CT-HEG: A Bidirectional, Timestamp-Attributed Event Graph for ICU In-Hospital Mortality Prediction - An Architectural Ablation Study

2026-08-01

AI Summary

This study addresses the challenge of in-hospital mortality prediction in intensive care units, where irregular sampling, heterogeneous entities, and dynamic temporal relationships complicate modeling. The authors propose a Continuous-Time Heterogeneous Electronic Health Record Graph (CT-HEG) framework that represents each hospitalization as a bidirectional event graph annotated with timestamps and event types, and introduce the CHIRP-Net model for risk prediction. A key innovation lies in the bidirectional, time-aware heterogeneous graph structure, which reveals the critical role of reverse edges in information aggregation; notably, simplifying edge types improves performance, challenging the presumed necessity of complex heterogeneity modeling. Evaluated on MIMIC-IV, CHIRP-Net achieves an AUROC of 0.8449 ± 0.0071, with an ensemble variant reaching 0.8618. Ablation shows that removing reverse edges reduces AUROC by 0.1968, and the calibrated model attains an expected calibration error (ECE) of 0.0307.

Abstract

Accurate ICU mortality prediction requires modeling irregular clinical observations across heterogeneous entity types. Existing sequence models handle irregular sampling but ignore typed relational structure; existing graph models assume fixed-interval inputs. We introduce the Continuous-Time Heterogeneous EHR Graph (CT-HEG) schema and evaluate which architectural choices drive predictive performance. CT-HEG encodes each ICU stay as a typed, timestamped graph with three node types (visit, vital, lab_event) and 2D edge attributes (t_hours/48, value_norm) encoding timing and value without imputation. We instantiate CT-HEG as CHIRP-Net, a four-layer heterogeneous GATv2Conv network, evaluated on MIMIC-IV v3.1 (31,142 ICU stays, LOS>=48h, 13.4% mortality) with five seeds and bootstrapped confidence intervals, against logistic regression, mTAND, a Transformer, and GRU-D, plus an ablation study. CHIRP-Net achieved 5-seed mean AUROC 0.8449+/-0.0071 (AUPRC 0.4958+/-0.0209); the ensemble achieved AUROC 0.8618 (95% CI: 0.8485-0.8745). Removing reverse edges disconnected observation nodes from the visit readout, cutting AUROC by 0.1968+/-0.0073. Time-attentive edge features contributed 0.0247+/-0.0093 AUROC. Collapsing heterogeneous edge types into one relation (7x fewer parameters) outperformed the full model on all seeds. Post-calibration ECE was 0.0307. Temporal and demographic subgroup analyses were explored but not reported here, pending follow-up work. Bidirectional connectivity was necessary for the model to use its inputs at all, and CT-HEG was reasonably well calibrated after validation-fitted temperature scaling. These results support CT-HEG for irregular EHR data, while external validation, a pre-specified temporal evaluation, and a demographic fairness audit remain necessary before any claim of robustness. Code: https://github.com/nasiruddinstudents-ctrl/chirp-net-mimic-iv.

Problem

• ICU mortality prediction
• irregular clinical observations
• heterogeneous event graph
• timestamp-attributed data
• EHR modeling

Innovation

• Continuous-Time Heterogeneous Graph
• Bidirectional Event Graph
• Time-Attentive Edge Features
• Architectural Ablation Study
• Irregular EHR Modeling