Lecture 5 - RCTs Flashcards
what are non-randomized experiments
- assignment not random
- when randomization not possible
- or when RCT incorrectly performed
keys to RCT
- random assignement
- must be uncertainly about which treatment is better
- results less generalizable since on one therapy at a time
basic steps
- create cohort - meet inclusion
- may observe pts. for period of time to see if stable
- assign randomly to groups
- followed and assessed
strenghts
- minimal diffs between groups
- removes bias with blinding
- can calculate incidence rates
- gold standard
weaknesses
- only ethical if not sure of treatment
- high cost
- poor external validity
- contamination
def. intent to treat analysis
analyze subjects by group assignment, even if they did not follow through
what happens if don’t do intent to treat
you have ruined the randomization!
what is problem with ITT
bias towards the null
- increases confidence if find result
def. cluster randomization
groups of particpants assigned randomly
when is cluster good
when can’t get indivs, and can help get large number
probs with cluster
increases bias
- weakens stats. inference
what is N of 1 trial
one indiv followed over time and time is blocked into discrete periods
when to use Nof1
tailoring treatment to pts
issues with nof1
good for inidivduals
- can;’t generalize
def. crossover design
same as RCT but switch groups at some point
