Interpretable Machine Learning Model for COVID-19 Diagnosis Demonstrates High Accuracy

Researchers at Nankai University, in collaboration with the Tianjin Union Medical Center, have developed an interpretable machine learning model that accurately diagnoses common respiratory virus infections, including influenza A and Mycoplasma pneumoniae, in patients. The study, published in the journal Immunity, Inflammation and Disease, utilized a dataset of 7,471 patients who presented with fever at Central Hospital between November and December 2023. The model achieved high accuracy in detecting Flu A with a recall rate of 81.0% and enhanced accuracy in identifying MP cases with a precision rate of 84.3%.

Key Takeaways:

• The interpretable machine learning model was developed using complete blood count (CBC) test data and was able to accurately identify common respiratory virus infections, including influenza A and Mycoplasma pneumoniae.
• The model was trained on a dataset of 7,471 patients who presented with fever at Central Hospital between November and December 2023.
• The high-recall model demonstrated superior performance in detecting Flu A, with a recall rate of 81.0%, while the precision-recall balanced model exhibited enhanced accuracy in identifying MP cases, with a precision rate of 84.3%.
• The model was refined through manual parameter adjustments and a comprehensive network search.
• The study's findings suggest that the use of interpretable machine learning models can aid clinicians in enhancing diagnostic efficiency and accuracy.
• The model's high accuracy rates could lead to reduced medical expenses by minimizing unnecessary tests and treatments.

Statistics:

• 7,471 patients were included in the study's derivation cohort, who presented with fever at Central Hospital between November and December 2023.
• The high-recall model achieved a recall rate of 81.0% for detecting Flu A.
• The precision-recall balanced model exhibited a precision rate of 84.3% for identifying MP cases.

Sources:

• "Clinical Characteristics of Patients With Respiratory Infections After Nonpharmacological Interventions for COVID-19 in China Have Ended: Using Machine Learning Approaches to Support Pathogen Prediction at Admission" by Yan-Hong Liu et al., Immunity, Inflammation and Disease, 2025;13(8).
• The journal Immunity, Inflammation and Disease can be contacted at: Wiley, 111 River St, Hoboken 07030-5774, NJ, USA.
• The Tianjin Union Medical Center can be contacted at: Nankai University, Tianjin, People's Republic of China.
• Additional authors include: Tian-Ning Li, Kwok-Leung Yiu, Lu Liu, Meng Han, Wei-Jia Ma, Chun-Lei Zhou, and Hong Mu.