I am completely new to meta-analysis. My objective is to look for predictors (effect) of injury (outcome) in the workplace. One issue I am facing is several studies (high quality ones) have several significant predictors. How do I tackle that?
E.g. One study found that age and job tenure predict injury. Another study found age and sleep predict injury.
Hi Basem,
This is a common question as you start out with your meta-analysis. It is also fairly easy to tackle, I believe.
If a question has multiple outcomes that are sufficiently dissimilar that you cannot assume they are the same, you would need to conduct a meta-analysis for each outcome. So, your paper could report multiple meta-analysis. A similar approach was employed here http://www.ncbi.nlm.nih.gov/pubmed/23915170.
If these variables are covariates of the main question, then you may need to explore their influence using meta-regression.
I believe age, sleep, and job tenure are sufficiently dissimilar and should be pooled separately.
Hope this helps.
I totally agree with Ajibola
All of my published meta-analyses on sentinel node mapping have two outcomes: false negative and detection rates and I pooled each separately
Ramin
Be advised if the studies are similar enough to pool the results. Consider time of intervention / observation, type of study, definitions of outcome and characteristics of the study sample in terms of prognostic factors.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis