The dependability of medical billing diagnostic information was evaluated using a rating scale developed by Studney and Hakstian. Generalizability theory offers a comprehensive and flexible framework for analyzing problems associated with measures derived from multiple raters. The medical record and billing system diagnostic information from 45 patient encounters were rated according to similarity and value by three physician judges on two occasions. Data were analyzed using a three-way ANOVA design with levels of judges (3), occasions (2), and encounters (45). Examination of variance components and associated generalizability indicated that the judges and occasions factors or the interaction with occasions and other factors contributed little variance. The greatest proportion of variance was attributable to interindividual differences among the encounters and the interaction of judges and encounters. This investigation illustrates the application of a relatively new approach to practical measurement problems, generalizability theory.