Simply comparing variations in the structure or process of care does not provide an evaluation of the quality of care, although it may point to potential quality problems that merit further inquiry. In one such study of variation, five-year survival rates during 1983-1991 varied markedly for several cancer sites. For women with breast cancer, for example, five-year survival ranged from 71.0 percent in Iowa to 79.9 percent in Hawaii (Farrow et al., 1996) (Table 4.2). These differences persisted after adjusting for age and stage. For all cancers other than ovary and bladder, one or more regions were found whose survival rates differed significantly from the overall mean. These differences persisted and were even more pronounced when the analysis was limited to patients less than 70 years of age with local-stage surgically treated disease. However, other important case-mix adjusters, such as the presence of comorbid illnesses, were not included in the model, so interpretation of these results is difficult. Thus, it is not clear whether these regional variations in survival from breast cancer reflect differences in patient populations, regional differences in quality of care, or other factors.

Data for quality assessment can come from several sources. First, administrative records are widely available, and although they are limited in clinical detail, they can be used to show intensity or patterns of utilization. For example, they can be used to determine whether a patient with large-cell non-Hodgkin's lymphoma (NHL) received at least six months of chemotherapy and whether the patient had a white blood cell count performed before receiving chemotherapy. Second, medical records can provide greater clinical detail, the recorded medical history, the results of laboratory tests, and the treatment plan. For example, the medical record can show