A geographical information system technique for record-matching in a studyof cancer deaths in welders

Retrospective studies are frequently complicated by incomplete worker ident ifiers. We encountered this problem when evaluating the risk of cancer deat h in a welding cohort. We dealt with it by developing birth-date ranges for each welder with unknown birth dates and using geographical information sy stem techniques in conjunction with last name, gender; and birth-date range to assign death certificates to welders on the basis of residential proxim ity to the worksite. Deaths for total malignant cancers and lung/tracheobro nchial/pleural cancers among these welders were not significantly different from those in county, state, and US populations, using standardized mortal ity ratios. The ratios in our study subjects were consistent with ratios So und in other published welder cohorts.