Fascinating discussion in the first few pages, and really interesting to see so many people on ARRSE who have taken a professional / education interest in this.
From a totally amateur perspective, but having lightly read some of the more serious literature, unlike
@jarrod248 I'd be totally unsurprised that screening wasn't shown to work. HOWEVER, I'd point out this doesn't mean that screening
cannot work, it may simply mean that we don't know what or how to screen for.
One of the criticisms of more recent, serious scientific psychologists is that what is broadly thought of as psychology has been, for a long time, incredibly unscientific. There have been some good reasons for this (complex interacting data sets, lack of equipment to accurately measure results, lack of statistical training, etc). But it would mean that a lot of the accepted truths and textbooks of the profession are unreliable. It's unfortunate that some conspiracy theorists (Scientologists) have picked up on this and run with it, but serious professionals like Kahneman increasingly note the same thing.
Looking at diagnoses in particular, and the systems that select and feed patients towards a diagnosis, it seems that a glaringly obvious problem is selection bias. People often get fed towards a particular diagnosis not so much by the symptoms they display, because symptoms tend to be quite general across a number of diagnoses, but because of the context of the diagnosis, patient, and system that has fed them in. So soldiers are much more likely to be diagnosed with PTSD. Selection bias doesn't mean that a higher than average number of soldiers
don't get PTSD. It means that if you had a perfectly normal population of people, and put them through that selection system, statistically you would expect them to have a higher than average rate of diagnoses of PTSD. Any screening process emphasises that - you are screening
for something. So you will tend find it.
What is not at all clear is that the process really understands what it is screening for. Someone posted a Aspergers / autism spectrum questionnaire earlier. The main problem with anything like that is: it's an Aspergers / autism spectrum questionnaire. The psychologist, or worse, sometimes the patient too, knows what is being tested for. But a lot of the questions on there are interchangeable with, say, questions on a antisocial personality disorder test. Replace or add some of the questions with ones about violence, and it becomes a different test set. Now apply that to a niche population like soldiers: how does that question set work? Particularly the violence questions? I understand those online tests aren't valid, but the point stands: the symptoms and methods of diagnosis are often very similar between radically different diagnoses. In any case like that where symptom sets may be representative of multiple causes, you need strong differential diagnosis, a high level of granularity and data, and eventually, you need some unique identifiers.
The problem is, there are very few such treatments or indicators available. Equipment which gives the required level of data isn't widely available, and therefore there usually isn't sufficient data to establish firm results. In other words: the science of psychology just isn't very advanced yet. Arguably, a lot of the results from many psychological tests should say: not enough data. Instead, some talk about the strongest correlation, even when it is too weak to be statistically significant.
My hunch would be that in a hundred years, people will understand psychology not as a set of distinct, diagnosable disorders, but a complex network of interacting symptoms, which produce highly individual results in each patient, much like where gene therapy is going. It will be less likely to diagnose people as "having" PTSD, Aspergers, autism, APD, and more likely that they will be marked as diverging from a middle range in some aspects: they are particularly good at this thing; particularly bad at that thing; pretty much normal elsewhere. I think that will look very, very different to how we expect people with any of those diagnoses to be and behave today. I also suspect our blanket treatments (particularly the US instant-chemicals approach) will be seen as medieval.
Curious to see what everyone else on here thinks, having collectively spent a lot more professional time on this.
TL : DR Screening tends to find what it is looking for, and requires that you know what to look for and how. It's not at all clear that we know either of those things.