Lecture 5 - RCTs Flashcards Preview

Undeleted > Lecture 5 - RCTs > Flashcards

Flashcards in Lecture 5 - RCTs Deck (16):
1

what are non-randomized experiments

- assignment not random
- when randomization not possible
- or when RCT incorrectly performed

2

keys to RCT

- random assignement
- must be uncertainly about which treatment is better
- results less generalizable since on one therapy at a time

3

basic steps

1. create cohort - meet inclusion
2. may observe pts. for period of time to see if stable
3. assign randomly to groups
4. followed and assessed

4

strenghts

- minimal diffs between groups
- removes bias with blinding
- can calculate incidence rates
- gold standard

5

weaknesses

- only ethical if not sure of treatment
- high cost
- poor external validity
- contamination

6

def. intent to treat analysis

analyze subjects by group assignment, even if they did not follow through

7

what happens if don't do intent to treat

you have ruined the randomization!

8

what is problem with ITT

bias towards the null
- increases confidence if find result

9

def. cluster randomization

groups of particpants assigned randomly

10

when is cluster good

when can't get indivs, and can help get large number

11

probs with cluster

increases bias
- weakens stats. inference

12

what is N of 1 trial

one indiv followed over time and time is blocked into discrete periods

13

when to use Nof1

tailoring treatment to pts

14

issues with nof1

good for inidivduals
- can;'t generalize

15

def. crossover design

same as RCT but switch groups at some point

16

when to use crossover

- studying physiological effects
- when not ethical to withold treatment

Decks in Undeleted Class (589):