Vital Signs Prediction and Early Warning Score Calculation Based on Continuous Monitoring of Hospitalised Patients...

Area of Science:

• Biomedical Engineering
• Digital Health
• Medical Informatics

Background:

• Traditional vital signs monitoring is infrequent, missing critical deteriorations.
• Wearable technology offers continuous, high-fidelity physiological data collection.
• Early Warning Scores (EWS) are crucial for timely clinical intervention.

Purpose of the Study:

• To evaluate continuous vital signs monitoring using wearable technology for real-time EWS.
• To assess the predictive accuracy of vital signs time-series using machine learning.
• To demonstrate the feasibility of real-time EWS estimation in clinical settings.

Main Methods:

• Prospective, interventional, international study involving hospitalised patients.
• Continuous monitoring of heart rate, blood pressure, respiration rate, and oxygen saturation.
• Hybrid kNN-LS-SVM machine learning algorithm for vital signs prediction.

Main Results:

• Real-time EWS estimation every minute is feasible.
• Promising vital signs prediction performance, outperforming LSTM.
• Low mean absolute percentage errors for heart rate prediction (e.g., 4.1% for 1-hour ahead).

Conclusions:

• Wearable technology enables continuous vital signs monitoring for enhanced patient care.
• Real-time EWS and vital signs prediction offer improved early warning capabilities.
• This approach promises cost-utility, mobility, and streaming analytics for patient monitoring.