John, cluster sampling is used to measure all sorts of things, not just “natural” phenomena (eg take-up of immunization vaccines). As long as the clusters are appropriately chosen then the sample should be representative of the experiences of the population as a whole, and therefore those that aren’t affected by war won’t show up as so being. I, for one, don’t have the knowledge to assess whether the clusters in this study are appropriate, nor I suspect do many people (including, quite possibly, the peer reviewers), but I’m not one to immediately suspect fabrication merely because I find the conclusions unpalatable. It would help us to know more about the situation if some other researchers repeated the experiment with different clusters.