Use of Machine Learning to Assess the Management of Uncomplicated Urinary Tract Infection

Area of Science:

• Infectious Diseases
• Pharmacology
• Health Services Research

Background:

• Uncomplicated urinary tract infections (UTIs) are frequently treated with outpatient antibiotics.
• National guidelines for UTI management were established in 2011.
• Current practices, patient demographics, and pathogen resistance patterns have evolved since 2011.

Purpose of the Study:

• To assess the effectiveness and safety of first-line antibiotics, fluoroquinolones, and oral beta-lactams for uncomplicated UTIs in current clinical practice.
• To compare treatment failure rates and adverse events associated with different antibiotic classes.

Main Methods:

• Retrospective, population-based cohort study using a large regional claims dataset (2012-2021).
• Included nonpregnant females aged 18+ with uncomplicated UTI treated with specific first-line, fluoroquinolone, or beta-lactam antibiotics.
• Primary outcome: 30-day composite endpoint for treatment failure (revisit for UTI, pyelonephritis, or sepsis). Secondary outcomes: specific adverse events.

Main Results:

• First-line therapies were associated with significantly lower 30-day revisit rates compared to fluoroquinolones (-1.78%) and beta-lactams (-6.40%).
• No significant differences in adverse events were observed between antibiotic groups.
• Automated OMOP feature extraction methods yielded results consistent with manual curation.

Conclusions:

• Current national guidelines for uncomplicated UTI treatment remain relevant and effective.
• First-line antibiotics demonstrate favorable outcomes and safety profiles in contemporary practice.
• Automated data analysis methods enhance reproducibility and generalizability in observational studies.