A Process Mining Pipeline to Characterize COVID-19 Patients' Trajectories and Identify Relevant Temporal Phenotypes...

Area of Science:

• Digital Health
• Health Informatics
• Data Science in Healthcare

Background:

• The COVID-19 pandemic significantly disrupted healthcare delivery systems.
• Effective re-organizational strategies are crucial for managing patient care during health crises.
• Understanding patient population dynamics within hospitals is vital for optimizing healthcare strategies.

Purpose of the Study:

• To develop an analysis pipeline for identifying healthcare process patterns in COVID-19 patients.
• To integrate demographic, physiological, and care process data for cohort stratification.
• To generate hypotheses for more effective healthcare strategies by analyzing patient subgroups.

Main Methods:

• An integrated pipeline combining process mining and trajectory mining (topological data analysis, pseudo-time) was employed.
• Data from 1,179 COVID-19 patients, including admission letters, EHR, and hospital infrastructure data, were analyzed.
• Heterogeneous data sources were integrated to identify frequent healthcare process patterns and patient trajectories.

Main Results:

• Five distinct temporal phenotypes were identified based on laboratory value trajectories, each associated with significantly different mortality risks.
• Process mining algorithms successfully stratified patients into sub-cohorts based on pandemic waves and temporal trajectories.
• Statistically significant differences in event characteristics were observed across these sub-cohorts.

Conclusions:

• The developed pipeline effectively identifies patient phenotypes and stratifies cohorts based on complex healthcare data.
• Temporal phenotypes derived from laboratory data offer valuable insights into COVID-19 patient prognosis and risk stratification.
• This approach provides a foundation for data-driven healthcare management and strategic planning during pandemics.