Predicting future hospital antimicrobial resistance prevalence using machine learning.

BACKGROUND: Predicting antimicrobial resistance (AMR), a top global health threat, nationwide at an aggregate hospital level could help target interventions. Using machine learning, we exploit historical AMR and antimicrobial usage to predict future AMR. METHODS: Antimicrobial use and AMR prevalence in bloodstream infections in hospitals in England were obtained per hospital group (Trust) and financial year (FY, April-March) for 22 pathogen-antibiotic combinations (FY2016-2017 to FY2021-2022). Extreme Gradient Boosting (XGBoost) model predictions were compared to the previous value taken forwards, the difference between the previous two years taken forwards and linear trend forecasting (LTF). XGBoost feature importances were calculated to aid interpretability. RESULTS: Here we show that XGBoost models achieve the best predictive performance. Relatively limited year-to-year variability in AMR prevalence within Trust-pathogen-antibiotic combinations means previous value taken forwards also achieves a low mean absolute error (MAE), similar to or slightly higher than XGBoost. Using the difference between the previous two years taken forward or LTF performs consistently worse. XGBoost considerably outperforms all other methods in Trusts with a larger change in AMR prevalence from FY2020-2021 (last training year) to FY2021-2022 (held-out test set). Feature importance values indicate that besides historical resistance to the same pathogen-antibiotic combination as the outcome, complex relationships between resistance in different pathogens to the same antibiotic/antibiotic class and usage are exploited for predictions. These are generally among the top ten features ranked according to their mean absolute SHAP values. CONCLUSIONS: Year-to-year resistance has generally changed little within Trust-pathogen-antibiotic combinations. In those with larger changes, XGBoost models can improve predictions, enabling informed decisions, efficient resource allocation, and targeted interventions.

Antibiotics play an important role in treating serious bacterial infections. However, with the increased usage of antibiotics, they are becoming less effective. In our study, we use machine learning to learn from past antibiotic resistance and usage in order to predict what resistance will look like in the future. Different hospitals across England have very different resistance levels, however, within each hospital, these levels remain stable over time. When larger changes in resistance occurred over time in individual hospitals, our methods were able to predict these. Understanding how much resistance there is in hospital populations, and what may occur in the future can help determine where resources and interventions should be directed.

GBS vaccines in the UK: a round table discussion.

Background: Group B streptococcus (GBS) remains a leading cause of infant sepsis, meningitis and death despite intrapartum antibiotic prophylaxis. A vaccine is urgently required, and two candidates are in advanced clinical trials. For successful GBS vaccine implementation, especially if a vaccine is licensed based on an immunological threshold, there must be cross-sector engagement, effective advocacy, robust plans for phase IV studies and equitable access. Meeting: A round-table discussion, held at St George's University of London, reviewed the current position of GBS vaccines in the UK context, focusing on phase IV plans, convening a diverse group of stakeholders from across the UK, with a role in GBS vaccine licensure, advocacy, implementation or effectiveness evaluation.Presentations outlined the latest UK epidemiology, noting the rising infant invasive GBS (iGBS) infection rates from 1996 to 2021 for both early and late onset disease, with the highest disease rates in Black infants (1.1/1000 livebirths vs white infants (0.81/1000 livebirths). Potential coverage of the candidate vaccines was high (>95%). Regulatory input suggested that EU regulators would consider waiving the need for a pre-licensure efficacy study if a putative correlate of protection could be adequately justified. Phase IV study methodologies for a GBS vaccine were considered, largely based on previous UK maternal vaccine assessments, such as a nationwide cohort study design using a vaccine register and a maternal services dataset. Other strategies were also discussed such as a cluster or stepped-wedge randomised trial to evaluate implementation outcomes. Opportunities for advocacy, education and engagement with additional key partners were discussed and identified. Conclusions: With an approved GBS vaccine a near possibility, planning of phase IV studies and identification of critical barriers to implementation are urgently needed. Cross-sector engagement is essential and will facilitate a successful pathway.

The laboratory investigation, management, and infection prevention and control of Candida auris: a narrative review to inform the 2024 national guidance update in England.

