Background: Studies linking large pregnancy cohorts with mortality data can address critical questions about long-term implications of gravid health, yet relevant US data are scant. We examined the feasibility of linking the Collaborative Perinatal Project, a large multiracial U.S. cohort study of pregnant women (n = 48,197; 1959-1966), to death records.
Methods: We abstracted essential National Death Index (NDI) (1979-2016) (n = 46,428). We performed a linkage to the Social Security Administration Death Master File through 2016 (n = 46,450). Genealogists manually searched vital status in 2016 for a random sample of women (n = 1,249). We conducted agreement analyses for women with abstracted data among the three sources. As proof of concept, we calculated adjusted associations between mortality and smoking and other sociodemographic factors using Cox proportional hazards regression.
Results: We successfully abstracted identifying information for most of the cohort (97%). National Death Index identified the greatest proportion of participants deceased (35%), followed by genealogists (31%) and Death Master File (23%). Estimates of agreement (κ [95% confidence interval]) between National Death Index and Death Master File were lower (0.52 [0.51, 0.53]) than for National Death Index and genealogist (0.66 [0.61, 0.70]). As expected, compared with nonsmokers, smoking ≥1 pack per day was associated with elevated mortality for all vital sources and was strongest for National Death Index.
Conclusions: Linking this historic cohort with mortality records was feasible and agreed reasonably on vital status when compared with other data sources. Such linkage enables future examination of pregnancy conditions in relation to mortality in a diverse U.S. cohort.