Research Methods Flashcards
What is a case report?
-Detailed description of an individual patient with unusual or rare condition/case
-leads to speculation about possible causes
ex thalidomide study, FAS (fetal alcohol syndrome), ether
What is a Case series?
Extension of case reports: cluster of individuals with same condition
ex. toxic shock syndrome from rely brand tampons (staph aureus)
What are ecological or cross sectional studies?
- Descriptive or analytical
- Descriptive if: no hypothesis, no specific exposure/outcome, no valid comparison groups
- Population level measures (unit of analysis)
3 common types of analytical studies?
1-ecological or cross-sectional studies
2-case control studies
3-cohort studies
What is a descriptive study
- Characterize outcomes (morbidity/mortality) by place-person-time, and have no a prior hypothesis
- 2 types (case reports & case series)
Type 2 error = ?
= Beta, i.e. False negative rate
=When you do not reject the null hypotheis, when you should have (there is in fact a difference)
Type 1 error = ?
Alpha, false positive rate
Shows a difference, when in fact there isn’t one
Power = ?
Power = 1 - beta
Rate of telling the true difference
What is an ecological fallacy?
- Making an association between 2 populations, when in fact at the individual level, there is no causation.
- Ex assuming causation in a popn that has higher rates of obesity and diabetes than another, but may miss causation at an individual level
What is a cross sectional study (different than ecological)?
- Assess exposure and outcome at the same time (in one time point)
- Example are surveys
Limitation of a cross-sectional study?
Establishing temporality
Retrospective case-control study principles ? (2 ideas)
- start by classifying subjects based on outcome status
- look back in time for exposure status
- uses ODDs RATIO (OR)
Can you measure relative risk in case control studies?
No, because you dont know the true population of those who were exposed.
Formula for OR (odds ratio)
OR = (a/b) divided by (c/d)
What are case control studies good for ?
- RARE conditions
- Usually need a larger population to capture a condition if its RARE (in the case of a prospective cohort).
What type of bias affects case control studies in particular?
- Recall bias
- Patients are more likely to recall an exposure (when there may not have been one), if they are diseased
Prospective cohort study?
- Classify based exposure status (exp vs non exposed)
- Follow forward in time (looking for an outcome)
Advantage of a prospective cohort study? Disadvantage?
Example being Framingham study
adv - Establish temporality
disadvantage - Very costly
What does Relative Risk > 1 imply?
Implies a hazard
RR
Implies protective factor
What is RETROSPECTIVE COHORT study?
- historical cohort, where data is already collected
- existing data sources (eg ICES)
What is ICES data?
- Institute for Clinical Evaluative Sciences
- Health admin data, collected for purposes besides just health
Two possible data sources for ICES?
- Acute care (ICU, ER)
- Outpatient sources (rehab, clinics, etc)
What is a nested case control study?
- is a Hybrid study
- Case control study inside a Cohort study
Panel study?
-
Experimental study?
- Intervention in control of investigator
- RCTs and community trials
3 basic types of RCTS?
- preventive trials (primary intervention)
- intervention trials (secondary intervention)
- therapeutic trials (tertiary, cure vs improve prognosis)
Relative Risk formula?
RR = (a/a+c) / (b/b+d)
Odds ratio formula?
OR = ad/bc
Forrest plot used in what kind of study?
Size of dot refers to what?
- Meta analysis, or systematic review
- Sample size
Publication bias?
-People Publishing studies that only show positive results
Qualitative study design characteristics?
- hypothesis generating, not testing
- develop concepts that aid in understanding natural phenomenon through description
Common tools of qualitative designs?
- Interviews (semi structured, structured, unstructured)
- gets rich info
- build rapport
How is sampling done in qualitative studies?
Small (les than 50) Not probabilistic (too expensive to do so), nor representative
How do you analyze qualitative data?
- be systematic
- theoretical constructs outlined
- detailed methods
What is triangulation?
- Looking at multiple sources/views for the same data
- try to come up with one single unifying theory
What are Bradford Hill’s Criteria for Causation?
- Strength of association (measured by RR, OR)
- Correlation coefficient (dose response, does Y increase with X)
- Consistency
- Temporality
Other Hill criteria (5)?
- Specificity (effect has only one or a few causes)
- Biological plausibility
- Coherence (with other knowledge, no competing theories)
- Experimental evidence -analogy (similar phenomenon)
What is confounding?
Confounder related to exposure and outcome, but not directly in a causal pathway
Famous example of a confounder?
- HRT (estrogen, progesterone)
- Early studies suggested that people on HRT had lower CVS outcomes
- Later studies showed the opposite, initial studies were in HCW
Example of confounding factors in link between obesity and heart disease?
SES, inactivity, built environment
they are outside of causal pathway
What is a mediator (as opposed to confounder)? Use example of obesity and heart disease
- Clogging of arteries
- would be in causal pathway (i.e. mediator)
Moderating factors?
- Similar to mediating, is in causal pathway
- interacts, potentiates, effect modifying
- eg salt intake, fat intake, inactivity
Effect modification?
Multiplicative effect, not additive
in causal pathway
When can you control for confounding?
Before the study starts (study phase), and
After the study starts (analysis phase)
3 ways in the study phase, to control for confounding?
- randomization
- restriction
- matching
Controlling for confounding in the analysis phase?
Multivariable analysis (multiple confounders)
Linear regression
Logistic regression