- many studies are difficult to interpret because dissimilar groups of patients were compared (e.g., there are insufficient controls for important clinical characteristics such as the presence of diseases other than cancer);
- studies are confined to a small group of patients, in one or a few institutions, states, or health plans, making it difficult to generalize to all patients with cancer; and
- studies are often based on data from cancer registries, which do not capture important aspects of care (e.g., outpatient treatment may not be included).
Two national databases are available with which to assess the quality of cancer care, but each has limitations.
- The Surveillance, Epidemiology, and End Results (SEER) cancer registry, maintained by the National Cancer Institute (NCI), when linked to Medicare and other insurance administrative files, has been valuable in assessing the quality of care for the elderly and other insured populations. It is also useful in identifying a sample of cases for in-depth studies of quality-related issues. The SEER registry, however, covers only 14 percent of the U.S. population in certain geographic locations, so it may not adequately represent the diversity of systems of care. Finding ways to capture measures of process of care, treatment information, and intermediate outcomes—and to improving the timeliness of reporting—would enhance the registry's use in quality assessment.
- The National Cancer Data Base (NCDB), a joint project of the American College of Surgeon's Commission on Cancer and the American Cancer Society, now holds information on more than half of all newly diagnosed cases of cancer nationwide and includes many of the demographic, clinical, and health system data elements necessary to assess quality of care. A limitation of the NCDB is the absence of complete information on outpatient care. The NCDB has not yet been widely used to assess quality of care, but has great potential for doing so.
Existing data systems must be enhanced so that questions about quality of care can be answered comprehensively, on a national scale, without delays of many years between data collection and analysis. An effective system would capture information about
- individuals with cancer (e.g., age, race and ethnicity, socioeconomic status, insurance or health plan coverage);
- their condition (e.g., stage, grade, histological pattern, comorbid conditions);
- their treatment, including significant outpatient treatments (e.g., adjuvant therapy, radiation therapy);
- their providers (e.g., specialty training);
- site of care delivery (e.g., community hospital, cancer center);
- type of care delivery system (e.g., managed care, fee-for-service); and
- outcomes (e.g., satisfaction, relapse, complications, quality of life, survival time, death).
It may be costly and difficult to obtain all of the desired data elements for all individuals with available sources, so sampling techniques could be used to make the task manageable for targeted studies. Alternatively, it may be feasible to link some databases (e.g., those describing structural aspects of care such as hospital characteristics) to other existing databases. It is un-