The emergent fungal pathogen Candida auris is increasingly recognised as an important cause of healthcare-associated infections globally. It is highly transmissible, adaptable, and persistent, resulting in an organism with significant outbreak potential that risks devastating consequences. Progress in the ability to identify C. auris in clinical specimens is encouraging, but laboratory diagnostic capacity and surveillance systems are lacking in many countries. Intrinsic resistance to commonly used antifungals, combined with the ability to rapidly acquire resistance to therapy, substantially restricts treatment options and novel agents are desperately needed. Despite this, outbreaks can be interrupted, and mortality avoided or minimised, through the application of rigorous infection prevention and control measures with an increasing evidence base. This review provides an update on epidemiology, the impact of the COVID-19 pandemic, risk factors, identification and typing, resistance profiles, treatment, detection of colonisation, and infection prevention and control measures for C. auris. This review has informed a planned 2024 update to the United Kingdom Health Security Agency (UKHSA) guidance on the laboratory investigation, management, and infection prevention and control of Candida auris. A multidisciplinary response is needed to control C. auris transmission in a healthcare setting and should emphasise outbreak preparedness and response, rapid contact tracing and isolation or cohorting of patients and staff, strict hand hygiene and other infection prevention and control measures, dedicated or single-use equipment, appropriate disinfection, and effective communication concerning patient transfers and discharge.

Rapid expansion and international spread of M1UK in the post-pandemic UK upsurge of Streptococcus pyogenes.

The UK observed a marked increase in scarlet fever and invasive group A streptococcal infection in 2022 with severe outcomes in children and similar trends worldwide. Here we report lineage M1UK to be the dominant source of invasive infections in this upsurge. Compared with ancestral M1global strains, invasive M1UK strains exhibit reduced genomic diversity and fewer mutations in two-component regulator genes covRS. The emergence of M1UK is dated to 2008. Following a bottleneck coinciding with the COVID-19 pandemic, three emergent M1UK clades underwent rapid nationwide expansion, despite lack of detection in previous years. All M1UK isolates thus-far sequenced globally have a phylogenetic origin in the UK, with dispersal of the new clades in Europe. While waning immunity may promote streptococcal epidemics, the genetic features of M1UK point to a fitness advantage in pathogenicity, and a striking ability to persist through population bottlenecks.

Surge of lower respiratory tract group A streptococcal infections in England in winter 2022: epidemiology and clinical profile.

BACKGROUND: Following low incidence of invasive group A streptococcal (iGAS) infections during the COVID-19 pandemic, marked increases were noted in many countries during 2022, particularly in children. In November 2022, severe presentations of lower respiratory tract infections (LRTIs), including empyema, were notified by clinicians across the UK. UKHSA investigated this rise with the aim of informing clinical management and public health response. METHODS: We undertook a case-series analysis using multiple routine data sources, exempted from ethics approval or patient consent. We identified iGAS cases in England in children younger than 15 years with an LRTI reported between Oct 1 and Dec 21, 2022, using UKHSA laboratory surveillance data (GAS detected in LRT specimens) and notifications by clinicians and Health Protection Teams (HPTs). Symptoms, diagnoses, health-care interactions, and outcome (death or recovery) data were obtained from HPT case management notes, the National Child Mortality Database, and the NHS Digital Emergency Care Dataset. FINDINGS: We identified 147 cases of LRTI iGAS in children across England (77 [52%] male, 70 [48%] female; median age 4 years [IQR 2-6]). Predominant ethnicities were White (74 [65%] of 113 with known ethnicity) and Asian (18 [16%] of 113). Most reported symptoms were fever (90 [75%] of 120 children with ≥1 symptom) and cough (60 [50%] of 120), and 71 (48%) of all 147 children had a diagnosed respiratory viral coinfection (most commonly hMPV and RSV). 127 (86%) of children attended an emergency department, 31% (n=36/114 with onset date) at least twice within 21 days after symptom onset. 37 (25%) of 147 children died, with a median time from symptom onset to death of 4 days (IQR 3-7). Of 32 children with sample dates, 16 (84%) were tested for GAS on or after the day they died. Over half of deaths (21 [57%] of 37 deaths) occurred in the community after rapid deterioration, of whom 18 had previous contact with health-care services documented. INTERPRETATION: The UK saw an unusual rise in iGAS LRTIs in children in late 2022. One in four cases died, over half in the community. Non-specific symptoms, viral symptoms, or positive virology might have lowered suspicion of bacterial infection. Although the use of multiple available data sources expedited the analysis, varying data completeness limited interpretation. Our study highlights the need for earlier detection and identification of effective measures to prevent death. FUNDING: None.

