Exam One - Clinical Relevance of Change Flashcards
5 standards of clinical relevance
- effect size
- standard mean difference
- standard error of the measurement
- minimum detectable difference
- minimum clinically important difference
alpha level
your tolerance for declaring a sample mean “significantly different” if it falls outside the expected range by choice
P<0.05
statistically significant
reject the null hypothesis
“unlikely enough” to conclude the observed result did not occur by chance
P>0.05
not statistically significant
fail to reject the null hypothesis
“not unlikely enough” to conclude the observed result was not a chance occurrence
problems with black and white philosophy??
P<0.05 is arbitrary and subjective
do we ONLY want to know if the difference could be attributed to chance alone?
no! We also care if the change was clinically relevant!
What measure evaluates treatment effect, independent of sample size?
effect size! (ES)
ES =
(change score)/(average standard deviation)
Es is the expression of the magnitude of the difference _____________
between 2 means
what measure is a comparison of 2 treatment effects, independent of sample size?
Standard mean difference (SMD)
SMD =
(treatment A change - Treatment B change)/(average SD)
SMD is the expression of the magnitude of the difference _______________-
between 2 CHANGE scores (4 means)
ES, SMD interpretation scale
> .8 BIG treatment effect
.4-.8 effect is big enough we can be confident a true effect exists
.2-.4 treatment effect is small enough that clinician may not perceive a benefit/harm
<.2 what treatment effect??
What are two other standards to measure clinical significance?
- Minimum clinically important difference (MCID)
- minimum detectable difference (MDD or MDC)
MCDI is…
- the smallest diff a patient would perceive as beneficial
- specific to the individual patient goals and values
with MDD/MDC we must ask ourselves…
is the effect outside the bounds of potential measurement error?
What formula do we use when considering the MDD/MDC?
Standard error of measurement (SEM)
SEM =
SD * sqrt (1-ICC)
MDD95
the difference required to be 95% confident the effects are not due to measurement error
MDD95 =
2.77*SEM
measures with low MDD are considered _________-
responsive
What are the measures to judge the relevance of continuous effects?
- effect size
- standard mean difference
- standard error of the measurement
- minimum detectable difference
- minimum clinically importance difference
What are the measures to judge discrete effects?
- absolute benefit increase/risk reduction
- relative benefit increase/risk reduction
- number needed to treat
Many studies use______ to measure effects of an intervention
discrete
ex - success vs failure
discrete results are often analyzed by _____-
count (# or %)
Number needed to treat (NNT)
number of patients who must receive the intervention to produce one positive outcome or avoid one adverse event (relative to control condition)
ARR =
absolute risk reduction = % controls w?prob - % therapy group w/ prob
RRR =
relative risk reduction = (% controls w?prob - % therapy group w/ prob)/(% of controls w/ prob)
NNT =
1/ARR
(1/% controls w?prob - % therapy group w/ prob)
ABI
absolute benefit increase = % therapy group w/ outcome - % control group with outcome
RBI
relative benefit increase = (% therapy group w/ outcome - % control group with outcome)/(% of controls w/ outcome)
NNT =
1/ABI
(1/% therapy group w/ outcome - % control group with outcome)
ceiling effects
scores that are already high are not likely to go higher
floor effects
scores that are already low are not likely to go lower
change scores should be interpreted with respect to…
baseline scores