Enhancing respiratory virus surveillance among hospitalised children: a machine learning-based predictive model

Area of Science:

• Pediatric Infectious Diseases
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Viral respiratory tract infections (vRTIs) are a major cause of pediatric hospitalizations.
• Current surveillance methods lack diagnostic accuracy in children due to varied symptoms.
• Need for improved diagnostic tools for vRTIs in pediatric populations.

Purpose of the Study:

• To develop and evaluate a machine learning (ML) model for predicting microbiologically confirmed vRTIs in hospitalized children.
• To assess the model's performance across different age groups and viral pathogens.
• To compare ML model accuracy against traditional syndromic surveillance definitions.

Main Methods:

• Retrospective study of 2050 pediatric patients with acute respiratory infections.
• Utilized admission data including symptoms, age, and comorbidities.
• Trained and tested six ML algorithms, selecting the best-performing logistic regression model based on auROC.

Main Results:

• The ML logistic regression model achieved an auROC of 0.754, outperforming traditional methods.
• Model demonstrated higher accuracy in infants (<1 year) and for respiratory syncytial virus (RSV) cases.
• High prevalence of RSV (38.7%) and enterovirus/rhinovirus (32.8%) detected among positive cases.

Conclusions:

• ML-based prediction models accurately identify pediatric vRTIs using readily available admission data.
• These models offer superior performance compared to syndromic surveillance, particularly for infants.
• Integration into electronic medical records can enhance surveillance, outbreak detection, and pandemic preparedness.