ESCMID 2026: Highlights from Day 2
Epidemiological Patterns of Carbapenem Resistance in Solid Organ Transplantation: A Contemporary Multi-Centre Analysis
Presenter: Ayse Hande Arslan
This multicentre retrospective study evaluated epidemiological patterns of carbapenem-resistant Gram-negative (CRGN) infections across solid organ transplant (SOT) recipients.
A total of 460 CRGN infection episodes (2014–2023) from seven transplant centres were analysed. The most common pathogens were Klebsiella spp (47%), followed by Acinetobacter spp (27%) and E. coli (13.7%). Considerable inter-centre variability was observed in pathogen distribution, diagnostics, and prophylaxis practices. Only two centres performed routine pre-transplant rectal screening.
Kidney transplant recipients accounted for ~65% of infections. Urinary tract infections were the most frequent overall. Liver recipients had higher rates of early (<3 months) and severe infections, particularly bloodstream and respiratory infections, often linked to ICU-associated Acinetobacter. In heart and combined kidney–liver transplants, wound/tissue infections were more common, suggesting surgical site involvement.
Two key clinical pathways were identified: early urinary Klebsiella infections in kidney recipients, and late ICU-associated respiratory Acinetobacter infections linked with higher mortality.
Thirty-day mortality was 25.9%, and ninety-day mortality was 32.6%, with the highest mortality seen in ICU-associated respiratory infections and early bloodstream infections in liver recipients. Overall, CRGN infections showed distinct patterns based on organ type, timing, and care setting.
Aztreonam/Avibactam Resistance Due To Synergistic Interplay of ampC-Related Mutations in Carbapenem-Resistant Enterobacter Spp.
Presenter: Yang Zeng
This study investigated resistance mechanisms to aztreonam-avibactam (ATM-AVI) in metallo-β-lactamase-producing Enterobacter xiangfangensis. ATM-AVI-resistant mutants were generated through induction experiments, followed by genome sequencing and expression analysis. Four resistant mutants were identified.
All mutants harboured ampC mutations causing amino acid substitutions at position 150 (Tyr150Cys or Tyr150Ser), along with ampD mutations leading to either amino acid substitution (Cys108Tyr) or truncation (Glu106X, Trp95X). Two mutants also had mutations in the ampC attenuator region.
Minimum inhibitory concentration (MIC, mg/L) increased progressively with combined mutations:
- AmpC (Tyr150Ser) + AmpD (Glu106X): 32/4
- AmpC (Tyr150Cys) + AmpD (Cys108Tyr): 64/4
- AmpC (Tyr150Cys) + AmpD (Cys108Tyr) + attenuator (+13G>T): 256/4
- AmpC (Tyr150Cys) + AmpD (Trp95X) + attenuator (+7T>C): 256/4
Individually, AmpC substitution or ampD mutation reduced susceptibility, but combined mutations conferred resistance. Attenuator mutations increased ampC expression by 2–4 fold, further elevating resistance levels. Overall, resistance was driven by a combination of structural changes in AmpC, regulatory gene mutations, and increased gene expression.
Antimicrobial Resistance and Risk-Stratified Empiric Therapy in Community-Onset Sepsis
Presenter: Winnie Lee
This retrospective study evaluated empiric antimicrobial prescribing patterns and outcomes in community-onset sepsis across a large UK hospital network.
A total of 3,338 patients met sepsis criteria. Overall, 30-day mortality was 11.1%, increasing with higher National Early Warning Score 2 (NEWS2) scores (p<0.05). Extended spectrum β-lactamase-producing pathogens (ESBL) accounted for 7% of cases.
Empiric therapy patterns:
Ceftriaxone was the most commonly used empiric antibiotic (12.5%). Dual therapy was frequent, most commonly amikacin + ceftriaxone in both ESBL (20.6%) and non-ESBL (30.6%) infections. Antibiotic regimens were frequently modified within 48 hours, with 17.9% of patients undergoing escalation in WHO AWaRe category within 24 hours.
ESBL subgroup:
30-day mortality was 12.1%, with no significant difference compared to non-ESBL patients. Ceftriaxone remained the most common empiric agent (15.8%), while meropenem use was 5.1% higher than in non-ESBL cases. AWaRe escalation within 24 hours occurred in 21.6% of ESBL-positive patients.
Overall, empiric prescribing was variable, with frequent early escalation and adjustments in therapy.
Modeling A Point-Of-Care Diagnostic Panel Implementation for Rapid Bloodstream Infection Identification in Remote Settings: A Cost-Effectiveness Analysis
Presenter: Kristian Bagge
This retrospective study evaluated the potential impact of implementing a rapid molecular diagnostic panel (QIAstat-Dx® BCID Plus AMR) for blood culture analysis in a remote hospital setting.
A total of 46,546 blood cultures from 6,119 patients (2020–2024) were analysed, with 2,392 positive bottles from 851 patients, representing 996 bacteremic events. The mean workflow times were: sampling to positivity 18.6 hours, positivity to microscopy 6.8 hours, and microscopy to MALDI-TOF 11.2 hours, resulting in a total delay of ~18 hours post-positivity.
The rapid panel would have detected 79.9% (796/996) of initial bacteremic events. Among follow-up samples (day 2–30), 95.2% (20/21) of discrepant Gram stain cases and 86.7% (78/90) of consistent cases were detectable, although only 4/78 represented new findings.
