Related Experiment Video
Updated: Aug 31, 2025
Constructing and Visualizing Models using Mime-based Machine-learning Framework
Published on: July 22, 2025
MedML: Fusing medical knowledge and machine learning models for early pediatric COVID-19 hospitalization and severity
Junyi Gao1, Chaoqi Yang1, Joerg Heintz1
1University of Illinois Urbana Champaign, Champaign, IL, USA.
Insights
A new machine learning model, MedML, predicts pediatric COVID-19 hospitalization and severity using electronic health records. MedML improves prediction accuracy by incorporating clinical knowledge, outperforming existing data-driven models.
Area of Science:
- Computational biology
- Medical informatics
- Epidemiology
Background:
- The COVID-19 pandemic caused significant socioeconomic disruption, necessitating accurate predictive models for patient outcomes.
- Effective resource allocation in healthcare requires reliable predictions of hospitalization and disease severity.
Purpose of the Study:
- To develop and evaluate MedML, a novel machine learning framework for predicting COVID-19 hospitalization and severity in pediatric patients.
- To integrate clinical domain knowledge into a predictive model using electronic health records.
Main Methods:
- MedML utilizes electronic health records, extracting predictive features from over 6 million medical concepts.
- It incorporates clinical knowledge graphs and graph neural networks to model inter-feature relationships.
- Propensity scores are used to refine feature selection for enhanced predictive power.
Main Results:
- MedML demonstrated superior performance on the National Cohort Collaborative (N3C) dataset.
- Achieved up to 7% higher Area Under the Receiver Operating Characteristic Curve (AUROC) and 14% higher Area Under the Precision-Recall Curve (AUPRC) compared to baseline models.
- The model proved more predictive and explainable than current data-driven approaches.
Conclusions:
- MedML offers a robust framework for predicting pediatric COVID-19 outcomes by integrating clinical expertise.
- The model's enhanced accuracy and explainability can aid in healthcare resource distribution during pandemics.
- This approach highlights the value of combining machine learning with medical knowledge for clinical decision support.
Abstract:
The COVID-19 pandemic has caused devastating economic and social disruption. This has led to a nationwide call for models to predict hospitalization and severe illness in patients with COVID-19 to inform the distribution of limited healthcare resources. To address this challenge, we propose a machine learning model, MedML, to conduct the hospitalization and severity prediction for the pediatric population using electronic health records. MedML extracts the most predictive features based on medical knowledge and propensity scores from over 6 million medical concepts and incorporates the inter-feature relationships in medical knowledge graphs via graph neural networks. We evaluate MedML on the National Cohort Collaborative (N3C) dataset. MedML achieves up to a 7% higher AUROC and 14% higher AUPRC compared to the best baseline machine learning models. MedML is a new machine learnig framework to incorporate clinical domain knowledge and is more predictive and explainable than current data-driven methods.
