RCTs and the Well Elderly Study Flashcards
What do I mean when I say that you should always
protect the integrity of your data?
Any dataset needs to be preserved in its original state, and any changes to
backups need to be documented.
What does a p-value really tell you when you are
testing null hypotheses?
If the null hypothesis were true and nothing was going on (e.g., the
groups were equal, or there was no improvement), then you would obtain your
result (or more extreme) this much of the time.
What are the two primary ways we’ve talked about
that randomness is exploited in research?
In observational studies, we use random sampling, and in experimental
work, we use random assignment. Rarely is there a reason to do both.
What are some good ways to limit drop-outs in
intervention research?
Follow your dropouts and take a lot of notes; also, design the study to limit participant time.
What is the problem with cherry-picking participants
for a study?
I think there are two problems, the first of which can be overcome. (1) By
recruiting folks who want to change, you may limit your generalizability to the
population of interest (e.g., instead of well older adults it becomes well older adults
who want to be helped by lifestyle redesign); (2) Regression to the mean will occur if
you select for extreme values (often the ones who want to change).
Imagine our OT graduate students are delivering an
intervention to well older adults, and they find that
the participants report improved self-perceived
health. What is the best reason to be suspicious of
the result?
A “niceness” effect is possible: participants may have been inclined to
report improvements simply to be nice—especially because the interventions were
facilitated by occupational therapy students who may have been perceived by
participants as being similar to their grandchildren.
What is the purpose of active control groups?
Active control groups allow you to compare your treatment with
something similar (e.g., yoga and exercise), which is often more realistic than
comparing something to nothing (yoga and no treatment)
What is the main problem with waitlist control in
RCTs?
Receiving anything may be better than receiving nothing
What would be the purpose of using a social group to
compare with a “life review” writing group?
To test whether improvements are simply related to the social benefits
What is intent-to-treat analysis for randomized
controlled trials?
Intent-to-treat seeks to analyze individuals based only on how they were
initially randomized, regardless of any deviations from treatment after baseline.
When should you conduct as-treated analysis for
randomized controlled trials?
After conducting intent-to-treat, and especially when nonadherence rates
are high.
What is Schelly et al.’s (2019, 2021) primary finding
with regard to the Well Elderly II data?
The individualized OT appeared to have a small indirect effect on
depression, but community outings were just as effective, and much cheaper.
From the following
study designs, what are the most obvious flaws?
Pick a retirement home; measure QL in the residents; conduct the
intervention and measure QL again. If QL improves, our intervention
works.
We have no way of knowing whether the intervention is better than weekly
bridge games and monthly visits by a high school student volunteer
Pick two nearby retirement homes with similar management (e.g.,
activities, schedules); conduct the intervention in one of them;
measure QL in both.
We should have measured QL before the intervention as well.
A quasi-experimental study compared outcomes (say, at age 30)
between students who had scored just above and just below a
college admissions cutoff in high school, where the former group
completed college. This study was designed to study the effects of
college on later life outcomes. Why is this an ingenious study design?
Because the groups are basically identical in terms of important lurking
variables, which is the aim of random assignment.
