Patient Management Assisted by a Neural Network Reduces Mortality in an Intermediate Care Unit

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Critical Care Medicine

Background:

• Mortality risk prediction in Intermediate Respiratory Care Units (IRCUs) requires specialized approaches distinct from Intensive Care Units (ICUs).
• Developing accurate predictive tools for IRCU patients is crucial for optimizing treatment strategies and improving patient outcomes.

Purpose of the Study:

• To develop and validate an IRCU-specific mortality predictor tool utilizing machine learning methodologies.
• To identify key clinical parameters associated with mortality risk in the IRCU setting.

Main Methods:

• A neural network was employed to identify significant mortality predictors from patient vital signs.
• Multivariate logistic regression was used to establish discriminatory cut-off points for identified parameters.
• The developed risk assessment guideline was applied, and mortality outcomes were tracked over one year.

Main Results:

• Key mortality predictors identified include thrombocytopenia, metabolic acidosis, anemia, tachypnea, age, sodium levels, hypoxemia, leukocytopenia, and hyperkalemia.
• Implementation of the predictive tool in the first year resulted in a 50% decrease in the patient failure rate.

Conclusions:

• A novel neural network model effectively identifies and classifies mortality predictors within a general hospital's IRCU.
• The combined approach of neural networks and multivariate regression provides a valuable real-time patient monitoring tool for specific mortality risks.
• The algorithm is scalable and adaptable to various care units, with potential for increased accuracy as more patient data is incorporated.