BACKGROUND: Race is associated with patterns of presentation and survival outcomes of gastric cancer in the United States. However, the impact of race on the receipt of guideline-recommended care is not well characterized. By using current recommendations, the authors examined the association between race and guideline-recommended treatments and identified factors that are predictive of variations in gastric cancer care. METHODS: By using the National Cancer Database for 1998 through 2005, 106,002 patients with gastric adenocarcinoma were identified. Multivariate analysis techniques were used to examine the association between race, the receipt of guideline-recommended care, and survival after adjusting for covariates. RESULTS: Although African-American and Hispanic patients were more likely to undergo adequate lymphadenectomy (> or =15 lymph nodes) and to receive care at comprehensive cancer centers and high-volume facilities (for all, P < or = .001), they were less likely to receive adjuvant multimodality therapy for American Joint Committee on Cancer stage IB through IV, lymph node-negative (M0) disease. Up to 60% of all patients who underwent gastrectomy failed to receive adequate lymphadenectomy and adjuvant multimodality therapy. The delivery of multimodality therapy varied significantly by stage and lymph node evaluation (P < or = .001). These findings persisted on our multivariate analyses, indicating that African-American and Hispanic patients received adequate lymph node evaluation (P < or = .001), whereas they were associated with receiving no adjuvant multimodality therapy (P < or = .025). CONCLUSIONS: There were significant variations in treatment for gastric cancer among ethnic groups in the United States. It was noteworthy that, although nonwhite race was associated with improved surgical care, gastric cancer care remained suboptimal overall. Cancer programs need to identify procedures to maximize the delivery of adequate gastric cancer care to all patients.

OBJECTIVE: To describe key methods and issues in conducting survival analyses, especially using Medicare (and other) administrative data. PRINCIPAL FINDINGS: Survival analyses are rich , informative, and underutilized methods for examining out comes whose timing is important . Medicare files contain the necessary information for conducting such analyses, including identification of cohorts, definition of events, censoring of observations, and adjustment for covariates. CONCLUSION: Survival analyses can readily be conducted using the information contained in administrative data files.