Comorbidity progression analysis: patient stratification and comorbidity prediction using temporal comorbidity...

Area of Science:

• Medical Informatics
• Computational Biology
• Health Services Research

Background:

• Understanding comorbidity progression is crucial for effective patient management and personalized treatment strategies.
• Existing methods often lack the ability to capture dynamic temporal patterns of disease co-occurrence.
• Identifying distinct population-specific comorbidity trajectories is essential for timely intervention.

Purpose of the Study:

• To develop and validate a framework for analyzing comorbidity progression patterns using temporal comorbidity networks (TCN).
• To enable timely detection of potential comorbidities and enhance understanding of comorbid condition development.
• To stratify patients into distinct subgroups based on their comorbidity progression.

Main Methods:

• Constructed temporal comorbidity networks (TCN) from longitudinal patient diagnosis data.
• Utilized TCN for patient stratification through preliminary and prescription analysis.
• Developed a distance-matched temporal comorbidity network (TCN-DM) for comorbidity prediction by identifying similar patients.

Main Results:

• The framework successfully identified four distinct heart failure (HF) subgroups using the MIMIC-III dataset.
• TCN revealed significant comorbidity progression patterns within these HF patient subgroups.
• The TCN-DM method outperformed other approaches in comorbidity prediction, achieving F1-Scores between 0.454 and 0.612.

Conclusions:

• The proposed framework effectively identifies population-specific comorbidity patterns and predicts future comorbidity developments.
• This approach offers valuable insights for both individual patient care and population health management.
• The TCN-DM method shows promise for improving the accuracy of comorbidity prediction in clinical settings.