Publications : 2018

Mumma MT, Cohen SS, Sirko JL, Ellis ED, Boice JD Jr. 2018. Obtaining vital status and cause of death on a million persons. Int J Radiat Biol 9:1–21.

Abstract

Purpose: To present the methodology used to determine vital status and obtain cause of death (COD) within the Million Person Study of Low-Dose Health Effects (MPS). Data sources and vital status tracing techniques used to obtain vital status and COD for six (n = 424,238 subjects) of the ∼20+ cohorts under study are described.

Methods and materials: A multistage approach using multiple sources of vital status information was used to determine vital status (or ‘trace’) study participants from as early as 1940 to the present. Mortality records from state departments of vital statistics and the Social Security Administration Death Master File (SSA-DMF) were matched to study participants by Social Security Number (SSN), full name, date of birth (DOB), and/or sex using deterministic and probabilistic algorithms. The National Death Index (NDI) and SSA Service for Epidemiological Researchers (SSA-SER) were used to obtain COD (after 1978) and verification of alive status, respectively. Online public records and ancestry services, death certificates, and specialized mortality sources were also utilized.

Results: For the MPS cohorts traced to date (nuclear power plant workers, industrial radiographers, atomic veterans, and workers at Rocketdyne/Atomics International, Mound nuclear facility, and Mallinckrodt Chemical Works), vital status was confirmed for over 90% of all study subjects in all but one cohort (88%). The ascertainment of COD was over 96% for all cohorts.

Conclusions: A hallmark of a high-quality epidemiologic cohort mortality study is a low percentage of subjects with unknown vital status and a low percentage of deaths without a COD. The sources and methods used for vital status tracing and COD determination for the MPS have been successful and should be useful for other investigators tracing large, historic study populations. Some of the approaches would be applicable for use in all cohort studies using regional-specific mortality data or modifications to the approach.