Evaluating quality of census tract data in the NPCR dataset.

BACKGROUND: Knowledge of data quality is essential for accurate interpretation and use of cancer data by cancer control and prevention programs and researchers.
OBJECTIVE: To assess the quality of census tract information in the Centers for Disease Control and Prevention's (CDC) National Program of Cancer Registries (NPCR) dataset. This assessment will guide analyses using census tract information from the NPCR dataset.
METHODS: The 2001-2005 data submitted for the 2008 NPCR Cancer Surveillance System (NPCR-CSS) were used to calculate the overall and registry-specific proportion of records that included information on the variable census tract 2000 (NAACCR item #130).
RESULTS: For diagnosis years 2001-2005 combined, valid information on census tract was submitted for 59% of the records in the NPCR dataset. Of the 41 NPCR-supported registries evaluated, 22 (54%) submitted valid census tract data on > = 90% of total incidence records, and 10 (24%) submitted valid census tract data on < = 5% of the records. Across individual diagnosis years, 53%, 59%, 60%, 60% and 61% of records included valid information for 2001, 2002, 2003, 2004, and 2005, respectively. CONCLUSION/IMPLICATIONS: The completeness of census tract data varied by registry and diagnosis year. The differences in completeness of census tract data can reflect changes in the number of registries that submit such data and changes in the number of registries that meet data quality standards. In addition, there may be resource issues or other un-identified barriers for not being able to submit these data. Complete data on census tract from all registries is desirable for national level research related to spatial analyses, although information from registries with complete census tract data may be usable for local, state, or regional studies or national estimates.