Validating tumor linkage using the NAACCR site pairs table

The objective of this study is to illustrate use of the Site Pairs Table developed by the North American Association of Central Cancer Registries (NAACCR) Record Linkage Work Group to validate tumor linkage in a central registry database and to identify potential cases with inaccurate tumor linkage. Central registries often receive reports for patients with multiple tumors, and they receive multiple reports from different sources for the same tumor. Tumor site pairs (pairs of unique tumors for patients with multiple tumors) ought not refer to the same tumor as represented in the Site Pairs Table. Likewise, abstract pairs (pairs of abstracts relating to the same tumor) ought to be identified during the tumor linkage process as belonging to the same tumor. Three central cancer registries represented on the work group contributed data to the study. The data included cases diagnosed 1992-2003 and represented 143,288 patients with multiple tumors and 280,227 tumors with multiple abstracts. Totals of 181,118 tumor site pairs and 391,670 abstract site pairs were generated from the data and compared to the Site Pairs Table. Of the abstract site pairs 381,389 (97.4%) were found in the Site Pairs Table. One registry reviewed its portion of the 2.6% not found in the table and determined 12% of the cases were incorrectly linked and should change from one tumor to two tumors. Of the tumor site pairs, 144,793 (80%) were not found in the Site Pairs Table. Further evaluation of the remaining 20% by paired site and laterality, histology and timing showed 19.3% were considered unique tumors and 0.7% were identified as potential cases with inaccurate tumor linkage. Two registries reviewed their portion of these cases. One registry changed two tumors to one tumor on 44% of the cases they reviewed. The other registry changed two tumors to one tumor on 53% of the cases they reviewed. Analyzing site pairs within the registry database using the Site Pairs Table assists in identifying inaccurate tumor linkages as was shown in this study.