1348798 - The Social Security Death Index is No Longer Good Measure of All-Cause Mortality in CMR Patients

Background: Early epidemiological studies established the accuracy of the Social Security Death Index (SSDI) for determining all-cause mortality with accuracies ranging from 80-96% (1). Since 2016, several CMR studies have used the SSDI as the sole determinant of all-cause mortality to assess the prognostic value of vasodilator stress (2), lateral mitral annular plane systolic excursion (MAPSE) (3), scar quantification (4), and left atrial volume index (5).  However, in 2011, the Social Security Administration stopped including state death records in its Death Master File (6). As a result, it contains many fewer death records. The purpose of this study is to re-assess the accuracy of SSDI in a CMR population.

Methods:

This single center study comprises 5,740 patients who had a CMR exam between 2001 and 2012. Date of death was determined in 2022 by two methods: 1) SSDI: searching the patient’s name, social security number, and date of birth, and 2) internet: searching the patient’s name, date of birth, residential address at the time of CMR, and the name of a first degree relative, if available.   For the internet search, a point system was created to determine if the publicly available information (usually an obituary) conclusively matched the patient (Figure 1).

Results:

SSDI detected only 32% of total deaths (367 vs. 1153). Figure 2A shows a progressive decrease in % all-cause mortality as follow-up time decreases, consistent with patients and their dates of death being correctly identified. Figure 2B shows that the SSDI reported a higher percentage of deaths in the earlier scan years, and Figure 2C shows that the average survival duration of patients detected by SSDI is relatively constant regardless of scan year. Both findings are consistent with the SSDI biased to detecting earlier deaths more than later deaths. Figure 3 shows survival curves from scan years 2001 and 2012, showing SSDI underestimates all-cause mortality 2-4 fold, depending on the scan year.  Only for 2001, does the SSDI match the total deaths for the first several years, consistent with high SSDI accuracy for deaths prior to 2011.  

Conclusion:

The SSDI severely underestimates all-cause mortality in the CMR population and may skew outcomes in studies relying on the SSDI. The 4-fold underestimation found in this study in 2012 is consistent with estimates from private industry (7). Because the accuracy of the SSDI varies by year, comparisons between contemporary studies and earlier ones are problematic. Part of SSDI’s popularity is its relatively low cost and ease of access. However, by missing so many deaths, studies using SSDI may require much larger patient numbers, which potentially increases cost and time.  An internet search can substantially increase the number of deaths detected although it is a more time-consuming process.

 

1.