Implementation of the rapid test could reduce identification time by at least 10 hours. The incremental cost-effectiveness ratio (ICER) was €124.6 per sample achieving rapid identification. Overall, rapid molecular testing demonstrated high detection rates with reduced turnaround time in a remote setting.
Rapid Pathogen and Antimicrobial Resistance Detection in Bloodstream Infections Using Short Subculture and Nanopore Sequencing
Presenter: Mariana Pitombeira Liborio
This experimental study evaluated early pathogen identification and resistance detection in bloodstream infections (BSIs) using short incubation and Oxford Nanopore sequencing. Blood samples spiked with CTX-M-15-producing E. coli (EC958) (<50 CFU/mL) were incubated and analysed hourly up to 7 hours. Two extraction methods were compared: MolYsis + DNeasy (B5+D) and MolYsis + BiOstic (B5+B).
Bacterial growth increased with time, reaching 10² CFU/mL at 3 hours and 10⁶ CFU/mL at 7 hours. With increasing incubation, cycle threshold values decreased for both methods, indicating higher bacterial load.
Using the B5+B method, E. coli was detected within 5 minutes of sequencing at 5 hours incubation (~10⁴ CFU/mL), and the resistance gene CTX-M-15 was identified in just over 3 hours of sequencing at 7 hours (~10⁶ CFU/mL). In contrast, the B5+D method did not detect the organism within the first hour of sequencing for any sample. Overall, rapid detection of pathogen and resistance markers was achievable within hours using optimized extraction and sequencing workflows.
Antimicrobial Stewardship Implementation Across Intensive Care Units: Insights from A National Survey
Presenter: Anna Marthe Van Boekel
This national survey evaluated the implementation and structure of antimicrobial stewardship programs (ASPs) across ICUs in the Netherlands.
A total of 213 respondents from 69 hospitals (100% hospital-level response) participated. ICU sizes were 4–9 beds (36%), 10–19 beds (41%), and ≥20 beds (23%). ASPs were reported in 100% of ICUs; however, only 45% were formally organized, while 55% had informal implementation of ASP elements.
Common AMS protocols included perioperative prophylaxis (~90%), therapeutic drug monitoring (~85%), initiation of antimicrobial therapy (~85%), IV-to-oral switch (~80%), optimisation/de-escalation (~70%), and stopping therapy (~60%).
Antimicrobial restriction strategies were implemented in all ICUs, most commonly via post-authorization review (~75%), post-prescription review (~60%), and automatic alerts (~30%). Other approaches included pre-authorization (~30%), post-authorization first prescription (~35%), and indication forms (~25%). Overall, while AMS interventions were universally present, there was substantial variability in the type, number, and level of formalization of ASPs across ICUs.
Optimising Intravenous to Oral Antibiotic Switch in Intensive Care Units: A Multi-Site Audit Using UKHSA IVOS Criteria Across Three West Midlands NHS Trusts
Presenter: Nirlep Agravedi
This multi-site audit evaluated IV-to-oral switch (IVOS) practices in ICU patients across three NHS Trusts using UKHSA criteria.
A total of 37 patients were assessed (MMUH (Midland Metropolitan University Hospital) n=11, Dudley n=10, QEH (Queen Elizabeth Hospital) n=16). Overall, only 21.6% (8/37) of patients were switched or eligible for IV-to-oral conversion, with variation across sites (QEH 37.5%, Dudley 20.0%, MMUH 9%).
Adherence to individual IVOS criteria varied: safe swallow (81.1%), absence of vomiting (97.3%), and temperature stability (75.7%) were commonly met, whereas clinical improvement (59.5%) and adequate gastrointestinal function (59.5%) were less frequent.
Special infection considerations requiring prolonged IV therapy were present in 67.6% of patients. Key barriers to IVOS included clinical deterioration and gastrointestinal dysfunction. Overall, IV-to-oral switch rates were suboptimal, with notable inter-site variation, highlighting the need for more consistent assessment practices.
Microbis: A Novel Artificial Intelligence and Machine Learning-Based Digital Platform for Rapid Bacterial Identification and Antimicrobial Resistance Prediction to Strengthen Diagnostics, Clinical Decision-Making, and Surveillance in Resource-Limited Healthcare Settings Across Africa
Presenter: Anna Musuvii Martin
This digital health study evaluated MicroBIS, an AI/ML-based platform designed to improve bacterial identification and antimicrobial resistance (AMR) prediction using biochemical assays and resistance datasets.
The platform was developed using a web-based MERN framework with ML models (Random Forest, PyBact). Simulated biochemical datasets (8-test and 20-test panels) and AMR data from the Pfizer ATLAS database were used for training and validation.
For bacterial identification, accuracy was 79% with the 20-test panel and 31% with the 8-test panel. When predicting the top five likely species, accuracy increased to 99% (20-test panel) and 74% (8-test panel). Certain biochemical markers, such as citrate utilization, showed higher predictive value.
For antimicrobial resistance prediction, the ML model achieved 56% accuracy in classifying susceptibility profiles. Overall, the platform demonstrated feasibility in combining biochemical testing with AI/ML for bacterial identification and AMR prediction.
ESCMID 2026, 17-21 April, Munich, Germany.
