Isolation, Identification, and Characterization of Bacteria from Urinary Tract Infections and Their Effect on Different Antibiotics
Keywords:
Urinary tract infections, Antibiotic resistance, antimicrobial stewardship, Random Forest model, Machine learning, Empirical therapy, Multidrug resistance.Abstract
Background: Urinary tract infections (UTIs) are among the most prevalent bacterial
infections worldwide, with Escherichia coli and Klebsiella spp. being the leading
causative agents. The increasing emergence of antibiotic resistance in uropathogens
has complicated empirical treatment strategies, necessitating continuous surveillance
of resistance trends. This study aims to isolate, identify, and characterize bacterial
strains from UTI patients, analyze their antibiotic susceptibility patterns, and explore
machine learning-based predictive models for resistance classification.
Methods: A descriptive cross-sectional study was conducted at Konaseema institute
of Medical Sciences and Research Foundation, Amalapuram, over a one-year period
(June 2023 – July 2024). A total of 1,720 urine samples were analyzed, of which 624
(36.2%) showed significant bacterial growth. Bacterial isolates were identified using
standard microbiological techniques, including culture on Blood Agar, MacConkey
Agar, and CLED Agar, Gram staining, and biochemical testing. Antibiotic
susceptibility testing (AST) was performed using the Kirby-Bauer disk diffusion
method, following CLSI guidelines. Statistical analysis included Chi-Square tests to
assess associations between bacterial species and resistance patterns, and a Random
Forest classification model to predict resistance trends based on susceptibility
profiles.
Results: Among the 624 culture-positive samples, E. coli (45.7%) was the most
prevalent uropathogen, followed by Klebsiella spp. (19.9%), Staphylococcus aureus
(13.6%), and Pseudomonas aeruginosa (7.4%). Antibiotic resistance rates were
highest among Non-Fermenters, particularly against β-lactam antibiotics.
Enterobacterales exhibited significant resistance to third-generation cephalosporins
and fluoroquinolones, whereas Gram-Positive Cocci demonstrated variable
resistance patterns, notably against β-lactams and macrolides.
Chi-Square analysis revealed no statistically significant association (p > 0.05)
between bacterial species and antibiotic resistance patterns, suggesting that resistance
trends may be influenced by factors beyond species classification. The Random
Forest model achieved an AUC of 1.00, demonstrating excellent discriminatory
power in predicting bacterial classification based on resistance profiles. CeftazidimeAvibactam, Levofloxacin, and Piperacillin-Tazobactam were identified as the most
influential antibiotics in resistance prediction.
Conclusion: This study highlights the high prevalence of multidrug-resistant
uropathogens, particularly among Non-Fermenters and Enterobacterales, reinforcing
the need for real-time susceptibility testing and antimicrobial stewardship programs.
The lack of significant species-resistance association emphasizes that predicting
antibiotic resistance requires broader epidemiological and molecular analyses rather
than relying solely on bacterial species. The successful application of machine
learning (Random Forest) in resistance prediction presents a promising approach for
future antimicrobial resistance surveillance. Further validation on larger datasets is
recommended to enhance predictive accuracy and clinical applicability.