Death data
Death data in hospital records
Epic data model captures patient death information in two table locations. First is the "Known decedent" (Tables such as EPT 115) and second is the "Possibly decedent" (Tables such as EPT 34205). The first is populated when patients die during their course of care at Stanford hospitals, or Stanford Autopsy Service is used. In a small proportion of cases, hospital staff may populate the "known decedent" tables e.g., when they become aware of death during their course of an outreach. The second, "possibly decedent" at Stanford is populated from various sources. It could potentially come from third party who support hospital billing services. Alternatively, these could be passed on to us from collaborating sites e.g., Sutter Health and Stanford collaboration on Oncology services. SHC started using "possibly decedent" tables since 2021. At the time of writing, Epic only supports "replacement" mode in the "possibly decedent" tables, a later entry can wipe out a former entry.
Use of "possibly decedent" is challenging. Often the provenance of this data is not well captured in source systems and the patient matching logic may be "fuzzy" at best.
Social Security Administration Death Data
STARR includes SSA death data, specifically Limited Access Death Master File (LADMF). Research IT receives updated LADMF data on a monthly basis. To learn details of SSA LADMF and STARR linkage, please refer to our manuscript.
In summary, SSA has three sources of death data, a public Death Master File (DMF), the Limited Access Death Master File (LADMF) and the full death file. SSA compiles files of death information from many sources, including close relatives and representative payees, funeral homes, financial institutions, postal authorities, States, and other Federal agencies (e.g., Department of Veterans Affairs, Centers for Medicare and Medicaid Services) and added about 2.9 million new death reports to its records in 2019. About ~78% of these death reports came from the states through a system called Electronic Death Registration (EDR) system. Using EDR, state vital records agencies first verify the decedent’s name and Social Security number (SSN) against SSA’s records and then submit a verified electronic death record that is automatically posted to SSA’s Master Files of SSN Holders and SSN Applications, also known as the Numident. Death reports through EDR are timely (95% within 30 days) and highly accurate and are used across SSA’s systems to automatically stop Social Security and SSI benefit payments to deceased individuals. Unfortunately, the LADMF does not contain state death records and only contains only those death records obtained from non-state sources (i.e., obtained from close relatives, representative payees, other federal agencies, other state agencies such as state welfare offices, friends, and neighbors).
The full death data file is only accessible to certain federal benefit-paying agencies such as the Department of Defense, Department of Veterans Affairs, Department of Agriculture, Department of Housing and Urban Development, Centers for Medicare and Medicaid Services, Internal Revenue Service, and starting 2020, to National Institutes of Health for research and statistical purposes. The public DMF, which, in addition to excluding death data from state vital statistics bureaus, also excludes records for individuals who died within the last three years. The LADMF is thus somewhat larger than the public DMF but still considerably smaller than SSA’s full file of death information. SSA does not guarantee the accuracy of the DMF file. Firstly, the SSA does not have a death record for all deceased persons. Secondly, in rare instances, it is possible for the records of a person who is not deceased to be included erroneously in the DMF.
A research study, titled "Alive or dead: Validity of the Social Security Administration Death Master File after 2011 (link)" concluded that researchers using the DMF may underestimate mortality. Here is a summary of changes to SSA death data explaining the change:
- SSA created the public DMF in 1980 in response to a 1978 Freedom of Information Act lawsuit.
- In 2002, SSA began disclosing state EDR records on the public DMF.
- In 2010/2011, SSA began examining their disclosure of protected state records. Section 205(r) of the Social Security Act prohibits SSA from disclosing State death records.
- Beginning Nov. 1, 2011, SSA decided it would withdraw state death records from the public file.
Finally, this LADMF source is regulated and imposes some additional restrictions on how the data is used downstream. Stanford University has signed a subscriber and associated licensing agreements. LADMF regulations place certain restrictions on the use and disclosure of the date of death (including individual data elements, such as month, day and year of death) of individuals during the three-calendar-year period beginning on the date of the individual’s death. Fact of death is not regulated. If there is independent death information from hospital source, confirming the date of death, the individual's data is not regulated under LADMF. The LADMF regulations are independent of HIPAA regulations and persist even in fully de-identified data. The LADMF regulated data can't go to a collaborator who does not have an independent LADMF subscription.
If you have any further questions on the LADMF source, please request a consultation service.
STARR Tools and death data
STARR Tools had historically provided SSA death data information to augment the in-hospital death information ("known decedent") available via EHRs. Historically, the two sources were combined into a single field in STARR Tools. Since 2021, STARR Tools, separates the two sources - the "known decedent" coming from Epic nd the LADMF coming from a "fuzzy" matching logic from LADMF source file. We recommend that the interested reader peruse our manuscript to seek a better appreciation of the fuzzy matching challenges.
In addition to the challenges posed by the incompleteness of death information, and inaccuracy posed by fuzzy matching, the LADMF data is highly regulated and researchers too must follow the associated regulations. Inappropriate use are subject to penalties under federal regulations (15 CFR 1110.200).
Researcher access to high-quality death data
Death is a critical outcome for a large number of research endeavors. If the hospital data and LADMF are incomplete and potentially inaccurate, one could potentially ask what are other viable sources for researchers. The gold standard is National Death Index (NDI). Access to NDI requires an eligible IRB approved study to file an application. The applicaiton process is rigorous and is both time consuming and expensive. For NIH funded studies, that require NDI data, the NDI access fee is covered by NIH.
Research IT is currently working with hospital administrators to integrate alternate high-quality sources. There is no specific timeline available.