Underreporting of myeloid malignancies by United States cancer registries.

BACKGROUND The recent decrease in myeloid leukemia incidence may be directly attributed to changes in the population-based cancer registries 2001 guidelines, which required the capture of only one malignancy in the myeloid lineage per person and the simultaneous adoption of myelodysplastic syndrome registration in the United States. METHODS We constructed four claims-based algorithms to assess myeloid leukemia incidence, applied the algorithms to the 1999-2008 Surveillance Epidemiology and End Results (SEER)-Medicare database, and assessed algorithm validity using SEER-registered cases. RESULTS Each had moderate sensitivities (75%-94%) and high specificities (>99.0%), with the 2+BCBM algorithm showing the highest specificity. On the basis of the 2+BCBM algorithm, SEER registered only 50% of the acute myelogenous leukemia cases and a third of the chronic myelogenous leukemia (CML) cases. The annual incidence of myeloid leukemia in 2005 was 26 per 100,000 persons 66 years or older, much higher than the 15 per 100,000 reported by SEER using the same sample. CONCLUSION Our findings suggest underreporting of myeloid leukemias in SEER by a magnitude of 50% to 70% as well as validate and support the use of the 2+BCBM claims algorithm in identifying myeloid leukemia cases. Use of this algorithm identified a high number of uncaptured myeloid leukemia cases, particularly CML cases. IMPACT Our results call for the commitment of more resources for centralized cancer registries so that they may improve myeloid leukemia case ascertainment, which would empower policy makers with ability to properly allocate limited health care resources.