A Clinical Decision Tree to Predict Whether a Bacteremic Patient Is Infected With an Extended-Spectrum β-Lactamase-Producing Organism

Retrospective validation study of a machine learning-based software for empirical and organism-targeted antibiotic therapy selection

Errors in antibiotic prescriptions are frequent, often resulting from the inadequate coverage of the infection-causative microorganism. The efficacy of iAST, a machine-learning-based software offering empirical and organism-targeted antibiotic recommendations, was assessed. The study was conducted in a 12-hospital Spanish institution. After model fine-tuning with 27,531 historical antibiograms, 325 consecutive patients with acute infections were selected for retrospective validation. The primary endpoint was comparing each of the top three of iAST's antibiotic recommendations' success rates (confirmed by antibiogram results) with the antibiotic prescribed by the physicians. Secondary endpoints included examining the same hypothesis within specific study population subgroups and assessing antibiotic stewardship by comparing the percentage of antibiotics recommended that belonged to different World Health Organization AWaRe groups within each arm of the study. All of iAST first three recommendations were non-inferior to doctor prescription in the primary endpoint analysis population as well as the secondary endpoint. The overall success rate of doctors' empirical treatment was 68.93%, while that of the first three iAST options was 91.06% (P < 0.001), 90.63% (P < 0.001), and 91.06% (P < 001), respectively. For organism-targeted therapy, the doctor's overall success rate was 84.16%, and that of the first three ranked iAST options was 97.83% (P < 0.001), 94.09% (P < 0.001), and 91.30% (P < 0.001), respectively. In empirical therapy, compared to physician prescriptions, iAST demonstrated a greater propensity to recommend access antibiotics, fewer watch antibiotics, and higher reserve antibiotics. In organism-targeted therapy, iAST advised a higher utilization of access antibiotics. The present study demonstrates iAST accuracy in predicting antibiotic susceptibility, showcasing its potential to promote effective antibiotic stewardship.

CLINICAL TRIALS

This study is registered with ClinicalTrials.gov as NCT06174519.

Predicting the immunomodulatory activity of probiotic lactic acid bacteria using supervised machine learning in a Cornu aspersum snail model

In the process of screening for probiotic strains, there are no clearly established bacterial phenotypic markers which could be used for the prediction of their in vivo mechanism of action. In this work, we demonstrate for the first time that Machine Learning (ML) methods can be used for accurately predicting the in vivo immunomodulatory activity of probiotic strains based on their cell surface phenotypic features using a snail host-microbe interaction model. A broad range of snail gut presumptive probiotics, including 240 new lactic acid bacterial strains (Lactobacillus, Leuconostoc, Lactococcus, and Enterococcus), were isolated and characterized based on their capacity to withstand snails' gastrointestinal defense barriers, such as the pedal mucus, gastric mucus, gastric juices, and acidic pH, in association with their cell surface hydrophobicity, autoaggregation, and biofilm formation ability. The implemented ML pipeline predicted with high accuracy (88 %) strains with a strong capacity to enhance chemotaxis and phagocytic activity of snails' hemolymph cells, while also revealed bacterial autoaggregation and cell surface hydrophobicity as the most important parameters that significantly affect host immune responses. The results show that ML approaches may be useful to derive a predictive understanding of host-probiotic interactions, while also highlighted the use of snails as an efficient animal model for screening presumptive probiotic strains in the light of their interaction with cellular innate immune responses.

Artificial intelligence and prescription of antibiotic therapy: present and future

INTRODUCTION

In the past few years, the use of artificial intelligence in healthcare has grown exponentially. Prescription of antibiotics is not exempt from its rapid diffusion, and various machine learning (ML) techniques, from logistic regression to deep neural networks and large language models, have been explored in the literature to support decisions regarding antibiotic prescription.

AREAS COVERED

In this narrative review, we discuss promises and challenges of the application of ML-based clinical decision support systems (ML-CDSSs) for antibiotic prescription. A search was conducted in PubMed up to April 2024.

