1 paper2 findings
Event-free survival
upin humans1
Patient-history embeddings predict event-free survival in humans (AUC 0.726 versus 0.722 for the comorbidity model).
“The embedding model achieved an AUC of 0.726 versus 0.722 for the comorbidity model.”
HumansRemaining alive without accumulating a second unrecorded ICD-10 disease block
Patients’ diagnosis histories modestly improved disease forecasts
NPJ digital medicine14 Oct 2026
Incident ICD-10 disease-block outcomes
upin humans1
Patient-history embeddings predict incident ICD-10 disease-block outcomes in humans (Improved prediction for 93 of 131 outcomes; median AUC gain 0.006).
“embeddings improved prediction over the Elixhauser-based comorbidity model for 93 of 131 incident ICD-10 disease-block outcomes, with a modest median AUC gain of 0.006.”
HumansPrediction compared with the Elixhauser-based comorbidity model
Patients’ diagnosis histories modestly improved disease forecasts
NPJ digital medicine14 Oct 2026
The model improved prediction for 93 of 131 disease groups compared with a model based on existing disease burden.The model improved prediction for 93 of 131 disease groups compared with a model based on existing disease burden.