Medical News Observer

Key Points:

• The machine learning model uses blood sample RNA sequencing and patient data to forecast secondary bacterial infections, focusing on the expression of seven specific genes.
• This predictive approach can identify patients at risk of developing secondary respiratory bacterial infections within 24 hours of hospital admission.
• The study highlights the potential of combining host transcriptomics with machine learning to refine antibiotic prescriptions, improve patient outcomes, and reduce antibiotic resistance risks.

Overview

Secondary bacterial infections often complicate viral respiratory infections. These cause significant morbidity and mortality in hospitalized COVID-19 patients. In addition, indiscriminate use of antibiotics leads to antibiotic resistance. This creates superbugs, an important problem worldwide. 

To address this problem, researchers at the University of Queensland have developed a machine learning model to predict the risk of secondary bacterial infections in hospitalized COVID-19 patients, potentially improving antibiotic usage decisions. 

Machine learning model to predict secondary infections

By analyzing blood samples RNA, sequencing, and patient data of individuals over 18 with COVID-19 symptoms and comparing them with a control group without COVID-19, the study developed a model to forecast bacterial superinfections in the respiratory tract. 

The least absolute shrinkage and selection operator (LASSO) model analyzes transcriptomic results and patient data, such as WHO severity score. The LASSO model identifies the likelihood of developing secondary respiratory bacterial infections within 24 hours of hospital admission. 

Results

Results show that the expression of seven specific genes (DAPP1, CST3, FGL2, GCH1, CIITA, UPP1, and RN7SL1) in a patient’s blood, along with their World Health Organization (WHO) severity score upon admission, can accurately predict the risk of developing a secondary bacterial infection in the respiratory tract more than 24 hours after admission. 

Implications

This study demonstrates how host transcriptomics combined with machine learning could significantly improve the precision of antibiotic prescriptions in viral infections, enhancing patient care and mitigating the risk of antibiotic resistance. 

References

Carney, Meagan, Tiana Maria Pelaia, Tracy Chew, Sally Teoh, Amy Phu, Karan Kim, Ya Wang, et al. 2024. “Host Transcriptomics and Machine Learning for Secondary Bacterial Infections in Patients with COVID-19: A Prospective, Observational Cohort Study.” The Lancet Microbe 0 (0). https://doi.org/10.1016/S2666-5247(23)00363-4.