of selection bias
can be toward or away from the null
What are sources of information bias?
Surrogate interviews
* Non-response of particular questions that
leads to incorrect classification (vs not being
completely excluded)
* Invalid instruments (do not measure what you
want to measure)
* Incomplete or erroneous data sources that
leads you to classify incorrectly
What are the types of differential misclassification?
If distribution of errors for one variable depend
on the other (e.g. detection bias, recall bias)
– Can occur both in measurement and in analysis
(e.g., when collapsing continuous variables)
What is non differential misclassification?
If distribution of errors for one variable do NOT
depend on the other
What is the bias direction in Misclassification of exposure that is
DEPENDENT on outcome?
bias either direction
What is the direction of bias in non differential misclassification of exposure?
Misclassification of exposure that is
INDEPENDENT of outcome (i.e. same for
D+ and D-), generally biases toward the
null
What is the direction of Differential misclassification of outcome
bias either direction
Non-differential misclassification of
outcome - direction of bias?
generally biases toward the
null
What happens with Confounders that OVERESTIMATE
shift the observed
measure of association AWAY from the null
What happens with Confounders that UNDERESTIMATE
shift measure of
association TOWARDS the null
What are the confounder criteria?
Cause of outcome
– Cause of exposure
– Not on E-D causal pathway (i.e. not a mediator)
how to control for confounding in study design?
Randomize the exposure
* Match exposed and unexposed on variable of interest
* Restriction
how to control for confounding in the analysis stage?
Stratified analysis
* Standardization
* Use multivariable adjustment techniques (e.g.,
regression)
* Advanced methods
What are the steps to evaluate confounding?
STEP 1: Calculate crude measure of association
* STEP 2: Calculate strata specific measures of
association
* STEP 3: Compare crude vs stratum specific estimates
How much of a difference between crude and
adjusted beta is enough to conclude that there is
evidence of confounding in the data?
10-20%
When to use a case cross-over design?
Exposure is dynamic
2. Exposure has a very short induction time
example: car crash, MI
There must be within person variation in the exposure of
interest in order to test the association in a case cross-
over design
What is the measure of association for case cross over design?
Compare exposure distributions for the case and control periods
* Use a matched pair case control analysis
* Instead of concordant and discordant pairs of subjects, each patient contributed a pair of intervals, i.e., a hazard and a control period, which were either concordant or discordant for exposure
What is incidence density sampling?
Collect information on each case
* Collect control at the same time each case is
observed; collect control from the underlying source
population giving rise to the cases
* Controls can become cases!
What are the measures of association in nested case control studies?
Odds ratio from nested case-control study
approximates the rate ratio
Do you need rare disease assumption for nested case control studies?
No
When is case based sampling used?
Case cohort studies
What is case based sampling?
All controls are selected at the beginning of the
study period
* Selected so that the ratio of exposed controls to the
number of exposed cohort members equals the ratio
of unexposed controls to the number of unexposed
cohort members (random selection)
* Used to study multiple disease outcomes with the
same control group
What are measures of association for case cohort studies?
Odds ratio from case-cohort study
approximates the risk ratio
Do not need rare disease assumption
What measures of association can be used for cross sectional studies?
can estimate prevalence risk ratios, risk differences, and odds ratios
What measures of association can be used in prospective cohort studies?
can estimate risk ratios, risk differences, rate ratios, and odds ratios
Wha measures of association can be used in RCTs?
risk ratios, risk differences, rate ratios, and odds ratios
What are the measures of association for case control studies?
odds ratios
What are the measures of association for ecologic studies?
. Prevalence risk ratios, risk differences, and odds ratios (at the unit of analysis
What are the measures of association for retrospective cohort studies?
Risk ratios, risk differences, rate ratios, odds ratios
What measures of association in Nested case-control with incident density sampling
odds ratio that approximates the rate ratio
What is an example of EMM synergism?
moking combined with asbestos exposure increases risk
of lung cancer (synergy)
▪ Effect of exposure and modifier create more risk of outcome than
expected based on joint effect of both
What is an example of EMM antagonism?
NSAID (ibuprofen) used with anti-diuretic reduces the
effectiveness of the anti-diuretic (antagonism)
What is the difference between EMM and confounding?
Confounding =
Correlated causes
acting through
independent
causal pathways
Effect measure
modification =
several causes
acting on same
causal pathway
What are the steps to assess EMM?
- Examine crude relationship
- Stratify based on third variable
- Compare stratum specific effect measures
If there are differences in the measures of effect
across strata, there is effect measure modification
What can you hope to find when there is EMM?
If there is effect measure modification: stratum specific measures are different from the crude and
different from each other.
– We should report the stratum specific measure (RR, RD, IRR,
OR)
What can you hope to find if there is confounding?
stratum specific measures are similar to each other but different
from crude (can use 10-20%)
– Report Adjusted measure (RR, RD, IRR, OR)
What is the difference between crude and strata specific measures for EMM and confounders?
EMM - stratum specific measures are different from the crude and
different from each other.
Confounding - stratum specific measures are similar to each other but different
from crude (can use 10-20%
How to assess multiplicative EMM?
Multiplicative Effect Modification: Evaluates differences in relative risk or rate
How do you assess additive EMM?
Additive Effect Modification: Evaluates differences in absolute risk or rate.
you have a logistic model: logit(p(Y=1)) = α + β1X1 + β1X2 + β3X1*X2, you have evidence ofadditive interaction if:
eβ1 + β2 + β3 ≠ eβ1 + e β2 - 1
