I have a study which reports PTSD diagnostic status at both 3 and 6 months. In my meta-analysis of the intervention, both 3 and 6-month assessments fall within the same outcome assessment (for example, short-term is 3-6 months and long-term is 9-12 months). Can I include both 3 and 6-month data within the same short-term outcome analysis? This seems wrong as the population is the same, just at different time points. If so, how do I deal with such data? Many thanks in advance.
It depends on your protocol (plan), so we usually decide before we start. The content experts included in the meta-analysis should decide a specific time point in the protocol rather than non-defined terms as short- or long-term outcomes.
So to answer your question, if you mean to include the 3 and 6 months outcomes as 2 independent sets (as if they came from 2 different studies), so no you can't. But if you will split the short term outcomes' analysis into 2 time points, then simply yes you can.
No, this will give a wrong estimate and double the weight of this study which is not correct statistically.
They can be added in two different independent models (forest plots) but they can NOT be pooled together.
Ahmed Negida thanks so much for your response - I thought as much. When you say they can be added in two different independent models, I would have to break up my outcomes. Currently, they are short-term (3-6 months) and long-term (9-12 months). For the sake of this study, I would have to break this up into 4 seperate time points. Is there any other way to handle such data that does not involve significantly increasing the number of analyses?
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis