Improving survival models in healthcare
Author Information
Author(s): Bertsimas Dimitris, Ning Catherine, Lønning Per Eystein, Baba Hideo, Endo Itaru, Burkhart Richard, Aucejo Federico N, Balzer Felix, Kreis Martin E, Margonis Georgios Antonios
Hypothesis
Can a novel matching approach enhance the prognostic power of Cox regression models in survival analysis?
Conclusion
The new matching approach significantly improves the performance of survival models compared to traditional methods.
Supporting Evidence
- The new approach improved bootstrapped C-indices by up to 20% compared to traditional models.
- Model 3 showed a 6- to 10-fold improvement in calibration over model 1.
- Robustness of the findings was confirmed through bias-corrected bootstrapping.
Takeaway
This study found a new way to make predictions about patient survival that works better than the old methods.
Methodology
A two-stage mechanism was used to select representative patient observations for model training.
Participant Demographics
Patients with resected colorectal cancer liver metastases, including 1,197 who received chemotherapy and 602 who did not.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website