ment, and follow-up. Participating hospitals submit all cases seen at their hospital for a particular data year. The system appears to have a bias toward hospitals with a computerized cancer registry, and it does not provide comprehensive outpatient data. After 1996, hospitals with ACoS accreditation (about 1,450 hospitals) were required to participate; it is estimated that in the year 2000, 1,750 hospitals caring for 80 percent of U.S. cancer cases will be participating. The NCDB provides comparisons of cancer management patterns and outcomes to national norms at the hospital, community, and state levels. The NCDB can also be used to track how well the results of major clinical trials are incorporated into clinical practice. There have been questions about quality control of the data collected by individual hospital registries, and NCDB is working to improve the quality.

The NCDB and the SEER program, when compared in 1992, provided similar patient descriptors (e.g., age, race, gender), cancer characteristics (e.g., stage), and types of surgical treatment for breast, colon, lung, and prostate cancer (Mettlin et al., 1997a). The two registries had similar distributions of cancer cases, by clinical characteristics.

Registries represent an exceptionally valuable opportunity to conduct quality assessment on a broad level. They could go further in collecting data on explicit process measures, intermediate outcomes, and treatment information, as well as characteristics needed to risk-adjust the outcomes. They would also be more useful for quality assessment if they were able to reduce the time lag between provision of care and availability of data. This information could be used both to provide detailed information on quality of care and to tie the processes of care to the outcomes. These data would also be useful in quality improvement.

There are a number of other ongoing data collection initiatives assessing the quality of cancer care:

• The American College of Radiology (ACR) conducts ongoing Patterns of Care Studies for a variety of cancers. Since 1971, ACR has collected information periodically from a national sample of radiation oncology facilities (Kramer and Herring, 1976). Patterns of Care data have been collected in 1972-1974, 1977, 1979, 1984, 1989, and 1994. Data collected at each of these periods have included structure and process information; some years have also included patient outcomes measures. For example, 1989 assessed only structure and process measures, whereas 1994, which has not yet been made publicly available, included outcomes assessments as well.
• The National Comprehensive Cancer Network (NCCN) has developed an outcomes database that pools information across NCI-designated comprehensive cancer center participants. In 1997, NCCN began the creation of a uniform outcomes reporting system (Weeks, 1997); results of this effort are pending.
• The American Urological Association is developing a Documented Outcomes Collection System (DOCS) to assess patient outcomes for a selected number of conditions and may collect information related to prostate cancer outcomes in the future.

Efforts to measure quality of cancer care in the United States are in the early stages. National organizations conducting quality assessment have focused primarily on prevention and