|
BACKGROUND/RATIONALE:
The BIRLS Death File is the main internal source utilized by VA researchers to ascertain deaths in their cohorts. Often, VA researchers use other sources to augment BIRLS death data due to their concerns regarding the completeness and accuracy of BIRLS data. Based on interest within the VA research community to acquire death dates from alternative sources including the National Death Index (NDI), previous research examining the sensitivity of BIRLS death dates, and our recent comparison of death dates in the BIRLS Death File to death dates provided by Medicare, this feasibility study was conducted to more accurately determine the costs and benefits associated with use of alternative sources for death ascertainment (e.g., NDI) and to develop a strategy for the possible development of a central database or registry of death data that can be used by VA researchers for death ascertainment.
OBJECTIVE(S):
1. Analyze and compare the following sources for death data: Medicare Files, SSA Death Master File, BIRLS Death File and PTF.
2. Determine the benefit of using the NDI to identify additional deaths and validate deaths in other sources.
3. Develop a methodology to use these sources most cost effectively for death ascertainment in research.
4. Propose a strategy to maintain a database of death date information and provide cost estimates for annual maintenance and updates.
METHODS:
A finder file of veterans “known to the VA” was created. Dates of deaths for these veterans were obtained from the death data available to the VA: Medicare Files, SSA Death Master File, BIRLS Death File and PTF. These dates were compared for consistency. A stratified random samples of the SSNs with a death dates were sent to the NDI to validate the dates of death. Veterans without a date of death were classified as either presumed living or unknown status based on their recent Medicare utilization, VHA utilization and receipt of Compensation and Pension Benefits. Stratified random samples of the presumed living and unknown status veterans were sent to the SSA for a presumed living search and the NDI to validate their status and identify deaths not recorded in death data available to the VA. The results of these searches were analyzed to validate the quality of the death date sources and assess the cost effectiveness of augmenting these sources with the NDI. Based on the results of the analysis, a methodology for death ascertainment for veterans using data available within the VA is proposed as well as the possible development of a Veterans Death Registry.
FINDINGS/RESULTS:
The pooled use of the Medicare Files, SSA Death Master File, BIRLS Death File and PTF to ascertain death compared very favorably with the NDI in all strata except those veterans over 65 classified with an unknown status for whom we had no Medicare data (only 0.3% of the total study population of 8.6 million). Using the NDI as the gold standard, we calculated the sensitivity and specificity of using the four sources available within the VA to identify deaths in a random sample of 3,000 veterans. The sensitivity and specificity were very high, 98.3% and 99.8% respectively. Our method to categorize a veteran as presumed living was validated by the NDI and SSA searches. No deaths were identified in the presumed living sample and the SSA confirmed over 90% as presumed living.
We conclude from our results that using the NDI to enhance death ascertainment would not be cost effective. A death registry could be built using sources already available within the VA. Based on these results, we recommend that (1) a death registry be built with data already available within the VA and (2) information be added to the registry that could be used to generate a “presumed living” status.
IMPACT:
This study will help the VA Research Community make an informed decision about the best way to address their concerns about the accuracy and completeness of BIRLS death information. Development of a central Veteran’s Registry of Death Information could greatly ease a researcher’s task of identifying cohort deaths and provide a standard methodology for such death identification reducing the possibility of measurement error in VA research.
PUBLICATIONS:
None at this time.
DRA:
Health Services and Systems
DRE:
none
Keywords:
Organizational issues
MeSH Terms:
none
|
