This is an idea I have considered for many years. It fits with my overall view of a lot of evidence from treatment studies.
For many actions in life, an event either happens, or it doesn't. This seems obvious, I guess. You either throw a ball, or you don't. You either show up for work, or you don't. (Mind you, in my own case, I would say that my own modest level of athletic skill causes me quite often to "sort of" throw a ball, or "sort of" swim.)
In medicine, many actions are similarly unambiguous. The surgical removal of an appendix either happens, or it doesn't. An infection either responds to an antibiotic, or it doesn't. Clear.
Yet, I find that many treatments in medicine are much less clear.
In the case of psychiatric treatments, it has been a theory of mine that the drug (or therapy) may reduce the probability of a symptom occurring, in addition to, or instead of, directly reducing the symptom (or not). This phenomenon may be apparent not only in studies of populations, but in an individual.
Many disease processes are driven by multiple variables, which, together, alter probabilities of symptom recurrence. The variables may include genetic factors, environmental stress, etc. There may be a core phenomenon in nature, as manifest on a chemical, or even quantum-mechanical, level, of minute, truly random events, influencing a cascade of effects. The presence of a medication in the body may be just one more variable, influencing the likelihood of a symptom occuring, or developing, or advancing.
Some medications may appear not to be working, if a short-term view is taken. But in a longer-term view, it may be seen that symptom frequency and intensity is diminished. This is consistent with the theory that the medication affects probabilities.
This theory supports the idea that medications, and other psychiatric treatments, could have an important preventative role, beyond their role in an acute situation. And it encourages giving treatments a long period of time to work--at least months, if not years-- in order to most accurately assess effectiveness.
There have been some long-term studies which support this idea, but unfortunately most of the treatment studies in psychiatry have been relatively short-term (only a few months of follow-up, rarely more than a year).