2275

Background: Epidemiological studies of spatially varying exposures benefit from accurate long-term comprehensive residential address histories of study participants in order to accurately link locations to spatially and temporally resolved exposure information. Commercially available residential address history databases might help researchers establish historical address data retrospective to study onset. The Ginkgo Evaluation of Memory Study (GEMS) was a randomized trial in four US sites (Hagerstown MD, Pittsburgh PA, Sacramento CA, and Winston-Salem NC) from 2000-2008. We aimed to assess the accuracy and historical reach of LexisNexis residential histories for the 3,069 GEMS participants from 1980 to 2008. Methods: We submitted to LexisNexis a list of 3069 participant names and 4050 known addresses and dates of residency from 2000-2008. The query returned 10,966 addresses. We developed a set of specific rules along with a fuzzy matching algorithm, to establish a sequence of probable residential histories. The average length of histories per participant were then evaluated and compared across gender, study site, and age. Results: Almost all (97.8%) GEMS participants had at least a partial address history returned by LexisNexis for years prior to 2000. The earliest historical addresses had a median year of 1983 and 35.2% of participants had their earliest historical address going back to at least 1980. The biggest differences in average historical lengths were between study sites. Participants from Hagerstown had an average of between 2.9 and 6.0 fewer years of residential history than the other three sites. Additionally, males had between .56 and 1.23 more years on average than females. Conclusions: Commercial residential address systems have the potential to enhance health studies that require historical residential information. As these types of databases grow in accessibility and accuracy, historical databases can be a key tool of environmental epidemiologists. Demographics and geographic locations, however, could impact accuracy and address length.