Evaluating the evidence for an intervention Flashcards
What are the 5 As?
Asking - question Accessing - search Appraising - critical evaluation Applying - extracting Assessing (re) - evaluate process
Essential components to a background question (2)
- Question root and a verb, e.g. who
2. Disorder, test, Rx etc
Essential components to a foreground question (2)
PICO
What is the best type of randomisation?
Off-site, computerised system
What does the standard deviation measure?
Measure of the distribution of values within a sample.
To correct for not having the full population for SD?
Sample - 1
What does the standard error measure?
Amount of variability in the sample mean.
It defines the limits of the distribution of sample means.
95% Confidence Interval = in SE?
~2SE
Hypothesis tests (aim, issues)
HTs determine if relationships are real or related to chance.
Problems
- no clinical importance shown
- may get type 2 errors
- p value is not a direct indicator of effect size
- size of the true effect is unknown
What is WMD?
Weighted mean difference
- different outcomes weighted based on sample sizes and differing precision.
Why use CIs? (3)
- Accuracy of sample statistic
- Helpful with clinical importance
- Function as a hypothesis test
How does sample size affect precision?
Higher precision with greater sample size.
Magnitude of treatment effect?
Difference between values
Precision of estimated treatment effect?
CIs
What are the three common measures of risk?
Absolute Risk Reduction (ARR)
Relative Risk Reduction (RRR)
Number Needed to Treat (NNT)
What is the CER and EER?
Briefly describe.
Control Event Rate
Experimental Event Rate
i.e. control group no. with event occurring/total control group, e.g. 26/32; 81.25%
What is ARR and how is it calculated?
Difference in risk between Rx and control groups.
ARR = CER - EER
What does a negative ARR indicate?
Greater risk in the Rx group (good if the event is positive)
How are CIs calculated for ARR?
95% CI ~= diff in risk +/- 1/sqrt(average ‘n’ of the two groups)
What is the NNT? (3)
The number of people you would need to treat (on average) to prevent the even of interest happening to one person.
i.e. a lower NNT is a more effective Rx
NNT should be related to a time period of Rx
How is the NNT calculated?
1/ARR
What is the RR?
Relative Risk.
- a measure estimating the size of the effect of a Rx vs another Rx/control.
- proportion of bad outcomes in the Rx/bad outcomes in the control group
- EER/CER
How are the results of RR interpreted?
Below 1 is good
1 is no difference
Above 1 is bad (more risk with Rx)
What does the RRR tell us and how is it calculated?
How much the Rx reduced the risk of bad outcomes relative the control group.
RRR = ARR/CER
Why should you be cautious of RRR? (example)
e.g. CER = 5.7% and EER = 4.4%; or CER = 0.000057 and EER = 0.000044
Both = 25% RRR
But much lower ARR in second one and way higher NNT.
How is an odds ratio calculated?
EER/(1-EER) / CER/(1-CER)
Is the evidence applicable to your patient?
- Subjects similar? (inc/exc criteria, prognostic indicators)
- Rx applied appropriately, and can I perform them?
- Are the outcomes useful?
- Does the therapy do more good than harm/are there side effects?
What does “Are the outcomes useful?” refer to? (3)
- surrogate measures aren’t usually, e.g. BMD for # risk
- outcome important to pt
- size of the effect clin sig?
What is an odds ratio? (2)
Odds that an outcome will occur given a particular exposure
COMPARED to odds of the outcome occurring without the exposure.