Accurate prediction of neonatal mortality remains one of the most challenging problems in critical care. This study developed a High Definition Phenotype (HDP) framework that integrates electronic medical records, laboratory results, bedside physiological monitoring, medications, nutrition, and derived physiological features into a unified time-series representation of each patient.
Clinical data from 1,546 neonatal admissions across eight Level III NICUs were collected, with 757 patients containing complete fixed, intermittent, and continuous datasets for model development. Two predictive approaches—a Logistic Regression Model (LRM) and a Long Short-Term Memory (LSTM) deep learning model—were developed and compared with established neonatal severity scores including CRIB, CRIB-II, SNAP-II, and SNAPPE-II.
Both AI models consistently outperformed traditional severity scores, with the LSTM achieving an AUC of up to 0.96, while also providing continuously updated mortality risk predictions throughout hospitalization. The framework demonstrates the potential of integrating multimodal clinical data into real-time bedside decision support systems capable of detecting patient deterioration earlier than conventional approaches.
Length of stay (LOS) is an important indicator of neonatal outcomes, healthcare resource utilization, and hospital costs. This study developed an artificial intelligence–based prediction model that estimates NICU length of stay using routinely collected electronic medical record data. Unlike previous approaches, the model incorporated antenatal and perinatal characteristics, deviations from ASPEN nutrition guidelines, deviations from NeoFax medication recommendations, and major neonatal clinical diagnoses. Data from 836 neonatal admissions across two tertiary NICUs were used to develop gestation-specific predictive models and validated using an independent cohort of 211 patients. The resulting bedside interface enables clinicians to visualize patient-specific risk factors contributing to prolonged hospitalization, supporting family counseling, clinical decision-making, and operational planning while demonstrating the feasibility of real-time AI-assisted predictive analytics within neonatal intensive care.
Objective
To evaluate an edge-hosted large language model framework capable of generating neonatal intensive care discharge summaries while preserving patient privacy.
Methods
MORPHEUS was evaluated using 401 NICU admissions. Performance was measured using a clinician-defined rubric covering 72 clinical documentation fields, including completeness, clarity, coherence, actionability, and factual accuracy.
Key Findings
• Higher completeness than clinician-authored summaries
• Improved documentation consistency
• Privacy-preserving edge deployment without cloud processing
• Demonstrated feasibility for real-world NICU workflows
Impact
This work demonstrates how edge AI can reduce clinician documentation burden while maintaining high-quality discharge summaries and protecting sensitive patient information.
