Concordance of Death Information Between Two Health Systems Serving the Same Region and the Social Security Administration Death Master File: Retrospective Observational Study.

Background: There are multiple sources that document the date of a person's death in the United States. Unfortunately, this seemingly simple information is either incomplete or costly to obtain. Nevertheless, information on such events is critical for both health systems and clinical studies to assess the outcomes of operational and therapeutic interventions. Objective: As part of a larger assessment of the quality of multisource death information, we compared the death data from two health systems serving the same region and the Social Security Administration Death Master File (SSADMF). Methods: This study linked death records for patients seen in either health system with the SSADMF from 2007 to 2020 to identify concordant and discordant death data among sources. Analyses included cross-system matching, classification of death records by overlap, and calculation of agreement using the Fleiss κ. Results: Among 904,581 matched patients, only 209 (0.02%) deaths were confirmed by all 3 sources. Large proportions of deaths were uniquely recorded by a single source: 54.32% (10,697/19,691) by health system A, 5.16% (1017/19,691) by health system B, and 20.17% (3972/19,691) by the SSADMF. The Fleiss κ was negative (-0.312), reflecting less agreement than expected by chance. Conclusions: While this study is not generalizable, it showed that, without processes in place to obtain external data regarding patient deaths, health care facility death information should not be relied upon as a complete list of those who have died. The discordances observed highlight the potential for significant gaps in death reporting within health care systems, which could impact the accuracy of mortality-based analyses and quality assessments.