EXPERT OPINION

Prescribing antibiotics is a complex process involving various dynamic phases. In each of these phases, the support of ML-CDSSs has shown the potential, and also the actual ability in some studies, to favorably impacting relevant clinical outcomes. Nonetheless, before widely exploiting this massive potential, there are still crucial challenges ahead that are being intensively investigated, pertaining to the transparency of training data, the definition of the sufficient degree of prediction explanations when predictions are obtained through black box models, and the legal and ethical framework for decision responsibility whenever an antibiotic prescription is supported by ML-CDSSs.

Development and external validation of a model to predict multidrug-resistant bacterial infections in patients with cirrhosis

With the increasing rate of infections caused by multidrug-resistant organisms (MDRO), selecting appropriate empiric antibiotics has become challenging. We aimed to develop and externally validate a model for predicting the risk of MDRO infections in patients with cirrhosis.

METHODS

We included patients with cirrhosis and bacterial infections from two prospective studies: a transcontinental study was used for model development and internal validation (n = 1302), and a study from Argentina and Uruguay was used for external validation (n = 472). All predictors were measured at the time of infection. Both culture-positive and culture-negative infections were included. The model was developed using logistic regression with backward stepwise predictor selection. We externally validated the optimism-adjusted model using calibration and discrimination statistics and evaluated its clinical utility.

RESULTS

The prevalence of MDRO infections was 19% and 22% in the development and external validation datasets, respectively. The model's predictors were sex, prior antibiotic use, type and site of infection, MELD-Na, use of vasopressors, acute-on-chronic liver failure, and interaction terms. Upon external validation, the calibration slope was 77 (95% CI .48-1.05), and the area under the ROC curve was .68 (95% CI .61-.73). The application of the model significantly changed the post-test probability of having an MDRO infection, identifying patients with nosocomial infection at very low risk (8%) and patients with community-acquired infections at significant risk (36%).

CONCLUSION

This model achieved adequate performance and could be used to improve the selection of empiric antibiotics, aligning with other antibiotic stewardship program strategies.

Epidemiology of Clostridioides difficile infection at one hospital 10 years after an outbreak of the epidemic C. difficile strain BI/027: changing strain prevalence, antimicrobial susceptibilities, and patient antibiotic exposures

In contrast to the epidemiology 10 years earlier at our hospital when the epidemic restriction endonuclease analysis (REA) group strain BI accounted for 72% of Clostridioides difficile isolates recovered from first-episode C. difficile infection (CDI) cases, BI represented 19% of first-episode CDI isolates in 2013-2015. Two additional REA group strains accounted for 31% of isolates (Y, 16%; DH, 12%). High-level resistance to fluoroquinolones and azithromycin was more common among BI isolates than among DH, Y, and non-BI/DH/Y isolates. Multivariable analysis revealed that BI cases were 2.47 times more likely to be associated with fluoroquinolone exposure compared to non-BI cases (95% confidence interval [CI]: 1.12-5.46). In addition, the odds of developing a CDI after third- or fourth-generation cephalosporin exposure was 2.83 times for DH cases than for non-DH cases (95% CI: 1.06-7.54). Fluoroquinolone use in the hospital decreased from 2005 to 2015 from a peak of 113 to a low of 56 antimicrobial days/1,000 patient days. In contrast, cephalosporin use increased from 42 to 81 antimicrobial days/1,000 patient days. These changes correlated with a decrease in geometric mean MIC for ciprofloxacin (61.03 to 42.65 mg/L, P = 0.02) and an increase in geometric mean MIC for ceftriaxone (40.87 to 86.14 mg/L, P < 0.01) among BI isolates. The BI strain remained resistant to fluoroquinolones, but an overall decrease in fluoroquinolone use and increase in cephalosporin use were associated with a decrease in the prevalence of BI, an increased diversity of C. difficile strain types, and the emergence of strains DH and Y.

