Although severity of illness may affect the usefulness of a diagnostic test, most previous work has considered test performance only in the dichotomous situation where the target disorder is considered present or absent. In this paper, a clinical example is provided and two methods presented whereby illness severity may be incorporated into the determination of likelihood ratios for a diagnostic test. The use of this approach is that given a test result, a clinician may determine the post-test probability of all levels of the illness in the patient under study.