A Machine Learning Framework for Predicting Nosocomial Escherichia coli Infections in Cervical Cancer

Area of Science:

• Oncology
• Infectious Diseases
• Medical Informatics

Background:

• Nosocomial infections pose a significant threat to patients undergoing treatment for cervical cancer.
• Characterizing the etiological profile of these infections is crucial for effective management.
• Escherichia coli is identified as a predominant pathogen in this patient group.

Purpose of the Study:

• To determine the causes of hospital-acquired infections in cervical cancer patients.
• To develop a machine learning model for predicting Escherichia coli infections.
• To aid clinical decision-making for anti-infective therapy and risk stratification.

Main Methods:

• Retrospective analysis of clinical data from 118 cervical cancer patients.
• Evaluation of pathogen distribution and antimicrobial resistance patterns.
• Development of a predictive model for Escherichia coli using logistic regression and support vector machine (SVM) algorithms.

Main Results:

• Gram-negative bacteria, particularly Escherichia coli (50.33%), were the most common pathogens, predominantly in urine samples (69.54%).
• High resistance rates of Escherichia coli to common antibiotics (e.g., ceftriaxone, ciprofloxacin) were observed, with susceptibility to carbapenems and amikacin.
• Machine learning models demonstrated strong performance (AUC up to 0.90) in predicting Escherichia coli infection based on clinical factors like disease stage and anemia status.

Conclusions:

• Urinary tract infections caused by gram-negative bacilli, especially Escherichia coli, are prevalent in cervical cancer patients.
• A machine learning model utilizing clinical parameters effectively predicts Escherichia coli infections in mid-stream urine samples.
• This predictive tool can enhance risk identification and guide targeted empiric therapy, potentially improving patient outcomes.