Machine learning-based prediction models for home discharge in patients with COVID-19: Development and evaluation...

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Public Health

Background:

• Hospitalized COVID-19 patients require disposition planning.
• Electronic Health Records (EHR) contain valuable data for predicting patient outcomes.
• Efficient resource allocation is critical during public health crises.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting COVID-19 patient disposition.
• To determine if hospitalized COVID-19 patients will be admitted to alternative care or discharged home.
• To leverage EHR data for improved healthcare decision-making.

Main Methods:

• Retrospective cohort study of 1,578 hospitalized COVID-19 patients.
• Development and validation of six supervised ML models (e.g., Random Forest, Logistic Regression).
• Model performance evaluated using ROC-AUC, precision, accuracy, F1 score, and Brier score.

Main Results:

• The Random Forest classifier achieved the highest accuracy (0.84) and AUC (0.72).
• Other models like Logistic Regression also showed strong predictive capabilities (accuracy: 0.85, AUC: 0.71).
• The models demonstrated valuable performance in predicting patient discharge status.

Conclusions:

• ML models utilizing EHR data can effectively predict COVID-19 patient disposition.
• These predictive tools are crucial for optimizing healthcare resource allocation during pandemics.
• Explainable ML methods can enhance understanding of factors influencing healthcare decisions.