Hi! I am conducting a meta-analysis and I have one study which used two different tools to measure the same outcome.
Will it be appropriate two make the two comparison and pool both results along with other studies measuring the same outcome? Or do I omit one - and include one in the meta-analysis whereas the other tool is narratively synthesised?
Many thanks!
I recommend including the outcome measure, which resembles the outcome measures in the other studies most, and potentially discuss the other in a narrative manner. If you included both measures, the present study would have greater weight in the meta-analysis, and study samples would not be independent, since the study population is included twice.
If there is no reason to think one measure is more valid than the other, you _should_ include both, especially if the effect on these measure differ. Unless you do that you are choosing to bias the final outcome of your meta-analysis. It is not at all uncommon to include more than one effect from studies with several outcome measures. You will, however, need to take the dependency between these measure into account, and adjust the weights accordingly.
What software are you using to analyse your data? I strongly recommend Metafor in R.
I have attached two of my studies, in which we have included many studies with more than one outcome.
No, you are not allowed to include two results of the same paper in your MA. This will bias your outputs. Also, if two separate studies have worked on the same data [considering an online public database], again, you gotta exclude one.
P.S.: I agree with Jon on using R [packages meta, metafor, dmetar] for pooling, bias checking, etc.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis