Developing and validating clinical features-based machine learning algorithms to predict influenza infection in...

Area of Science:

• Infectious Diseases
• Medical Informatics
• Public Health

Background:

• Seasonal influenza is a significant public health concern, often leading to underdiagnosis and undertreatment in emergency departments (EDs).
• Accurate and timely diagnosis of influenza-like illness (ILI) is crucial for effective patient management.
• Machine learning (ML) offers potential for improving diagnostic accuracy in clinical settings.

Purpose of the Study:

• To develop and compare clinical feature-based ML algorithms for predicting influenza infection in ED patients with ILI.
• To evaluate the performance of ML models against existing clinical prediction models.
• To identify key clinical features associated with influenza infection.

Main Methods:

• A prospective cohort study was conducted in five EDs in the US and Taiwan (2015-2020).
• Adult patients presenting with ILI symptoms were recruited and tested for influenza using real-time RT-PCR.
• Seven ML algorithms were evaluated and compared with established clinical prediction models.

Main Results:

• The eXtreme Gradient Boosting ML algorithm demonstrated superior performance, achieving an AUC of 0.82.
• The model showed high sensitivity (0.92) and specificity (0.89) in predicting influenza infection.
• Key predictors included body temperature, cough, rhinorrhea, illness duration, and vaccination status.

Conclusions:

• Clinical feature-based ML models significantly outperform conventional models in predicting influenza infection.
• These ML tools can enhance the accuracy of influenza diagnosis in emergency departments.
• Improved diagnostic capabilities can lead to more timely and effective patient treatment.