Background: Cancer registries are used to report cancer care trends and outcomes. Information from these data sets is utilized to craft practice guidelines and management recommendations. Limited knowledge is available regarding the quality of the data contained within registries. We sought to determine the accuracy of a single variable, 'surgery of the primary site', in the Tennessee Cancer Registry (TCR).
Methods: A retrospective review of the TCR thyroid database was performed. Hospital facilities were classified as either Commission on Cancer (CoC) or non-CoC accredited. Certified Tumor Registrars at the TCR reviewed the abstracted text and/or telephoned the reporting facility staff to confirm the definitive thyroid procedure.
Results: A total of 921 thyroid cancer cases, diagnosed/treated at TN facilities during 2004-2011, were coded with thyroid lobectomy (TL). Overall, 369 (40 %) were incorrectly coded, of which 247(67 %) were changed to total thyroidectomy. The majority of cases (80 %) were reported by CoC facilities. When compared by facility type, 42 % of records submitted from CoC facilities contained incorrect codes for the variable 'surgery of the primary site' TL compared with 34 % of records submitted by non-CoC facilities (p = 0.047).
Conclusion: In this study of the TCR, 40 % of records contained inaccurate coding of the variable 'surgery of the primary site'. Upon validation, 27 % of all records were changed from TL to total thyroidectomy. The rate of incorrect coding was higher in CoC reporting facilities than in non-CoC facilities. Using text-to-code re-abstraction audits and facility contact these discrepancies can be validated and corrected to improve data quality.