Tracking Temporal Clusters from Patient Networks.

Area of Science:

• Health Informatics
• Data Science
• Epidemiology

Background:

• Medico-administrative databases are valuable for understanding health determinants but contain truncated patient care pathways.
• Analyzing these truncated pathways is challenging for comprehensive patient journey mapping.

Purpose of the Study:

• To develop and validate a patient network-based approach for constructing complete care trajectories from truncated medico-administrative data.
• To assess the performance of this novel approach against existing methods using a real-world dataset.

Main Methods:

• Constructed patient networks for each age group using antithrombotic treatment data (2008-2018) from the EGB.
• Applied the Markov clustering algorithm to identify patient clusters within each network.
• Matched clusters across consecutive networks to reconstruct longitudinal care trajectories.

Main Results:

• Successfully identified 12 distinct care trajectories associated with specific pathologies.
• The network-based approach achieved the highest silhouette score, outperforming three other existing methods.
• Demonstrated the ability to capture the longitudinal, multidimensional, and truncated nature of the data.

Conclusions:

• The proposed patient network approach effectively reconstructs care trajectories from truncated medico-administrative data.
• This method offers a robust tool for analyzing patient journeys and understanding health determinants.
• Highlights the potential of network analysis in health informatics for longitudinal data interpretation.