The study examined how physiological trajectories and observation patterns influence ICU mortality prediction across different datasets (MIMIC‑III, MIMIC‑IV, eICU). Models incorporating longitudinal physiological summaries (R2) consistently improved external prediction, while observation‑pattern features (R3) mainly benefited locally developed models and were highly sensitive to measurement timing and availability. The findings highlight that the transportability of clinical prediction models depends not only on patient populations but also on the conditions under which measurements are collected.
arXiv:2608. 02663v1 Announce Type: cross Abstract: Accurate ICU mortality prediction requires modeling irregular clinical observations across heterogeneous entity types.
The paper introduces a bounded updating framework for ICU prediction models that freezes the physiological encoder while allowing updates only to the treatment pathway and fusion head. Experiments on 84,792 MIMIC-IV ICU stays across four temporal shifts show that selective adaptation yields more stable explanations—higher rank correlation, better top‑5 feature agreement, and improved retrieval stability—compared to full model adaptation. Predictive performance varies by outcome, but the results demonstrate that explanation stability is governed by the structural boundaries of allowed adaptation rather than merely by freezing components.
PGP-Clinical-TimeKAN is a trajectory-first framework for joint probabilistic forecasting of multivariate physiological data, combining missingness-aware temporal encoders, a soft organ-system prior, patient-specific relations, nonlinear Kolmogorov‑Arnold messages, and a low‑rank multivariate Student‑t head. Evaluated on a MIMIC‑IV cohort of 6,882 patients, it achieves the second‑lowest normalized MAE and the lowest RMSE among 13 models, while providing calibrated probabilistic forecasts with empirical coverage at 50%, 80%, and 95% intervals. Ablation studies show that relational structure is critical for performance, and increasing covariance rank improves likelihood but not point accuracy. whyItMatters":"The model demonstrates that joint trajectory forecasting can yield highly accurate, calibrated predictions of physiological trajectories, offering a potentially inspectable intermediate task for clinical deterioration prediction."
The study introduces a graph‑enhanced framework for discovering disease trajectories and forecasting temporal endpoints using longitudinal EHR data from MIMIC‑IV v3.1. It constructs patient‑specific motifs from consecutive admissions, applies a time‑aware Transformer to select up to 24 motifs per endpoint, and reports modest improvements in micro‑AUPRC and event‑time MAE, with larger gains for events occurring after 730 days. The authors also explore branching‑motif conditioning and GNN‑based edge repair, finding that motifs mainly provide long‑range timing support rather than superior endpoint discrimination or survival prediction.
arXiv:2607. 02938v1 Announce Type: new Abstract: Clinical time series prediction in intensive care units remains challenging due to heterogeneous physiological variables and informative missingness.
arXiv:2608.29301v1 Announce Type: new Abstract: Predicting future organ dysfunction in Intensive Care Unit (ICU) patients is critical for early clinical intervention, yet existing machine learning ap...
This study develops a rolling day‑wise mortality prediction model for critically ill patients with acute kidney injury on continuous renal replacement therapy (CRRT) by incorporating minute‑level machine pressure waveforms alongside electronic health record (EHR) data. After cleaning the raw machine data—aligning it with therapy intervals, removing priming and downtime minutes, and denoising artifacts—the authors trained a transformer‑based stacked ensemble that fuses a sequence transformer with classical models using circuit‑instability features and clinical variables. In a multi‑center benchmark of 976 patients, the combined model achieved a one‑day mortality AUROC of 0.766, outperforming machine‑only (0.625) and EHR‑only (0.717) models, and SHAP analysis highlighted key pressure‑related features such as filter pressure, transmembrane pressure, and access‑to‑return difference. whyItMatters":"By integrating previously discarded CRRT machine data, the study provides a continuous, real‑time risk signal that could enable earlier detection of patient deterioration and improve mortality prediction in a high‑risk population."
The study evaluates how acquisition context and photoplethysmography (PPG) signals contribute to session‑calibrated blood pressure tracking using a wearable‑state model on 653 AuroraBP participants. Adding contextual information reduced mean absolute errors (MAEs) for systolic and diastolic BP by 0.436 mmHg and 0.126 mmHg, respectively, while PPG alone provided modest improvements. Exploratory follow‑ups with sensor‑history ensembles and pulse‑rate fusion yielded only slight gains, and a separate intraoperative model (VitalDB) showed limited PPG benefit.
arXiv:2608. 00058v1 Announce Type: cross Abstract: Accurate measurement of ECG intervals, including PR, QRS duration, and QT/QTc, is central to cardiac diagnosis, yet the published ECG delineation literature evaluates performance almost exclusively as fiducial-point timing errors on small curated databases, rather than as clinical interval accuracy on large unselected cohorts.
The paper introduces Curriculum‑Aware Interpolate‑then‑Refine (CAIR), a two‑stage framework for imputing physiological time‑series data. CAIR first learns a coarse base curve with a bidirectional‑GRU interpolator and then refines it through three Transformer passes, trained under a random‑gap curriculum that mimics realistic missingness. Evaluations on continuous glucose monitoring and arterial pressure datasets show CAIR outperforms all baselines across MCAR, MAR, and NMAR mechanisms, especially for long gaps and value‑dependent dropout, while also preserving clinically relevant burden metrics.
The paper introduces NECA, a model‑agnostic augmentation technique that retrieves similar patients from the training set and transforms their organ‑dysfunction event onset times into cumulative event‑curve features for multiple future horizons (6h, 12h, 24h, 42h). These features are concatenated with the original patient representation and fed to standard downstream predictors, leading to improved PR‑AUC and ROC‑AUC on PhysioNet 2012 and a MIMIC‑IV subset across several organ‑dysfunction tasks. Ablation studies show that clinically meaningful retrieval and horizon‑specific neighbor risk are the main drivers of performance gains, especially for neurological dysfunction events.
The paper introduces CALIBRA, a calibration-first multimodal temporal learning framework designed for transferable asthma‑risk forecasting across varying patient cohorts and sensor ecosystems. It processes multiple data streams—environmental, pulmonary, symptom, medication, wearable, and context—using dedicated recurrent encoders, a reliability‑conditioned gate, and gradient‑reversal training to mitigate cohort bias. CALIBRA employs a shrinkage‑based hierarchical logistic layer for probability calibration and split conformal prediction for abstention‑capable prediction sets, achieving competitive performance on a semi‑synthetic three‑cohort benchmark with controlled distribution shift.