PS3-22: HMORN VDW Mortality QA: Evaluating Completeness and Consistency of Death and Cause of Death Data

  • September 2013,
  • 178.2;
  • DOI: https://doi.org/10.3121/cmr.2013.1176.ps3-22

Abstract

Background/Aims Patient mortality data is a critical component of health research data, especially where death is a clinical outcome of interest. Mortality data is also important for cohort selection and to avoid communications with families of deceased patients. The HMORN Virtual Data Warehouse (VDW) has a DEATH file and a CAUSE OF DEATH file that contain data on whether a patient is dead or alive, date of death, confidence of death status, and causes of death. VDW mortality data is synthesized from multiple sources, such as state death certificates, federal data, and health plan utilization and membership data. These sources are combined using complex, probabilistic algorithms. There is variation in both source data and the transformation logic used by sites. This poses potential challenges for researchers to derive accurate results from multi-site studies. Our aim is to measure the quality of VDW mortality data across HMORN sites, highlight areas of consistency and variation, report issues for remediation, and identify opportunities to improve data specifications and guidelines.

Methods A multi-site QA program was distributed to all HMORN sites by the VDW Mortality Ad Hoc Workgroup. Results were returned to the workgroup, which combined, analyzed, and reported the results.

Results 14 sites returned results for DEATH and 12 sites for CAUSE OF DEATH. The DEATH file had good overall data quality: 3 sites had minor issues related to missing values or duplicates. Records at 11 sites were checked for post-death patient activity and all sites scored 93% or better. There was significant variation in the CONFIDENCE and SOURCE variables, indicating potential areas for improvements to the specifications. The CAUSE OF DEATH file had mixed quality: 7 sites had issues related to missing values and disagreement with the DEATH file, but there was good agreement on the most common causes of death. Additional research is needed to understand the sources of variation in the data and to recommend specification improvements.

Conclusions The VDW DEATH and CAUSE OF DEATH files provide quality mortality data that satisfy a number of uses for multi-site studies. Improving the quality of VDW mortality data will require sustained effort by all HMORN sites.

Loading
  • Share
  • Bookmark this Article

Keywords