This paper describes the design and preliminary results of research being conducted by Blue Cross of Western Pennsylvania to measure hospital case mix. The model of patient management used in this research interrelates symptoms, diagnosis and treatment. Analyses of detailed patient data have indicated that patient classifications that are based on discharge diagnosis, singly or in combination with other variables such as secondary or multiple diagnoses, procedure and age, do not necessarily result in patient categories that require similar management or similar hospital services. Patients who are clinically similar, and even the same patient, can have a number of diverse, but appropriate, reasons for being in the hospital, and their use of hospital resources in each hospital episode will differ accordingly. The implications of including all reasons for hospitalizing these patients under the same rubric are clear: the resultant category would not be homogeneous with respect to resource use or hospital costs. Any case-mix index constructed using such categories as its basis could not only be misleading, but could also be financially damaging or extremely profitable to selected hospitals if used in hospital reimbursement. Both the model presented and preliminary analysis will be useful in designing other strategies for research and application in the area of case mix.