Modelling physiological deterioration in post-operative patient vital-sign data

Area of Science:

• Medical Informatics
• Clinical Monitoring
• Surgical Outcomes

Background:

• Upper-gastrointestinal surgery patients frequently experience post-operative complications.
• These complications often necessitate intensive care unit (ICU) admission.
• Early detection of deterioration is crucial for improved patient outcomes.

Purpose of the Study:

• To explore vital sign changes in post-operative patients.
• To develop a normality model for early detection of patient deterioration.
• To assess the utility of vital sign trajectory analysis in post-operative care.

Main Methods:

• Utilized observational vital sign data from 171 post-operative patients.
• Employed univariate and multivariate analyses to study vital sign trajectories.
• Constructed a normality model using kernel density estimation on "normal" recovery data.
• Tested the normality model with "abnormal" data from patients requiring ICU re-admission.

Main Results:

• Vital sign distributions in "normal" recovery patients varied from admission to discharge.
• No significant vital sign distribution changes were observed from mid-stay to discharge in "normal" patients.
• The normality model successfully identified patient deterioration using unseen "abnormal" data.

Conclusions:

• Vital sign monitoring can reveal patient deterioration post-surgery.
• Kernel density estimation provides a robust method for normality modeling.
• This approach offers potential for early warning systems in post-operative care.