I don't remember seeing any relevant research but, in thinking about your question, I realized that I handle magical thinking a bit differently in psychotic clients than in non-psychotic clients.

With a non-psychotic client I'd probably spend a brief amount of time talking over the question of whether thinking about bad things (or talking about bad things) really makes them happen and then quickly move to transforming this belief into a testable hypothesis ("If I think/talk about ___________ then ____________ will happen.") and set it up as a behavioral experiment (On some days think/talk about _______ and observe what happens. On other days, refrain from thinking/talking about ______ and observe what happens.).

With a psychotic client (who's on appropriate meds) I'd be much gentler and take it more gradually. I'd want to find out if they mean that talking about it makes if feel more real or if they mean that talking about it makes it more likely to happen. Either way I'd want to find out what convinces them that this is true. I'd also want to find out what their experiences have been... Have there been times in the past when they talked about their fears? If so, what happened then? Have there been times in the past when they intentionally refrained from talking about their fears? If so, what happened then?

I might also want to find out about relevant beliefs in her household or in her family of origin. Many cultures have some for of the idea that what you think or talk about can shape the future. In the southern US (where I grew up), if you mention something you plan to do in the future you say "God willing and the creek don't rise" to keep from jinxing it. My understanding is that in eastern european Jewish tradition, when a baby is to be born you don't choose a name or fix up the baby's room, etc. to keep from jinxing it. Important people in her life may endorse the idea that she needs to be careful what she talks about or it will cause something bad to happen.

I probably would eventually get around to examining the pros and cons of keeping her fears to herself and to testing whether talking about her fears makes them come true (and whether keeping them to herself prevents them from coming true). But I'd approach it gradually and gently, I'd slow down if there are signs that this is raising her anxiety too much, and I'd start out the behavioral experiments with something that's not very threatening at all. For example, we could talk about getting a flat tire on some days and intentionally refrain from talking about this on other days then observe if this makes a difference in how often we get flat tires.

I don't know of any evidence that this is the "right" approach, it's what seems to work best in my somewhat limited experience with psychotic clients.