Development and Validation of a High-Quality Composite Real-World Mortality Endpoint

Area of Science:

• Health Informatics
• Oncology Research
• Real-World Evidence (RWE)

Background:

• Electronic Health Records (EHRs) are crucial for real-world evidence (RWE) generation.
• Accurate mortality data is essential for reliable RWE, particularly in oncology.
• Existing EHR data often has incomplete mortality information.

Purpose of the Study:

• To develop a high-quality, linkable mortality dataset derived from EHR data.
• To supplement EHR data with external sources for improved mortality ascertainment.
• To benchmark the quality of the developed mortality dataset against the National Death Index (NDI).

Main Methods:

• Amalgamated oncology EHR data with external commercial and US Social Security Death Index data.
• Developed and validated a composite mortality variable (version 2.0).
• Benchmarked the composite mortality variable against the National Death Index (NDI) for completeness and date accuracy.

Main Results:

• Mortality data sensitivity improved from 66% in structured EHR data to 91% in the composite dataset for advanced non-small-cell lung cancer.
• Sensitivity ranged from 85% to 88% for advanced melanoma, metastatic colorectal cancer, and metastatic breast cancer.
• Improved mortality data completeness minimized overestimation of survival compared to NDI-based estimates.

Conclusions:

• High-quality EHR-derived data is necessary for reliable RWE generation.
• Mortality data completeness significantly impacts survival endpoint accuracy.
• Benchmarking mortality data against the NDI is crucial for data quality assessment.