Predictors of LexisNexis residential history availability and registry data concordance for childhood cancer research.

Background: Use of a commercial database to obtain residential history information in environmental epidemiologic studies of cancer can lead to information bias if data availability varies by individual socio-demographic factors or case status. Residential data that is not missing at random and data that is discordant with cancer registry or birth record address data can impact subsequent exposure assessments. In our study of childhood cancers, we aimed to determine if availability of residential history information differs by case status or other potential confounders and if there was agreement with cancer registry and birth records address data.

Methods: We worked with LexisNexis to retrieve residential histories for mothers of 3,573 childhood cancer cases and 7,160 controls born 2000-2015 in Los Angeles and Orange Counties in Southern California. We used linear regression to determine independent predictors of having residential history returned by LexisNexis. We assessed concordance between maternal address at birth and child's address at diagnosis available from registry data and the LexisNexis residential history by comparing street addresses and geocoded coordinates.

Results: Maternal characteristics (birthplace, race and ethnicity, education, insurance provider) and child's case status were associated with the mother having any address returned by LexisNexis. When comparing geocoded coordinates of cases, < 10% of LexisNexis addresses during the diagnosis year matched cancer registry addresses. Birth record addresses matched LexisNexis-provided addresses for 47% of mothers.

Conclusions: This study elucidates potential implications of using commercial databases such as LexisNexis to reconstruct residential histories and derive exposure measures in cancer case-control studies.