What I am looking for is a dataset that has a number of independent variables, such as age, sex, smoker/non-smoker etc. for which I intend to carry out supervised learning for the dependent variable which would be a mental health diagnosis such as major depression, schizophrenia or a personality disorder.
I am aware of the SAMHDA data on drug abuse but that is not quite what I am looking for.
(I do not have institutional access to any such thing.)