Does calculating a weighted average right approach ?....epidemiological data sometimes overlaps..how to report this in systematic reviews?
If the studies are similar (design, diagnostic methods...) I would combine (stack) the 2 datasets, adding a dummy variable to indicate the origin of the data (survey 1 or survey 2). I would then run a global analysis, with the dummy variable among the set of explanatory variables. If the amount of overlapping is substantial (say, > 10%), it might be wise to use a statistical model adapted for repeated measurements (eg mixed-effect logistic regression).
Tendríamos que saber en qué consiste la similaridad para poder determinar la actuación, es decir, para ser más precisos en la intervención. Se podría incluir una variable control, un escenario distinto, pero utilizando los elementos similares. Otro aspecto es trabajar con la hipótesis contraria a la de uno de los estudios.
The question is not clear for me. It seems that the intention for combining these two studies is becuase they have different prevalence results, despite they were done in the same region . That is not a big deal, that is normal consequence of the sample variation, nothing more than that. Assuming that the two studies were well done. The easier way to solve this problem is to compare the central point estimation (the prevalence), and their confidence intervals.
Gus provides great input. I would expect the confidence intervals to overlap, suggesting the differences are not significant and the estimate essentially the same.
Well, why should prevalence estimate and confidence interval be similar between the two studies ? If the goal was to assess the spread of an infectious disease, we would rather expect to find differences. In this case we are more in the frame of repeated cross sectional studies than meta analyses.
The most important issue is WHY do these results differ. If the published prevalences report on the same time period, this is the result of differences in design or diagnostic methods, and you should not combine them. When the reported prevalences are on different (overlapping) time periods, the difference could be real, like with a fluctuating prevalence of an infectious disease. In the latter case you should decide whether you want the average prevalence over a longer period of time, or whether the most recent results are more useful.
It is hard to answer your question without a more precise description of the problem. I recommend that you provide some more detail, in particular explaining the objective of your study (I have the impression that you are reviewing the literature but it is not clear to which end), the kind of overlap that you suspect (e.g. overlap in time periods, overlap in the population presenting with the disease etc.), and which data you have access to relating to the two studies.
- Badges
- Science topic
- Similar topics
- Medicine
- Epidemiology