Assessing the discriminative ability of risk models for more than two outcome categories

Assessing the discriminative ability of risk models for more than two outcome categories