6004 Background: We have previously developed and reported on performance measures assessing the aggressiveness of cancer treatment near the end of life for selected cancers during 1993–96. Methods: We compared the care delivered in the 77 Health Care Service Areas (HCSAs) monitored by the Surveillance, Epidemiology, and End Results (SEER) program to all Medicare-eligible patients aged 65 and over who died of cancer (any diagnosis) between 1991 and 2000. We used hierarchical regression models to estimate regional variation in both levels and trends of each indicator. We then ranked each region according to the model-estimated rate of each indicator and computed the correlation among relative ranks of each region over the ten-year study period. Results: 215,488 patients met eligibility criteria. Within this broader and more recent sample we confirmed previous observations of steadily and significantly increasing use of chemotherapy within 2 weeks of death, emergency room visits, and intensive care unit admissions in the last month of life, and, among those admitted to hospice, an increasing proportion of late admissions within 3 days of death. There was significant regional variation in all measures, but the relative rankings of health care service areas from one year to the next were stable, with correlations of ranks ranging from .91–.98 from 1991–1992, and .66–.84 over the 5-year span from 1991–1995. Because of significant regional variation in trends, we found only moderate correlations ranging from .40–.61 over the entire decade. Conclusions: Cancer treatment near the end of life continued to become increasingly aggressive over the 1990s, however, there was significant regional variation in trends. The stability of regional practice patterns supports the reliability of these measures for quality surveillance purposes. [Table: see text] No significant financial relationships to disclose.