How do missing deaths in real-world data impact overall survival analyses?

Carrigan G, Whipple S, Taylor MD, Torres AZ, Gossai A, Arnieri B, Tucker MG, Hofmeister PP, Lambert P, Griffith SD, Capra WB. An evaluation of the impact of missing deaths on overall survival analyses of advanced non-small cell lung cancer patients conducted in an electronic health records database. Pharmacoepidemiol Drug Saf. 2019;28(5):572-581

Our summary

Researchers from Roche and Flatiron Health performed a deep dive into evaluating the impact of missingness in our mortality dataset. The findings, using National Death Index data up to the end of 2015, suggest that impact on survival analyses is minimal when sensitivity (completeness) of mortality data is high.

This is a follow-up publication to the May 2018 Health Services Research publication which detailed the compilation of our mortality dataset and subsequent benchmarking to the National Death Index, which is considered the national gold standard for death data. This validation based on 2015 data found that Flatiron’s aNSCLC data had a sensitivity of 90 percent.

Interested in working with Flatiron data?

Contact our Life Sciences Partnerships team to get started.

arrow

Sales CTA Form

Why this matters

The quality of real-world evidence in oncology depends on the robustness and reliability of endpoints such as overall survival. This study shows that the highly sensitive Flatiron composite mortality variable can be applied to EHR-derived datasets to yield overall survival estimates that are largely accurate, and subject to negligible bias.

Read the research

Related articles