Abstract 6346A: Privacy preserving record linkage of mortality data for oncology survival analyses

Abstract

Abstract Though survival is an essential endpoint in cancer research, mortality data is often incomplete in cancer registries and RWE sources used in cancer research, including medical claims data and electronic health records. Government mortality data has inadequate coverage, with single sources like the SSA being reappraised for completeness. To overcome this problem, RWE mortality composites have been developed using obituary data sourced from funeral homes, newspapers, & other online obituary sources, but no single source has complete coverage. We explored a privacy preserving record linkage strategy using a set of commercial mortality composites and governmental data sources to improve population coverage. Tokenization technology was used to generate a high coverage, deduplicated mortality data resource from disparate datasets, including a government source, a private claims source, two digital obituary sources, and a private media source. The CDC-reported number of total deaths were used as a benchmark, and we compared each of the five mortality data sources against this number to understand individual coverage. We then assessed coverage when these data sources were linked and de-duplicated to form a single dataset. We demonstrate that these mortality datasets captured over 90% of deaths in the United States over the previous four years. The de-duplicated incorporation of multiple sources is required for accurate real-world overall survival analyses. Through tokenization technology, the mortality signal can be linked to other datasets to produce longitudinal survival curves. Private Claims is a commercially available dataset sourced from medical claims. Obituary 1 & 2 are two distinct, commercially available datasets sourced from funeral homes, newspapers, and online obituary sources. Government is SSDI. Private Media is a commercially available dataset sourced from newspapers, county death records, and classifieds. Table: Sources Combined Compared to CDC Counts – Coverage by Month: Month Obituary 1+Government+Obituary. 2 Obituary. 1+Private Claims+Government Obituary 1+Government Obituary 2+Government Obituary 1+Government+Obituary 2+Private Claims Government+Private Claims Government+Private Claims+Obituary 2 Private Media+Government+Obituary. 1 Private Media+Government+Obituary 2+Private Claims Private Media+Government+Obituary 1 1+Obituary 2 All Feb 2021 ~85% ~60% ~50% ~80% >90% ~30% ~85% ~70% ~75% >90% >90% Feb 2020 ~85% ~70% ~60% ~85% >90% ~45% >90% ~80% >90% >90% >90% Feb 2019 >90% ~70% ~55% ~85% >90% ~45% >90% ~85% >90% >90% >90% Feb 2018 >90% >90% >90% >90% >90% ~50% >90% >90% >90% >90% >90% Citation Format: Devin Gilliam, Christine Horne, Jacob Kean, Vera Mucaj. Privacy preserving record linkage of mortality data for oncology survival analyses [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 6346A.