Epidemiology Flashcards
(26 cards)
Define Selection Bias
The most worried about
Deals with how the researcher selects or acquires study subjects
Creates a systematic difference in the composition between groups
Define Self-Selection/Participant/Responder bias
Those that wish to participate/volunteer may be different in some way to those that don’t volunteer or self-select to participate
i.e. clipboard at the mall example
Define Recall/Reporting bias
Something done by the study subject
A differential level of accuracy/detail in provided info. b/w study groups
i.e. exposed or diseased subjects may have greater sensitivity for recalling their history or amplify their responses (exaggerate)
“Hawthorne Effect” - Individuals can report their “effects” of exposure, disease symptoms or treatment differently bc they are apart of the study; want to please the experimenter
Define Compliance bias
Groups being interventionally studied have different compliances
Define cause (association & causality)
A precursor event/condition/characteristic required for the occurrence of the disease
Associations are relationships b/w an exposure and an outcome = artifactual, non-causal, causal
What are the 5 Hill’s Criteria for causality?
- Strength
- Consistency
- Temporality
- Biologic Gradient
- Plausibility
The greater criteria that are met, the more likely that an association may be causal
Delineate between cohort, case-control, and cross-sectional observational studies
Case-control = ALWAYS RETROSPECTIVE; based on the presence of disease (case) and no disease (control)
Cohort = Group allocation based on exposure; BOTH prospective and retrospective (unable to randomize, exposure is rare, etc.)
Cross-sectional = NATIONAL; snapshot in time; quick and easy
What is the primary difference between simple and factorial study designs?
Simple contains only ONE act of randomization
In factorial you are testing multiple hypothesis at same time
What is the primary difference between parallel and cross-over study designs?
Parallel contains no switching of groups after the initial randomization
Cross-over contains switching between and within groups (such as wash-outs) after randomization
List the 3 forms of randomization
- Simple - ensures an equal probability during allocation within one of the study groups
- Blocked - ensures an equal # within each interventional group; researcher doesn’t know
- Stratified - ensures a balance within the known confounding variables; i.e. want population characteristics to be equal (race, gender, etc.)
What are the 3 ways in which researchers can handle lost to followups or drop outs?
- Include them anyway = INTENT TO TREAT
- Ignore them = PER-PROTOCOL/EFFICACY ANALYSIS; pre-defined compliance
- AS-TREATED; ignores group assignments
Benefits of intent to treat?
Most conservative decision; convert all subsequent yet missed assessments for a subject to a null-effect
Preserves randomization process, preserves baseline characteristics, maintains statistical power/original sample size
Benefits/Disadvantages of treating subjects as per-protocol in studies?
Biases estimates of effect
Reduces generalizability
What is the purpose of case-control studies and when might you use them?
(Observational)
Always RETROSPECTIVE
Dependent on disease status
Used when disease is rare, testing for multiple exposures compared to one outcome, time and cost effective, ethical issues and cannot use interventional study (randomize)
What is the purpose of cohort studies and when might you use them?
(Observational)
Studies done based on exposure and commonalities
Used for RARE EXPOSURES, can determine the risk ratio as a measure of association, when you cannot randomize
What is the purpose of cross-sectional studies and when might you use them?
(Observational)
Large population studies = SNAPSHOT in time; collects data simultaneously and seeks associations and generates and tests hypothesis
NATIONAL
Define sensitivity; what is the calculation?
How well a test can detect disease presence when the disease is actually present; no prediction
A highly sensitive test can have a low false negative rate
Sensitivity = TP / (TP + FN;all disease) x 100% S = A / A+C
Define specificity; what is the calculation?
How well a test can detect the ABSENCE of disease when the disease is actually absent; no prediction
A highly specific test has a low false positive rate
Specificity = TN / (TN + FP) x 100% Sp = B / B+D
Define positive predictive value (PPV); what is the calculation?
How accurately a positive test PREDICTS the presence of disease
PPV = TP / (TP + FP) x 100% ppv = TP / (all positive tests) x 100%
Define negative predictive value (NPV); what is the calculation?
How accurately a negative test PREDICTS the absence of disease
NPV = TN / (TN + FN) x 100% npv = TN / (all negative tests) x 100%
Define diagnostic accuracy; what is the calculation?
Proportion of time that a patient is correctly identified as either having a disease or not having a disease with a positive or negative test.
DA = (TP + TN) / (TP+TN+FP+FN) x 100%
DA = (TP + TN) / (all patients) x 100%
= (A + D) / (A+B+C+D)
What type of studies are CONSORT checklist is used for?
Interventional studies
What type of studies are strOBe checklists used for?
Observational studies
Nominal data characteristics
Dichotomous/Binary
No magnitude; No consistency of scale; No order
i.e. gender, hair color, eye color, chocolate preference, Low BP/High BP
