P22: Development of a Composite Mortality Score Addresses Racial Bias in Mortality Data Reporting
Poster Presenter
Benjamin Holmes
Senior Clinical Data Analyst
Syapse United States
Objectives
1. Articulate the importance of complete mortality data to support oncology outcomes research
2. Evaluate the impact of a composite mortality score in overcoming racial bias associated with individual data sources
Method
This study developed a composite mortality score integrating data cross six real-world sources. Accuracy and completeness for individual sources and the composite score were evaluated vs. the National Death Index (NDI; gold-standard source).
Results
The study included data from 95,856 unique patients. Overall, the sensitivity for vital status across individual data sources ranged from 76.8% (EHR) to 98.7% (SEER). Specificity ranged from 74.5% (SSDI/DMF) to 97.8% (SEER). The composite mortality score achieved sensitivity of 94.8% and specificity of 93.2%.
Compared with white patients, black patients were less likely to have accurate death data captured in the digitized obituaries and tumor registries (all p<0.05). However, after development of a composite score there were no significant differences in mortality data capture by race (p=0.71).
Conclusion
A composite score for mortality that maintains strong sensitivity, specificity, and concordance was developed leveraging scalable mortality data sources available in the real-world setting. To the authors’ knowledge, this is the highest level of sensitivity and concordance that has been attained to date, within a large-scale, multi-purpose real-world oncology dataset. This composite score was further able to overcome biases observed in mortality data coverage.
