Healthcare pathway discovery and probabilistic machine learning.

Area of Science:

• Health Informatics
• Process Mining
• Machine Learning

Background:

• Healthcare pathways are crucial for quality of care, defining clinical activity sequences.
• Existing methods lack interpretability for clinicians without process mining expertise.
• Predictive models for recovery times need to account for performance indicator volatility.

Purpose of the Study:

• To combine healthcare pathway discovery with predictive models for individualized recovery times.
• To develop interpretable pathway models for clinicians.
• To incorporate stochastic volatility of performance indicators into predictive models.

Main Methods:

• Utilized business process-mining software (ProM) for pathway discovery and enrichment.
• Developed a process mining pipeline using hospital records.
• Applied probabilistic machine learning models for recovery time prediction.

Main Results:

• Discovered interpretable appendicitis pathway models from hospital records.
• Identified factors influencing patient recovery time.
• Developed a probabilistic regression model for estimating patient recovery time, aiding hospital scheduling.

Conclusions:

• Established ProM as a tool for improving healthcare pathway mining.
• The proposed pipeline supports the development of machine learning models linking pathways to performance indicators like recovery time.