Week 2 & 3 Flashcards
Epidemiological studies look at…
how often various health conditions occur in specific populations - quantitative method
Etiology = ______
causation
Efficacy = ….?
the best possible outcome
Variable = …?
measurement that differs across individuals (weight, height, age)
What is an independent variable?
- Treatment or the intervention - often to bring improvement
- Presumed cause, one that you want to test
What is a dependent variable?
- Outcome (of treatment); often what clinician wants to improve
- Presumed to be affected by the independent variable
Intervention = ….?
the researchers do something to bring about change
Which are easier for researchers to do; experimental studies or non-experimental studies?
Non-experimental studies (only weakly support causal inferences)
How many groups do experimental designs have?
At least two (treatment group, non-treatment/placebo etc group)
The treatment effect = ___ minus _____
after minus before (result minus previous status of condition)
A true experimental design has…
at least two groups with comparisons on outcomes, and randomly allocated subjects to treatment and control groups.
Observation = ______
measurement
In quantitative research, observational studies can include (3)…?
- Descriptive studies (such as survey designs)
- Correlational studies
- Designs used in epidemiological research about hazards to health
List of epidemiological observational designs (3)?
- Cohort study
- Case-control study
- Screening studies to identify possible health cases for attention
Are single group case series experimental or non experimental?
Non-experimental
“any effect that prevents the study conclusions from running true”, “any systematic error in collecting or interpreting data” - are known as…
Bias’
Is bias one directional?
No, can operate in each direction (overestimating or underestimating effect)
NHMRC levels of evidence hierarchy rank studies from highest quality (______ number) to lowest quality (_____ number)
lowest, highest
Level ____ is the best for single studies of interventions (treatments), diagnostic accuracy, etiology, prognosis or screening
two
Systematic reviews are level ____ EXCEPT when the studies are…?
level 1, except when the studies are a lower quality than level II (e.g. level III or IV)
The higher the number is (or lower the level of quality is) the greater risk of ____
bias
What are some High-level evidence (low bias risk) study examples ? (4)
- Systematic review superior evidence
- Intervention studies that are experimental
- Diagnostic studies
- Studies that are prospective and cohort designs rather than retrospective or case-control
What are some low-level evidence (high bias risk) study examples ? (5)
- Lack of randomisation for intervention study
- Observational designs (hard to know or control the influence of other factors linked with events of interest)
- Retrospective - looking backwards in time (outcome already known in study rather than future, risk of sampling bias)
- Controls not concurrent (different timing could bias results)
- No controls at all (no comparison conditions)
Level II studies are what kind of study?
Randomised controlled trials; treatment and concurrent control groups with subjects randomly assigned
Level III-1 studies are what kind of study?
Controlled trials without truly randomised allocation - treatment and control groups but using an approx random method
Level III-2 studies are what kind of study?
Comparative study with concurrent control group - control measured concurrently (same time) as intervention but without random allocation to groups
Level III-3 studies are what kind of study?
Comparative study without concurrent control group - Intervention and control conditions could have occurred at different times
Level IV studies have…?
Case series with no control group; treatment group only
If a study is high NHMRC level, does it mean it’s an all around well-conducted study?
Not necessarily; evidence level is study design only, not actual procedures for the study
What are other sources of bias in studies that are not counted in the NHMRC level of quality? (4)
- sampling, including attribution
- measurement errors
- How intervention are conducted
- How data analysed and results reported
Does the NHMRC levels of evidence apply to all research?
No, only quantitative studies
Is it possible for a poorly conducted RCT (Level II) to be more biased than a well-conducted comparative study (Level III)?
Yes
Why reporting quality is important?
Research reports may be the only way to know how research is done
- We rely on research reports to tell us all we need to know
- Method sections describe design and procedures for that study
As humans do we tend to not assume causality or over assume causality?
Over assume without really studying it
It is more important to know if something causes a reaction or to know if something DOESN’T cause a reaction?
Both equally important
Is the phrase “after the event, therefore, because of the event” true?
It can be the case, but the phrase itself is false. (After does not = due to)
inference =
a decision or conclusion about truth based on evidence
Causal inference =
inference about a causal relationship (e.g. treatment causal patients condition to improve)
Relations between ideas…
- Can be true or false
- logical only
- A triangle can NEVER have four sides
- Therefore, no reason to research if it can
- not empirical
Matter of fact…
- is empirical
- known through data and experience
- about cause and effect, e.g. a virus causes the common cold
Three rules of inferring cause and effect (Hume):
- Contiguity in time and space (same time same place)
- Cause precedes effects (Cause first then effect)
- Constant conjunction = reliability of cause and effect repeating
Can we prove cause and effect (hume)?
Can’t be proved because observation alone gives us no way to show that constant conjunction will always happen
Is cause and effect an empirical or logical process?
Empirical
What is a problem with Hume’s theory?
We can often infer cause and effect by single experience (touching fire burns you, no need to repeat)
Explain Mill’s Method of agreement
That the cause is EFFICIENT for the effect (e.g. when it rains, skies are always cloudy–> clouds cause rain)
Explain Mill’s Method of difference
That cause is NECESSARY for effect (e.g. rain does not occur without clouds; rain occur only with clouds // patient does not improve without treatment)
The method of _______ states that the treatment group tests whether patients improve when they have treatment
agreement
The method of _____ states that the control group tests what happens when patients don’t receive treatment
Difference
Moderator relationships occur when strength of relationship between two events ….?
Depend on another event (e.g. winning more sports medals increases the amount of sponsorship dollars an athlete receive, especially when the medals are gold [[colour of medal magnifies (moderates) relationship between the number of medals won and sponsorship dollars received]]
What are the Bradford Hill criteria for establishing causation from evidence? (5)
- Time - sequencing, order of events (Cause always first)
- Strength of association, correlation
- Dose-response for clinical trials (Size of dose of treatment matched with size of improvement)
- Replication of findings
- Plausibility
Aetiology = ….?
Cause, including causes of disease and death