Data-source choice in diabetes mortality surveillance and equity monitoring in the United States, 1999-2023.

AIMS: To determine whether data-source choice alters conclusions about diabetes mortality trends, high-burden jurisdictions, and income-related inequality in the United States. METHODS: We compared crude mortality rates (CMRs) and age-standardised mortality rates (ASMRs) from the CDC WONDER and Global Burden of Disease 2023 (GBD 2023) across 51 US jurisdictions during 1999-2023. Agreement, trends, high-burden classification, source discrepancy, spatial autocorrelation, and inequality were assessed using concordance analysis, Joinpoint regression, mixed-effects models, Moran's I, the slope index of inequality, and the concentration index. RESULTS: Across 1275 jurisdiction-year observations, CMR estimates were correlated but incompletely concordant (Pearson r = 0.76; concordance correlation coefficient = 0.72); concordance declined from 0.82 in 1999-2019 to 0.50 in 2020-2023. National CMR increased in CDC WONDER (AAPC, 0.60) but decreased in GBD 2023 (AAPC, -0.42). Opposite CMR trends occurred in 10 jurisdictions, and six had discordant high-burden classifications in 2023. The adjusted post-2019 signed discrepancy was -3.09 per 100 000, with no consistent spatial autocorrelation. Both sources showed higher ASMR in lower-income jurisdictions, with a steeper absolute gradient in CDC WONDER. CONCLUSIONS: Source choice changed conclusions about mortality trends, jurisdictional priority, and inequality. Discordant findings highlight the need for task-specific source selection and multisource review.