I think you are on the right track here. If there is enough data available, the best method would be to do separate analyses for different follow-up periods (1 mo, 3 mo, 6 mo...). Periods that that do not have enough data would just be left out (although some people average as little as two studies...).

I do not quite follow why you think interventions should not be judged by the longetivity of the effect, but regardless of that, if you do a disaggregate analysis first (separate time periods), then you can always aggregate that over all values in the end. This kind of step-by-step aggregation allows the reader so see exactly what you have done, and extract the information that is of interest to this person.

However, as always, your dataset might have some features which make such an approach difficult or beside the point. If you present some more details about your research it will be easier to comment upon this.

regards

Thanks for that answer Anders. My point about longevity of effect was that an intervention may be considered effective even if it is only a short term effect, although ideally we would prefer longer term effects. This may sound counter-intuitive but perhaps this is because I'm not looking at medical interventions. The review and meta-analysis is exploring the effectiveness of interventions seeking to improve wellbeing for 16-24 year olds in terms of their impact on learning outcomes. I am at the stage of extracting the data for the meta-analysis and was just thinking about the follow up issue ahead of doing this. I don't know whether that info sheds any more light on my original question?

One point should be kept in mind that if all articles in your metaanalysis are mentionning same level of followup then you need to mentions the same like at 1, 2,3 months but if it is different in different articles then you need to take the common levels of follow ups from all.