| Terms |
Definitions |
|
odds ratio
|
- prob will occur/prob will not occur
- (a/b)/(c/d) = (ad)/(bc)
|
|
why do a cross-sectional study?
|
-for PH planning
-etiologic research
|
|
3 factors to consider in measurement of disease frequency
|
1. # of people affected
2. # of people in population from which cases arise
3. length of time followed
|
|
problem of loss to follow-up
|
-reduces power of study to detect associations
|
|
ecological study
|
-examines rates of disease in relation to a factor described on a population level
-identifies groups by place/time
|
|
Risk difference
|
RD=Re-Ru
e = exposed
u = unexposed
|
|
ways to minimize recall bias
|
-select a diseased control group
-comprehensive questions for exposure ascertainment
-self-administered questionnaire
-rely on measurements and records
|
|
observation bias
|
a flaw in measuring exposure or outcome data that results in different accuracy between comparison groups
|
|
Efficacy analysis
|
determines treatment effects under ideal conditions (i.e. compliance)
|
|
why use a case-control study?
|
-little is known about disease etiology
-less time, $ than a cohort
-rare disease
-diseases with long induction and latent periods
-when exposure data is difficult to obtain
|
|
Attributable proportion among the exposed
|
APe=[(Re-Ru)/Re]x100
-etiologic fraction
-proportion of disease among exposed that would be eliminated if exposure were eliminated
|
|
Observational study
|
-studies causes, preventions, treatments
-passive
-cohort, case-control,
|
|
Organization in a 2 x 2 table
|
Disease on x-axis
Exposure on y-axis
|
|
Possible timing of events in a cohort study
|
-prospective
-retrospective
-ambidirectional
|
|
Precision
|
lack of random error, which leads to false association
|
|
Cohort study
|
-examines multiple health effects
-good for rare exposures
-followed according to exposure levels and followed for disease occurrence
-observational
|
|
Dynamic population
|
Membership is based on a condition and is transitory
|
|
survivor sampling
|
select controls from non-cases at the end of the case diagnosis and accrual period
|
|
purpose of case-control study
|
investigate risk of disease in relationship to a wide variety of exposures
|
|
advantage to using hospital controls
|
-easy to identify
-good participation rates
-less expensive to identify
-come from the same population
-recall is similar to cases
|
|
Bias
|
error committed by investigator in design or conduct that leads to a false association between exposure and disease
|
|
purpose of run-in period
|
to ascertain which potential participants are able to comply with the study regimen
|
|
avoiding interview bias
|
-mask interviewers to subjects' disease or exposure status
-closed-ended questions
|
|
problem with using prevalence in case-control studies
|
impossible to tell if exposure is related to the inception of the disease, its duration, or a combination
|
|
cohort study occurrence outcome measure
|
-cumulative incidence
-incidence rate
|
|
cons of using dead controls
|
-not representative of living population
-more likely to have used tobacco, alcohol, drugs
|
|
cons of retrospective cohort study
|
-minimal info is available on exposure, outcome, confounders
-difficult to establish temporal relationship
|
|
tips for writing understandable questions
|
-avoid ambiguous wording
-avoid 2 questions in one
-avoid jargon
-avoid impossible questions
-avoid questions that are not self-explanatory
-avoid questions w/ too many ideas
|
|
Experimental study
|
-studies prevention and treatment
-active manipulation
|
|
Masking
|
unawareness of treatment assignment of patient
|
|
Population risk difference
|
PRD=Rt-Ru
PRD=RDxPe
Pe = proportion exposed
-excess # of cases in total pop
|
|
How may biased results come about as an outcome from "special" controls (e.g. siblings)?
|
if study hypothesis involves a shared activity
|
|
cons of case-control study
|
-greater chance of bias
-not good with weak associations
-difficult to establish temporal relationship
|
|
recall bias
|
-differential level of accuracy in the info provided by compared groups
|
|
pros of ecological study
|
-quick
-inexpensive
-simple, easy to understand
-wider-range of exposure
|
|
misclassification
|
-"measurement error"
-occurs when broad definitions are used
-can reduce by using multiple measurements, validation, or using sensitive and specific criteria
|
|
purpose of placebos
|
to match as closely as possible the experience of the comparison group with that of the treatment group
|
|
Intent-to-treat
|
-treatment assignment analysis
-all individuals in treatment group are analyzed
|
|
Incidence Rate
|
# new cases / person-time in candidate population
-measure of transition between health/disease of dynamic populations
-research on causes, prevention, treatments
|
|
Case-crossover study
|
-used when risk of outcome has a short window
-cases serve as own controls
|
|
pros and cons of using Health Records as sources of info
|
pros: low expense, high accuracy
cons: other key characteristics missing
|
|
cons of ecological study
|
-ecological fallacy or ecological bias: not generalizable to individuals
-inability to detect subtle or complicated relationships
|
|
why use a cohort study?
|
-little is known about health consequences of exposure
-rare exposure
|
|
When evaluating a study for the presence of bias, investigators must:
|
-identify its source
-estimate its magnitude
-assess its direction
|
|
Referent group
|
unexposed, or comparison group
|
|
Case-control study
|
-good for rare diseases
-examines multiple exposures
-exposure histories compared
-observational
|
|
cons of cross-sectional study
|
-no temporal inference when exposure is changeable
-bias: identifies high proportion of cases w/ long duration
-healthy-worker effect
|
|
why use an observational study?
|
study effects of a wider range of exposure
|
|
cross-sectional study
|
-measures exposure prevalence in relation to disease prevalence
|
|
Fixed population
|
Membership is based on an event and is permanent
|
|
a study is considered valid only when what 3 things are eliminated?
|
-bias
-confounding
-random error
|
|
Relative Risk
|
RR=Re/Ru
|
|
"would" criterion
|
a member of the control group who gets the disease under study would end up as a case
|
|
selection bias
|
an error due to systematic differences in characteristics between those who are selected for study and those who are not
|
|
sources for comparison group
|
-internal
-general pop
-comparison cohort
|
|
Prevented factor
|
PF=[(Ru-Re)/Ru]x100
-used when exposure is thought to be protective
|
|
Absolute measures of comparison
|
Risk difference, incident rate difference, etc.
|
|
best way to avoid self-selection bias
|
obtain high participation rates among both groups
|
|
risk set sampling
|
-controls selected from population at risk as cases are being diagnosed
-may include future cases
|
|
issues in experimental study design
|
-noncompliance
-must maintain followup rates
-high costs
-reluctance to participate
-ethical issues (equipoise)
|
|
Cross-sectional study
|
-examines relationship between exposure and disease prevalence in a population at a single point in time
-observational
|
|
pros of cross-sectional study
|
-highly generalizable
-short time
-low cost
|
|
Validity
|
lack of bias and confounding
|
|
Key features of observation bias
|
-occurs after subjects have entered study
-pertains to data collection
-it often results in incorrect classification
|
|
Randomization
|
an act of assigning or ordering that is the result of a random process
|
|
Prevalence
|
# of existing cases / # in total population
-point or period
-resource planning
- P=IR*D
(D = duration)
|
|
why use an experimental study?
|
-high degree of validity from randomization (control for confounders)
-prevention or treatment with small effect
|
|
Cohort
|
a group of people with a common characteristic or experience
|
|
case-based or case-cohort sampling
|
-select controls from population at risk at the beginning of case diagnosis and accrual period
-may include future cases
|
|
Excess Relative Risk
|
ERR=(RR-1)x100
|
|
Incidence
|
-occurence of new cases over a specified period of time
-measured in a candidate population
|
|
Direct method of standardization requires...
|
1) Age-specific rates in each group
2) Age structure of a standard population
|
|
Closed cohort
|
fixed cohort with no losses to followup
|
|
Ecological study
|
-examines relationship between exposure and disease with population-level data
|
|
Cumulative Incidence
|
# new cases/ # in candidate population over a specified period of time
-used in fixed populations, few losses to followup
-research causes, prevention, treatment of disease
- CI=IR*t
|
|
Confounding
|
reflects on the fact that research is on free living humans with unevenly distributed characteristics
|
|
purpose of case-control study
|
provide information on exposure distribution in population
|
