PREDICTING MORTALITY IN ADULTS USING TRANSFORMERS: INSIGHTS FROM THE HEALTH AND RETIREMENT STUDY
2024

Predicting Mortality in Adults Using Transformers

publication Evidence: high

Author Information

Author(s): Weiss Jordan, Azhir Alaleh, Ram Nilam, Rehkopf David

Primary Institution: Stanford University

Hypothesis

Can a natural language processing model predict mortality in adults based on their demographic and health histories?

Conclusion

The NLP-based prediction model outperforms traditional statistical models in predicting mortality.

Supporting Evidence

  • The model achieved an average precision score of 0.900, significantly higher than the 0.395 score of conventional statistical models.
  • The study opens new avenues for personalized healthcare in observational settings.

Takeaway

This study uses a computer program to guess who might die next based on people's life stories and health information, and it does a better job than older methods.

Methodology

A transformer architecture was used to create embeddings from respondents' demographic, social, behavioral, and health histories to predict mortality.

Participant Demographics

Respondents from the US-based Health and Retirement Study.

Digital Object Identifier (DOI)

10.1093/geroni/igae098.0559

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication