ESCMID 2026: Updates on Epidemiology and Therapeutics of Urinary Tract Infections
Predicting Antibiotic Resistance Risk in Urinary Tract Infections Using Machine Learning to Support Safer Empirical Prescribing in Hospitalised Children and Adults
Presenter: Lina Aerts
This study developed and evaluated machine learning (ML) models to predict antibiotic resistance in urinary tract infections across adult and paediatric hospital populations. Electronic health records (2018–2025) from two Swiss hospitals were used. Models included demographic, clinical, and infection-related variables and were developed separately for adults and children, with assessment of cross-population transferability.
E. coli was the most common pathogen (52.3% adults; 66.9% children). Resistance prevalence varied widely: 4% (nitrofurantoin) to 49% (cefuroxime) in adults, and 2% (fosfomycin) to 24% (trimethoprim–sulfamethoxazole) in children.
Model performance:
- Adults: AUROC 0.65±0.01 (cotrimoxazole), 0.71±0.01 (ciprofloxacin), 0.76±0.01 (amoxicillin–clavulanate)
- Children: Lower and more variable performance, e.g. 0.69±0.14 (ciprofloxacin) and 0.60±0.16 (amoxicillin–clavulanate)
Prior resistance history was the strongest predictor, with male sex associated with slightly higher predicted risk. Adult-trained models showed partial transferability to paediatric data (e.g. AUROC 0.65 for amoxicillin–clavulanate).
Overall, ML models demonstrated moderate predictive ability for antibiotic resistance using routine clinical data.
Development of a Non-Invasion Intranasal Vaccine for Klebsiella Pneumoniae Urinary Tract Infection
Presenter: L. Kristopher Siu
This study evaluated an OmpK36-based vaccine targeting Klebsiella pneumoniae using subcutaneous and intranasal immunization approaches in a UTI model. Eight female BALB/c mice (5 weeks old) underwent a structured schedule including primary immunization, Boost I (day 14), and Boost II (day 28), followed by sample collection and sequential challenge phases. Blood samples and weight assessments were conducted at defined intervals before and after challenges.
Subcutaneous immunization with OmpK36 (without adjuvant) increased serum IgG levels, with titres rising from baseline to approximately 8 × 10⁶ after Boost II.
Intranasal immunization showed formulation-dependent responses:
- IgG (intranasal): The rOmpK36 + PELC + CpG formulation produced the highest titres (>10⁶) compared to control and other formulations.
- IgA (intranasal): The same formulation demonstrated the strongest mucosal immune response (measured by OD450), outperforming other groups.
Overall, intranasal immunization with PELC-OmpK36 induced protective immunity against K. pneumoniae-associated UTIs.
Incidence of Urinary Tract Infections Among Diabetic and Non-Diabetic Patients Following Treatment with SGLT-2 Inhibitors Compared with GLP-1 Receptor Agonists
Presenter: Oren Biham
This large real-world cohort study compared the incidence of urinary tract infections (UTIs) in patients initiating Sodium-glucose cotransporter 2 (SGLT2) inhibitors versus glucagon-like-peptide-1 receptor-agonists (GLP-1). A total of 275,073 patients were included (SGLT2i: 130,477; GLP-1: 144,596), with 191,408 diabetic and 83,665 non-diabetic patients. Overall, 16,740 UTI events were recorded.
In non-diabetic patients, UTI incidence was higher with SGLT2i compared to GLP-1 both before matching (65.22 vs 36.37 per 1,000 patient-years; RR 1.79 [95% CI 1.65–1.94]) and after matching (47.67 vs 40.22; RR 1.19 [95% CI 1.00–1.40]).
In patients with diabetes, UTI rates were similar between groups before matching (52.27 vs 51.87; RR 1.01 [95% CI 0.97–1.04]) and after matching (53.86 vs 52.62; RR 0.98 [95% CI 0.94–1.02]).
Overall, increased UTI risk with SGLT2i was observed in non-diabetic patients, while no difference was seen in those with diabetes.
Characterising Complicated Urinary Tract Infections (POS-Cuti): Preliminary Results of A Multinational Prospective Cohort (POST-Cuti / ECRAID)
Presenter: Jose M. Bravo-Ferrer
This multinational prospective cohort study evaluated characteristics and outcomes of patients with complicated UTI (cUTI), urinary-tract bacteraemia, and acute pyelonephritis. A total of 4,980 patients from 39 hospitals across 16 European countries were included. Of these, 83.7% had cUTI (4,166), 40.2% bacteraemia (2,001), and 33.4% pyelonephritis (1,661), with overlapping conditions.
Among cUTI patients, 36.9% had bacteraemia and 29.2% pyelonephritis. Among bacteraemia cases, 76.9% had cUTI and 27.2% pyelonephritis. Among pyelonephritis cases, 73.1% had cUTI and 32.8% bacteraemia. 30-day mortality was 6.5% in cUTI, 4% in bacteraemia (p=0.2 vs cUTI), and 2.2% in pyelonephritis (p=0.01 vs cUTI). Rates of recurrence, reinfection, and readmission were similar across groups.
Overall, significant overlap exists between these conditions, with comparable outcomes except for lower mortality in pyelonephritis.
Clinical Effectiveness of Four Different Short-Course Antibiotic Regimens in Adult Women with Bacteriological Confirmation of Urinary Tract Infection
Presenter: Ramon Monfa
This phase IV multicentre randomised trial compared the effectiveness of four short-course antibiotic regimens in women with uncomplicated UTI and confirmed positive urine cultures.
Among 768 women, 421 (57.0%) had positive cultures. Distribution across treatment arms was: single-dose fosfomycin 109 (58.3%), two-dose fosfomycin 102 (55.1%), nitrofurantoin 91 (49.7%), and pivmecillinam 119 (65%).
At day 7, clinical cure rates were significantly higher with all short-course regimens compared to single-dose fosfomycin. Absolute differences were:
- Nitrofurantoin: +34.9% (95% CI 22.9–46.8%)
- Pivmecillinam: +22.0% (95% CI 9.6–34.5%)
- Two-dose fosfomycin: +13.7% (95% CI 0.3–27.2%)
Nitrofurantoin showed the highest clinical effectiveness, while single-dose fosfomycin was the least effective, therefore it’s a role as a first-line treatment for uncomplicated UTI should be re-evaluated.
The Effectiveness of a Weekly Antimicrobial Stewardship Intervention on Antibiotic Prescriptions in Dutch Nursing Homes: A Randomised Controlled Trial
Presenter: Laura Van Buul
This multicentre randomized trial evaluated whether an antimicrobial stewardship (AMS) intervention reduced inappropriate antibiotic prescribing in nursing homes. A total of 57 clinicians across 8 nursing home organisations were randomized (intervention: 29; control: 28), contributing 1998.9 clinician-weeks. The intervention involved weekly audit and feedback over 36 weeks.
A total of 951 antibiotic prescriptions were recorded: 430 (45.2%) in the control group and 521 (54.8%) in the intervention group. Amoxicillin-clavulanate was most commonly prescribed (29.5% control; 33.2% intervention), followed by nitrofurantoin (21.2%; 19.6%) and ciprofloxacin (11.2%; 12.5%). Other antibiotics included fosfomycin (10.7%; 8.3%), flucloxacillin (9.1%; 9.4%), Trimethoprim/sulfamethoxazole (TMP-SMX) (7.9%; 6.1%), and amoxicillin (~2% in both groups). Most prescriptions were oral (98.8% control; 98.7% intervention), with minimal intramuscular (1.2%; 0.9%) and intravenous use (0; 0.4%).
UTI was the most common indication (60.9% control; 58.0% intervention), followed by RTI (28.1%; 28.0%) and SSTI (10.9%; 14.0%). Inappropriate prescriptions were 49.5% (258/521) in the intervention arm and 59.8% (257/430) in the control arm. The incidence rate (IR) was 14.2 vs 15.0 per prescriber per week, with an incidence rate ratio (IRR) of 0.95 (95% CI 0.79–1.13).
Overall, weekly audit and feedback did not significantly reduce inappropriate antibiotic prescribing.
ESCMID 2026, 17-21 April, Munich, Germany.