Nitrofurantoin resistance as an indicator for multidrug resistance: an assessment of Escherichia coli urinary tract specimens in England, 2015-19.

Objectives: To determine whether MDR occurs more frequently in nitrofurantoin-resistant Escherichia coli urinary isolates in England, compared with nitrofurantoin-susceptible isolates. Methods: Using routine E. coli urine isolate antibiotic susceptibility laboratory surveillance data for England, 2015-19 inclusive, the percentage of MDR or XDR phenotype was estimated for nitrofurantoin-susceptible and nitrofurantoin-resistant laboratory-reported urinary tract samples by region, patient sex and age group. Results: Resistance to nitrofurantoin among E. coli urinary samples decreased slightly year on year from 2.9% in 2015 to 2.3% in 2019. Among E. coli UTIs tested for nitrofurantoin susceptibility and  ≥3 additional antibiotics, the percentage that were MDR was consistently 15%-20% percentage points higher for nitrofurantoin-resistant isolates compared with nitrofurantoin-susceptible isolates. Similarly, the percentage of isolates with an XDR phenotype was higher among nitrofurantoin-resistant versus -susceptible isolates (8.7% versus 1.4%, respectively, in 2019); this disparity was greater in male patients, although variation was seen by age group in both sexes. Regional variation was also noted, with the highest MDR percentage amongst nitrofurantoin-resistant E. coli urinary samples in the London region (36.7% in 2019); the lowest was in the North East (2019: 16.9%). Conclusions: MDR and XDR phenotypes occur more frequently in nitrofurantoin-resistant E. coli urinary isolates in England, compared with nitrofurantoin-susceptible isolates. However, nitrofurantoin resistance is low (<3%) overall. This latest study provides important insights into trends in nitrofurantoin resistance and MDR, which is of particular concern for patients ≥75 years old and those who are male. It also emphasises geographical heterogeneities within England in nitrofurantoin resistance and MDR.

Determining the impact of professional body recommendations on the screening of acquired carbapenemase-producing Enterobacterales in England.

Introduction: Acquired carbapenemase-producing Gram-negative bacteria are an increasing public health concern globally and have been mandatory to report in England since October 2020. However, in light of the COVID-19 (SARS-CoV-2) pandemic, the Royal College of Pathologists (RCPath) released new guidance "for reducing the need for screening of CRE (carbapenem-resistant Enterobacterales) […] in low-risk areas", without defining "low risk". Methods: To assess the impact of the RCPath recommendations on screening of carbapenemase-producing Enterobacterales (CPE), an online Select Survey was sent to all NHS acute hospitals in England. The initial survey distribution was between March and April 2021 and the survey was relaunched between November 2021 and March 2022. Results: In total, 54 hospitals completed the survey, representing 39.1% of 138 eligible Trusts. All hospitals had a CPE screening policy in place, and the majority of these reflect UKHSA's Framework of actions to contain CPE. Of the 23 hospitals who reported a reduction in CPE screening, only three (13.0%) indicated that this was due to the RCPath recommendations, with 21 (91.3%) indicating that there had been a natural reduction in the number of patients admitted to the Trust who would have previously been screened due to the COVID-19 pandemic. Conclusion: For most surveyed hospitals, CPE screening was not reduced due to the RCPath recommendations. However, the results highlighted that there is a large amount of individual variation in CPE screening practices and diagnostic testing between hospitals.

Molecular characterization of the prototrophic Salmonella mutants defective for intraepithelial replication.

Three MudJ prototrophs demonstrated that intracellular replication is a Salmonella virulence trait (K. Y. Leung and B. B. Finlay, Proc. Natl. Acad. Sci. USA, 88:11470-11474, 1991). mutS and mutH are disrupted in mutants 3-11 and 12-23, and ssaQ is disrupted in mutant 17-21. Further analysis revealed that loss of Salmonella pathogenicity island 2 function underlies the intracellular replication defect of 3-11 and 17-21.