Using Decision Trees as an Expert System for Clinical Decision Support for COVID-19

Area of Science:

• Infectious Diseases
• Medical Informatics
• Computational Biology

Background:

• The global impact of COVID-19 necessitates advanced tools for patient management.
• A lack of accessible tools exists for determining COVID-19 infection severity and associated risk factors.
• Understanding biological system responses and comorbidities is crucial for predicting severe COVID-19 outcomes.

Purpose of the Study:

• To address the absence of a patient and provider-facing chatbot for COVID-19 severity assessment.
• To construct a decision tree framework for stratifying COVID-19 cases based on key patient and disease characteristics.
• To explore the feasibility of a stratified decision support system for severe COVID-19 prediction.

Main Methods:

• Literature review to identify relevant factors for COVID-19 severity.
• Development of a binary classification decision tree using a suite of tools.
• Stratification of the decision tree by age, body system, viral infection, comorbidities, and manifestations.
• Construction of a framework with 212 nodes, encompassing potential patient scenarios.

Main Results:

• A decision tree framework with 212 nodes was established, generating 63,360 potential scenarios.
• Stratification by body system, comorbidities, and manifestations significantly strengthens the predictive framework.
• The study identified the complex interplay of factors contributing to severe COVID-19 cases.

Conclusions:

• The developed decision tree prototype provides insight into data requirements for effective COVID-19 decision support.
• Stratifying viral infection, body systems, comorbidities, and manifestations enhances the framework's robustness.
• Further validation is needed to create a clinically viable decision tree for COVID-19 management.