Medicine as practiced now has both scientific and non-scientific parts. Original research, as in bioinformatics, drug discovery, etc. constitutes the scientific part as it uses systematic experimentation, logic, statistics, etc. to deduce universal laws that can be independently verified.

However, the application of this knowledge to actual individuals in an doctor's office or hospital is an inherently subjective and non-scientific process, involving subjective inputs such as 'Rate your pain on a scale of 1-10.', 'How are you feeling after we upped your dose?', 'Is this treatment plan allowing you to get back on the job?' etc. The value a doctor provides is taking the vast corpus of medical knowledge (both scientific and not) and applying it to a single individual. This is not science, but is critical for the medical treatment process.

So what's the argument, that we should just wait a few centuries to help anyone who is mentally ill because we don't know enough about the underlying mechanisms?