You can talk about MRIs, or genes, or serotonin all day, but that doesn't mean you know what depression (or other mental state) is - using the current tools you assume you know what it is and that is the problem.

One way to measure objectively is to look at whether someone is employable and maintains employment. But that is far too low resolution and non specific. People change their mood during a day and from one day to a next, and there must be physical correlates in principle.

There is a huge human cost to the fact that nobody knows what they are doing, because they cannot relate treatment to anything objective. Sometimes a drug seems to be not working well after several years, and the dosage is raised, and all the while the side effect of the drug is what's causing the problem, so it gets worse the higher the dosage goes. In the mean time, that person becomes unable to work, and is caught in a series of catch-22s.