As techniques to assess the quality of care have proliferated, appropriate concern has arisen about study methods employed by researchers. In particular, studies that focus on a single variable, such as mortality, length of stay (LOS), or complication rate, have sometimes been justly criticized because they have failed to account for differences in the case mixes of providers and the differing severity of illnesses among patients treated. The issue has become even more controversial as third-party payers have attempted to tie reimbursement mechanisms to a single variable, often median or average LOS or average charges for given diagnoses. In the February 1980 issue of the Quality Review Bulletin, the authors of this article discussed a method to investigate variations from areawide average LOS (see "Length of Stay Variations: A Focused Review," pagae 6). The authors concluded that actual differences exist among plhysician practice patterns for the treatment of the same conditions. Although not discussed in that article, the authors devised a method of analysis that incorporated case mix and severity of illness variables to determine whether the correlations they reported were valid. In this article, the authors explain the theory and application of the illness severity index they devised. The index may be of special interest to researchers concerned about whether they are comparing "apples with oranges" or truly discovering differences among providers' practice patterns.--The Editors.