Methods/Statistics
Concordance of LexisNexis Accurint address information with address at diagnosis recorded in a population-based registry Rebecca Nash* Rebecca Nash Lauren E. McCullough Deirdre Cronin-Fenton Joellen Schildkraut Kevin C. Ward Timothy L. Lash
Background: The LexisNexis Accurint database is a commercially available source of residential data that has been used to obtain follow-up information, including emigration status, for disease registries and cohort studies. However, accuracy of Accurint address information has not been examined among adults in a population-based registry.
Methods: We assessed concordance between address at diagnosis recorded in the Georgia Cancer Registry (GCR) and an address identified in Accurint to cover a period that included diagnosis date among a cohort of breast cancer survivors (N=25,730). Concordance was assessed by text string matching of city and state, allowing for up to two-character differences in city name. To account for Accurint providing addresses with overlapping coverage periods, we assessed concordance by selecting a single address using different selection criteria: (1) longest duration, (2) most recent end date, and (3) prioritizing better matches. Concordance was quantified overall and by race and ethnicity.
Results: LexisNexis provided a total of 51,447 address records that covered a period including the diagnosis date among 25,323 (98%) breast cancer patients. Two-thirds of these records (66%) matched the city and state of the address recorded in the GCR, 22% matched the state only and 11% were non-matches. Most patients (95%) had at least one record that matched both the city and state; 30% had at least one address that matched on state only and 16% had an address that did not match the GCR-recorded address. Concordance on city and state ranged from 75% (longest duration) to 95% (prioritizing better matches). American Indian and non-Hispanic White patients had the highest and Asian and Pacific Islander patients had the lowest concordance for all three methods.
Conclusion: Concordance between addresses recorded in GCR and Accurint was moderate among a cohort of breast cancer survivors. There were differences by selection criteria and race and ethnicity.