Take a look at Section 19.3 of http://hbiostat.org/doc/bbr.pdf and especially the links that are listed in that section, starting with this. This is not categorization but instead is a translation of risk into an action. In the absence of a utility function the usual approach is for the physician to give the patient the best estimate of risk of an outcome under two treatment options, and to let the patient with some assistance from the physician decide if risks are the best course. Even though the utility function is not enunciated, it’s still there.
I haven’t seen an article about translating probabilities of various severities of outcomes into multiple possible actions. I hope someone can point us to such an article.