Presentation (CTI)

OBJECTIVES: High-quality mortality data is essential for accurate research and improved patient outcomes. Mortality data can be obtained from a variety of sources but no single source has been found to be sufficiently robust for analysis. This study shows that while no single source reaches a desired level of accuracy, completeness, or timeliness for analytical action, a thoughtfully crafted composite can achieve research-grade quality by supporting corroboration, backfilling, and auto-correction of underlying data. More importantly a well-constructed composite can adhere to FAIR (Findable, Accessible, Interoperable, and Reusable) principles.
METHODS: Our study uses nine unique source categories (comprising internment, government and obituary source types) to show how each source is improved incrementally as more sources are added. For each baseline source we determine the next best source to add that would improve the baseline scores the most along three dimensions:

Timeliness: Median days between death date and death record capture, termed “capture lag”.

Completeness: Percentage of records containing first name, last name, date of birth, and date of death.

Accuracy: Level of data agreement with state-level government records.

RESULTS: Composite mortality datasets increased completeness of single-source, incomplete records by 41.3%, with gains observed across all contributing sources. Timeliness improved significantly: median capture lag decreased from 7 to 4 days, and 92.5% of records were captured within 14 days - a 10% increase over single-source reporting. Accuracy also improved: among records initially incorrect using a single source, 25% were corrected after incorporating additional sources.
CONCLUSIONS: A composite mortality database built from a variety of sources greatly enhances accuracy, completeness and timeliness of the underlying data. By combining multiple sources, data becomes inherently more robust and trustworthy and is much better suited for analytical action.