Development of Predictive Models to Inform a Novel Risk Categorization Framework for Antibiotic Resistance in...

Area of Science:

• Infectious Diseases
• Clinical Pharmacology
• Health Informatics

Background:

• Antibiotic resistance is a growing concern in treating uncomplicated urinary tract infections (uUTIs).
• Identifying patients at risk for antibiotic nonsusceptibility is crucial for appropriate prescribing.
• Current clinical practice faces challenges in accurately assessing this risk.

Purpose of the Study:

• To develop predictive models for antibiotic nonsusceptibility in uUTIs.
• To identify key predictors of nonsusceptibility to common antibiotic classes.
• To create a risk categorization framework for guiding treatment decisions.

Main Methods:

• Utilized a de-identified Electronic Health Record dataset of females aged ≥12 years with Escherichia coli uUTI.
• Developed four predictive models for nonsusceptibility to nitrofurantoin, trimethoprim-sulfamethoxazole, β-lactams, and fluoroquinolones.
• Constructed a risk categorization framework classifying isolates as low, moderate, or high risk.

Main Results:

• Models were developed using data from 87,487 patients.
• Key predictors included prior UTI history, previous antibiotic resistance, and patient demographics.
• The framework identified varying proportions of high-risk isolates across antibiotic classes (e.g., 17.4% for β-lactams).

Conclusions:

• Predictive models and a risk framework effectively highlight factors associated with antibiotic nonsusceptibility in uUTIs.
• These tools offer valuable insights for clinicians to inform empiric antibiotic prescribing.
• The findings contribute to strategies for mitigating antibiotic resistance in uUTI management.