optimal treatment - non-linear function of the symptomes

["nikolay" <niskrev@umich.edu>]

I apologies for asking a question not directly related to S.
I wondered if there is a result (in medicine) that the optimal treatment is
a non-linear function of the symptomes, e.g. no treatment prescribed for
mild symptomes, but some treatment - for more severe symtomes. Can anyone
recommend good reference - both theoretical and empirical on this topic.
Thanks