Basic principles HS Flashcards
(487 cards)
Diagnosis question =
Cross-sectional analytic study = observational study that analyses data from a
population at a specific period of time
Aetiology question =
- Cohort study = longitudinal study that follow a population, often one that has a
particular exposure – e.g. smoking - Case-control study = population study in which two existing groups differing in
outcomes are identified & compared based on basis of some supposed causal
attribution
Prognosis question
cohort
Tx question
- RCT / Systematic Review of RCT
- Note: can test for superiority between two variables, or for equivalence
(i.e. equally as effective) - Cluster trials is where randomize clusters of patients, e.g. GP surgeries
evaluation questiuons
Systemic review/ MA
cohort +
- Best information about causation
- Able to examine a range of outcomes
- Good for rare exposure
Disadvantages of cohort studies
- Long follow up = expensive & time-consuming
- Bad for rare outcomes
- Bad for long latency periods
- Can have different follow-up for exposed / non-exposed
- Confounders not recognized
CCS +
Simple / easy to conduct = don’t require long follow ups
* Best for rare outcomes
* Good for long latent periods
disadvantages of a CCS
Inferior to a cohort study
* Bad for rare exposure
* Controls may not represent where sample is from = often requires
several controls per case
* Cases don’t represent full disease spectrum = cured / died
* Recall bias
* Confounders not recognized
BIAS
Systematic introduction of error into a study that can distort the results in a non-random way
selectio bias
Error in assigning individuals to groups, leading to differences in
group’s qualities that may influence the outcomes
* To make inferences from results, we require sample to be
representative of population = requires random sampling (or if
smaller group some stratified randomization)
ways in which selection bias can arise
- Sampling selected subjects not representative of population
2. Volunteer volunteer subjects not representative of population
3. Non-responder responders are not representative of the
populat
recall bias =
- Difference in accuracy of recollection of study participa
- could be due to time, greater recall in those w the disease (issue in CCS)
Publication bias
- Failure to publish / include certain studies because they have
negative results = important in systematic revie
hawthorne effect
- Group changing its behavior due to knowledge it is being studi
procedure bias
- Subjects in different groups receive different care, other than just the intervention
info bias
- Information about participants is incorr
methods of randomisation - simple
computer random number generator (may have chance bias)
stratified randomisation
– to make sure you have equal numbers of e.g. different severities of illness in each
minimisation bias
computer algorithm forms groups to ensure comparability (smaller studies)
key principle of randomisation
t must ensure unpredictable assignment to groups
concealed randomisation tends to
o yield less impressive results – because without this you over-estimate the effect by
selection bias
regression to the mean =
People often get better or worse regardless of intervention = not intervention causing change
regression to mean example
If you are measuring something that randomly varies (e.g. BP which is fluctuant) and you’re always
looking at the extremes, a significant proportion of the people who are in the extremes are there by
chance
B. Therefore, when you re-measure the group, on average they will “regress to the mean”
C. In the example of BP = if you measure a group of people with high blood pressure, when you next
measure them, they will improve (decrease BP) on avera