Predicting clinical pathways of traumatic brain injuries (TBIs) through process mining

Area of Science:

• Healthcare Management
• Clinical Informatics
• Machine Learning in Medicine

Background:

• Healthcare quality is challenged by unpredictable events and complex patient conditions.
• Clinical pathways (CPs) are essential for effective resource planning and operational efficiency.
• Accurate prediction of patient journeys is vital for managing healthcare services.

Purpose of the Study:

• To develop a decision support system for predicting clinical pathways (CPs) and outcomes in traumatic brain injury (TBI) patients.
• To enhance healthcare resource planning and service efficiency through predictive modeling.
• To provide insights into the factors influencing patient journey predictions.

Main Methods:

• Utilized a machine learning framework combining an optimal decision tree and Markov-based trace clustering.
• Employed a Shapely value approach to analyze feature contributions at individual and population levels.
• Validated the predictive model using real-life event data.

Main Results:

• The proposed system demonstrated high accuracy in predicting clinical pathways (CPs) and patient outcomes for TBI.
• The Shapely value analysis provided interpretable insights into prediction drivers.
• The model's performance was validated on real-world healthcare data.

Conclusions:

• The developed decision support system effectively predicts clinical pathways (CPs) and outcomes for TBI patients.
• The interpretability of the model facilitates clinical adoption and trust in machine learning applications.
• Accurate CP prediction supports improved healthcare management and patient care strategies.