Session 6 Flashcards
What strengthens a case for a causal link?
Evidence based medicine?
3 features of evidence based medicine?
Biologically plausible mechanism
Use of current best evidence to make decisions about the care of individuals.
Clinical judgement, patients values and preferences,relevant scientific evidence.
Unfiltered information in hierarchy of medicine?
Filtered?
Scientifically proven means?
RCT’s, cohort etc
Critically appraised individual articles, critically appraised topics, systematic reviews.
It’s better than the other treatment
How do you calculate the confidence interval?
How does odds ratio relate to confidence interval?
If odds ratio is >1?
Odds ratio x fro max and / error rate for min
Odds ratio if the best guess of the exact middle value in the confidence interval
New treatment is better than the other
What’s the null hypothesis?
If null hypothesis (1) is in the confidence interval?
If 1 is in CI what is the p value
No statistical significance
Not significant as you accept the null hypothesis,so it could be due to chance
P value is >0.05 as we know data is inconsistent
Define a systematic review?
Meta analysis?
Benefit of meta analysis?
Overview of primary studies that used explicit and reproducible methods
Systematically collates study results. Quantitative synthesis of the results of two or more primary studies that addressed the same hypothesis in the same way. (It’s like quantitative pooling of results) So meta analysis is a type of systematic review.
Reduces problems of interpretation due to variations in sampling size
Describe the process of meta analysis
In a forest plot what is the relevance of the square size?
Releavance of the diamond in a forest plot?
Solid line in forest plot?
All studies have there odds ratio’s and CI’s calculated. These are then combined to give a pooled estimate odds ratio. Studies are weighted on their size and uncertainty.
Odds ratio and the size is the proportion of weight given to the study.
The centre is the pooled odds ratio and the width is the pooled CI
Null hypothesis
Two types of models for meta analysis?
Odds ratio same for both?
Why use random effects model?
Fixed effect model-assumes studies are estimating the same true effect size, they are using the same things and the same outcome. All studies are trying to get the same true effect and the only reason they don’t. Get there is due to random variation.
Random effects model-assumes studies are estimating a similar true effect size but not the same.therefore it has a wider confidence interval. Moreover more equal weighting’s so small studies get more weighting. They take the the mean true effect and the outcomes can be slightly different but have a common link.
Usually
It accounts for variation
What is used to explain heterogeneity?
In meta analysis what can cause the variable quality of studies?
Other issue of meta analysis?
How to solve these two issues?
Sub group analysis
Poor study design,poor study protocol, poor protocol implementation
Certain studies are more prone to confounding and bias for example RCT’s are less prone and case controlled are very prone.
Define a quality standard and only include certain studies (RCT’s). Or score each study for its quality and then take that into account when weighting.
Main problem when selecting studies for meta analysis?
How to prevent publication bias?
Publication bias where studies especially small are picked to suit the desired outcome, so more likely picked if statistically significant.
Check meta analysis protocol for identification of studies.
Check statistical tests as they are often weaker.
Funnel plots (measure of study size/standard error against measure of effect/odds ratio)
On a funnel plot what indicates no bias?
What do smaller studies look like on funnel plots?
Symmetrical funnel, with studies which are not statistically relevant
They stray further from the central effect size
