Positive predicted value and negative predicted value are used by clinicians to evaluate how likely a disease stage is present given the test results. In contrast, positive and negative likelihood ratios (LRs) are used in practice to assess the potential utility of a specific diagnostic test and the likelihood of a patient having the condition. This article introduces the concepts and the derivation of generalized predictive values and LRs from binary diseases to ordinal multistages diseases. We evaluate the performance of the proposed methods with numerical examples. We illustrated the proposed methods provided using real data.