It's quite possible that the causality goes the other direction, and a baseline population of brain disorders is being ascribed to the last notable thing that happened to them (which has also happened to ~ the entire population at this point).
In most wars until comparatively recently it was common to have 25yos in charge of 100s to 1000s of men, including on occasion general officer rank. Even with modern peacetime career trajectories it is possible to have a 25yo captain commanding a company of ~200 soldiers.
People have been managing garbage collection schedules for decades now. It's quite possible for many systems to have completely deterministic performance, with the allocation/deallocation performance made extremely fast, gc restricted to certain times or a known constant overhead, etc. Ironically, from a programming perspective it's incredibly easy in a language like Java to see exactly what allocates and bound those cases.
Conversely, it's also possible for reference counting to have perverse performance cases over a truly arbitrary reference graph with frequent increments and decrements. You're not just doing atomic inc/dec, you're traversing an arbitrary number of pointers on every reference update, and it can be remarkably difficult to avoid de/allocations in something like Python where there's not really a builtin notion of a primitive non-object type.
Generally speaking, memory de/allocation patterns are the issue, not the specific choice of reference counting vs gc.