Forecasting of Continuous Vital Sign Using Multivariate Auto-Regressive Models.

Area of Science:

• Biomedical Engineering
• Clinical Informatics
• Time Series Analysis

Background:

• Continuous monitoring of vital signs is crucial for hospitalized patients.
• Accurate forecasting of vital signs can improve patient management and outcomes.
• Existing forecasting methods may have limitations in capturing complex physiological dynamics.

Purpose of the Study:

• To assess the efficacy of multivariate auto-regressive (MAR) models for forecasting continuous vital signs.
• To evaluate the performance of MAR models in predicting heart rate and respiration rate in postoperative patients.
• To explore the potential of MAR models for real-time clinical decision support.

Main Methods:

• Utilized 20 hours of continuous heart rate and respiration rate data from eight postoperative patients.
• Fitted a centered MAR model for forecasting in 15-minute windows.
• Employed Markov Chain Monte Carlo (MCMC) sampling for model fitting.
• Validated the model using data from five additional patients.

Main Results:

• Achieved an average Root Mean Square Error (RMSE) of 11.4 beats per minute for heart rate forecasts.
• Achieved an average RMSE of 3.3 breaths per minute for respiration rate forecasts.
• Demonstrated the model's predictive capability within the specified forecast window.

Conclusions:

• MAR models show significant potential for forecasting continuous vital signs in hospitalized patients.
• The findings suggest that MAR models can be a valuable tool for enhancing patient monitoring.
• Further research can explore the integration of MAR models into clinical workflows for proactive patient care.