I'm working on a meta analysis of prospective studies assessing changes to mental health in the year following an event. For the purposes of example, say it's the severity of depressive symptoms (expressed as a mean and SD) in the year after giving birth. To be included in the analysis, each study had to report on depression severity at at least two assessment points. All participants in the studies gave birth- there are no comparison groups. The depression measures and number and timing of assessment points differ across studies.
I would like these analyses to yield an expected severity of depression at each range of time post-birth and explore potential moderators of the severity of depression severity within each range of time.
My questions are as follows:
- Given that the studies use different measures, I recorded the range of each depression scale and am considering transforming all means to a 0-100 scale and meta-analyzing them as if they were proportions (with additional Freeman-Tukey transformations for analysis). I would like confirmation that this approach makes sense as a way to handle repeated-measures means and advice regarding how or whether to use the SDs in this process.
- To get summary statistics and test moderators for each time range, I'm not sure whether it would be better to run separate models for each group of effect sizes (e.g., one model for means at 0 to 1 month, a second model for means at 1 to 3 months) or enter dummy-coded time since birth into a single model containing all effect sizes. The latter is obviously more parsimonious, but I want to be sure it doesn't have limitations I'm not considering.