Development and validation of a machine learning-based diagnostic system for 22 pediatric respiratory pathogens: a...

Area of Science:

• Pediatric infectious diseases
• Medical informatics
• Computational biology

Background:

• Respiratory tract infections (RTIs) are a major cause of childhood illness.
• Accurate pathogen identification is essential for effective treatment of pediatric RTIs.
• Current diagnostic methods can be slow, delaying targeted therapy.

Purpose of the Study:

• To develop and validate an interpretable diagnostic system, Pathog-PDx, for early identification of respiratory pathogens in children.
• To assess the real-world applicability of the Pathog-PDx model.
• To provide a decision support tool for clinicians treating pediatric RTIs.

Main Methods:

• Developed an interpretable Pathogen Diagnostic System for Pediatric Respiratory Infections (Pathog-PDx) using 42 clinical and laboratory features from electronic health records (EHR).
• Conducted a multicenter study involving 134,500 hospitalized children for model development.
• Performed prospective validation on an independent cohort of 1338 children.

Main Results:

• Pathog-PDx accurately distinguished 22 pathogen subtypes, outperforming conventional models in identifying single and mixed infections.
• Achieved high classification performance for key pathogens like influenza virus (AUC=0.95, Sn=0.88, Sp=0.86).
• Demonstrated a mean AUC of 0.88 for various RTI pathogens.

Conclusions:

• Pathog-PDx is a validated tool for early identification of respiratory pathogens in pediatric patients.
• The system provides actionable predictions before conventional test results are available.
• Pathog-PDx has been deployed as a freely accessible web-based decision support system to guide timely and targeted therapy.