[Paper Review] Machine learning techniques to identify antibiotic resistance in patients diagnosed with various skin and soft tissue infections

Farnaz H. Foomani, Shahzad Mirza+5 more|arXiv (Cornell University)|Feb 28, 2022

Bacterial Identification and Susceptibility Testing4 citations

TL;DR

This study develops machine learning models to predict antibiotic resistance in patients with skin and soft tissue infections (SSTIs) using clinical, demographic, and antibiotic susceptibility testing (ABST) data. The models achieve AUC scores of 0.68–0.98 for Gram-positive cocci and 0.56–0.93 for Gram-negative bacilli, demonstrating that patient-level features can predict resistance with moderate to high accuracy, depending on the antimicrobial class.

ABSTRACT

Skin and soft tissue infections (SSTIs) are among the most frequently observed diseases in ambulatory and hospital settings. Resistance of diverse bacterial pathogens to antibiotics is a significant cause of severe SSTIs, and treatment failure results in morbidity, mortality, and increased cost of hospitalization. Therefore, antimicrobial surveillance is essential to predict antibiotic resistance trends and monitor the results of medical interventions. To address this, we developed machine learning (ML) models (deep and conventional algorithms) to predict antimicrobial resistance using antibiotic susceptibility testing (ABST) data collected from patients clinically diagnosed with primary and secondary pyoderma over a period of one year. We trained an individual ML algorithm on each antimicrobial family to determine whether a Gram-Positive Cocci (GPC) or Gram-Negative Bacilli (GNB) bacteria will resist the corresponding antibiotic. For this purpose, clinical and demographic features from the patient and data from ABST were employed in training. We achieved an Area Under the Curve (AUC) of 0.68-0.98 in GPC and 0.56-0.93 in GNB bacteria, depending on the antimicrobial family. We also conducted a correlation analysis to determine the linear relationship between each feature and antimicrobial families in different bacteria. ML techniques suggest that a predictable nonlinear relationship exists between patients' clinical-demographic characteristics and antibiotic resistance; however, the accuracy of this prediction depends on the type of the antimicrobial family.

Motivation & Objective

• To address rising antibiotic resistance in skin and soft tissue infections (SSTIs), which contribute to treatment failure, morbidity, mortality, and increased healthcare costs.
• To develop machine learning models that predict resistance patterns in Gram-positive cocci (GPC) and Gram-negative bacilli (GNB) using patient-level clinical and demographic features.
• To evaluate the predictive performance of both deep and conventional machine learning algorithms on ABST data collected over a one-year period.
• To identify which patient features correlate most strongly with resistance across different antimicrobial families.
• To assess whether nonlinear relationships between patient characteristics and resistance can be reliably modeled.

Proposed method

• The study trained individual machine learning models for each antimicrobial family, distinguishing between Gram-positive cocci (GPC) and Gram-negative bacilli (GNB).
• Features included clinical and demographic patient data alongside antibiotic susceptibility testing (ABST) results from a one-year clinical dataset.
• Both conventional and deep learning algorithms were applied to predict resistance outcomes for each antibiotic class.
• Model performance was evaluated using the Area Under the Curve (AUC) metric to assess discrimination ability.
• Correlation analysis was performed to quantify linear relationships between individual features and resistance outcomes for each antimicrobial family.
• The analysis focused on identifying patterns in patient characteristics predictive of resistance across diverse bacterial pathogens.

Experimental results

Research questions

• RQ1 Can machine learning models accurately predict antibiotic resistance in patients with skin and soft tissue infections using clinical and demographic data?
• RQ2 How does model performance vary across different antimicrobial families for Gram-positive and Gram-negative pathogens?
• RQ3 What is the strength and nature of the relationship between patient-level features and resistance outcomes?
• RQ4 To what extent can nonlinear patterns in patient data predict resistance trends?
• RQ5 Which clinical and demographic features show the strongest correlation with resistance in specific antibiotic classes?

Key findings

• The machine learning models achieved AUC scores ranging from 0.68 to 0.98 for Gram-positive cocci (GPC), indicating strong to moderate predictive performance depending on the antimicrobial family.
• For Gram-negative bacilli (GNB), AUC values ranged from 0.56 to 0.93, showing variable but often strong predictive capability.
• A nonlinear relationship was identified between patient clinical-demographic features and antibiotic resistance, suggesting complex interactions beyond linear models.
• The predictive accuracy varied significantly by antimicrobial family, with higher performance observed for certain drug classes.
• Correlation analysis revealed that specific patient features were more strongly associated with resistance in particular bacterial groups and antibiotic types.
• The results demonstrate that machine learning can effectively model resistance trends using routinely collected clinical data, supporting early intervention and antimicrobial stewardship.