Predicting onset of disease progression using temporal disease occurrence networks

Area of Science:

• Medical Informatics
• Computational Epidemiology
• Network Science

Background:

• Early recognition and prevention are critical for managing disease progression.
• Predictive modeling of disease trajectories remains a significant challenge in healthcare.

Purpose of the Study:

• To develop a novel technique for analyzing and predicting disease progression using temporal disease occurrence networks.
• To create a method that identifies frequent disease sequences and associated risks.

Main Methods:

• Utilized 3.9 million patient records, transforming them into temporal disease occurrence networks.
• Employed a supervised depth-first search algorithm to identify frequent disease sequences, incorporating patient demographics.
• Generated a ranked list of diseases with conditional probabilities and relative risks.

Main Results:

• The proposed network-based method demonstrated improved predictive performance.
• Achieved an Area Under the Receiver Operating Characteristic Curve (AUC) of 0.65 for single disease prediction.
• Obtained an AUC of 0.68 for predicting sets of diseases relative to ground truth.

Conclusions:

• The generated ranked list provides physicians with actionable insights into sequential disease development.
• Enables timely preventive measures based on predicted disease trajectories and risk scores.