ASCENDgpt: A Phenotype-Aware Transformer Model for Cardiovascular Risk Prediction from Electronic Health Records

2025-08-31

AI Summary

This study addresses the challenge of multi-endpoint cardiovascular risk prediction from longitudinal electronic health records (EHRs). To this end, we propose a phenotype-aware Transformer model. Methodologically, we design a clinically interpretable phenotype tokenization scheme that compresses tens of thousands of diagnosis codes into 176 semantically cohesive phenotype tokens—reducing vocabulary size by 77.9%. The model integrates masked language modeling pretraining with time-to-event fine-tuning to jointly predict five time-to-event endpoints: myocardial infarction, stroke, major adverse cardiovascular events (MACE), cardiovascular death, and all-cause death. On an independent test set, the model achieves a mean C-index of 0.816 (0.842 for cardiovascular death), significantly improving predictive accuracy, clinical interpretability, and computational efficiency. This work establishes a novel paradigm for dynamic, EHR-driven cardiovascular risk assessment.

Technology Category

Machine Learning: Hardware-aware MLComputer Vision: Interpretability, Explainability, and TransparencyNatural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models

Application Category

User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalizationSemantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMsEconomics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI

Abstract

We present ASCENDgpt, a transformer-based model specifically designed for cardiovascular risk prediction from longitudinal electronic health records (EHRs). Our approach introduces a novel phenotype-aware tokenization scheme that maps 47,155 raw ICD codes to 176 clinically meaningful phenotype tokens, achieving 99.6% consolidation of diagnosis codes while preserving semantic information. This phenotype mapping contributes to a total vocabulary of 10,442 tokens - a 77.9% reduction when compared with using raw ICD codes directly. We pretrain ASCENDgpt on sequences derived from 19402 unique individuals using a masked language modeling objective, then fine-tune for time-to-event prediction of five cardiovascular outcomes: myocardial infarction (MI), stroke, major adverse cardiovascular events (MACE), cardiovascular death, and all-cause mortality. Our model achieves excellent discrimination on the held-out test set with an average C-index of 0.816, demonstrating strong performance across all outcomes (MI: 0.792, stroke: 0.824, MACE: 0.800, cardiovascular death: 0.842, all-cause mortality: 0.824). The phenotype-based approach enables clinically interpretable predictions while maintaining computational efficiency. Our work demonstrates the effectiveness of domain-specific tokenization and pretraining for EHR-based risk prediction tasks.

Problem

• Predicting cardiovascular risk from longitudinal electronic health records
• Mapping raw ICD codes to clinically meaningful phenotype tokens
• Achieving interpretable predictions while maintaining computational efficiency

Innovation

• Phenotype-aware tokenization mapping raw ICD codes
• Transformer model pretrained with masked language modeling
• Fine-tuned for time-to-event cardiovascular outcome prediction