Artificial intelligence applications in the diagnosis and treatment of bacterial infections

The diagnosis and treatment of bacterial infections in the medical and public health field in the 21st century remain significantly challenging. Artificial Intelligence (AI) has emerged as a powerful new tool in diagnosing and treating bacterial infections. AI is rapidly revolutionizing epidemiological studies of infectious diseases, providing effective early warning, prevention, and control of outbreaks. Machine learning models provide a highly flexible way to simulate and predict the complex mechanisms of pathogen-host interactions, which is crucial for a comprehensive understanding of the nature of diseases. Machine learning-based pathogen identification technology and antimicrobial drug susceptibility testing break through the limitations of traditional methods, significantly shorten the time from sample collection to the determination of result, and greatly improve the speed and accuracy of laboratory testing. In addition, AI technology application in treating bacterial infections, particularly in the research and development of drugs and vaccines, and the application of innovative therapies such as bacteriophage, provides new strategies for improving therapy and curbing bacterial resistance. Although AI has a broad application prospect in diagnosing and treating bacterial infections, significant challenges remain in data quality and quantity, model interpretability, clinical integration, and patient privacy protection. To overcome these challenges and, realize widespread application in clinical practice, interdisciplinary cooperation, technology innovation, and policy support are essential components of the joint efforts required. In summary, with continuous advancements and in-depth application of AI technology, AI will enable doctors to more effectivelyaddress the challenge of bacterial infection, promoting the development of medical practice toward precision, efficiency, and personalization; optimizing the best nursing and treatment plans for patients; and providing strong support for public health safety.

Machine-learning-based risk assessment tool to rule out empirical use of ESBL-targeted therapy in endemic areas

BACKGROUND

Antimicrobial stewardship focuses on identifying patients who require extended-spectrum beta-lactamase (ESBL)-targeted therapy. 'Rule-in' tools have been researched extensively in areas of low endemicity; however, such tools are inadequate for areas with high prevalence of ESBL-producing pathogens, as almost all patients will be selected.

AIM

To develop a machine-learning-based 'rule-out' tool suitable for areas with high levels of resistance.

METHODS

Gradient-boosted decision trees were used to train and validate a risk prediction model on data from 17,913 (45% ESBL) patients with Escherichia coli and Klebsiella pneumoniae in urine cultures. The predictive power of different sets of variables was evaluated using Shapley values to evaluate the contributions of variables.

FINDINGS

The model successfully identified patients with low risk of ESBL resistance in ESBL-endemic areas (area under receiver operating characteristic curve 0.72). When used to select the 30% of patients with the lowest predicted risk, the model yielded a negative predictive value ≥0.74. A simplified model with seven input features was found to perform nearly as well as the full model. This simplified model is freely accessible as a web application.

CONCLUSIONS

This study found that a risk calculator for antibiotic resistance can be a viable 'rule-out' strategy to reduce the use of ESBL-targeted therapy in ESBL-endemic areas. The robust performance of a version of the model with limited features makes the clinical use of such a tool feasible. This tool provides an important alternative in an era with growing rates of ESBL-producing pathogens, where some experts have called for empirical use of carbapenems as first-line therapy for all patients in areas with high prevalence of ESBL-producing pathogens.

Laboratory Tests, Bacterial Resistance, and Treatment Options in Adult Patients Hospitalized with a Suspected Urinary Tract Infection

Patients treated for systemic urinary tract infections commonly have nonspecific presentations, and the specificity of the results of the urinalysis and urine cultures is low. In the following narrative review, we will describe the widespread misuse of urine testing, and consider how to limit testing, the disutility of urine cultures, and the use of antibiotics in hospitalized adult patients. Automated dipstick testing is more precise and sensitive than the microscopic urinalysis which will result in false negative test results if ordered to confirm a positive dipstick test result. There is evidence that canceling urine cultures if the dipstick is negative (negative leukocyte esterase, and nitrite) is safe and helps prevent the overuse of urine cultures. Because of the side effects of introducing a urine catheter, for patients who cannot provide a urine sample, empiric antibiotic treatment should be considered as an alternative to culturing the urine if a trial of withholding antibiotic therapy is not an option. Treatment options that will decrease both narrower and wider spectrum antibiotic use include a period of watching and waiting before antibiotic therapy and empiric treatment with antibiotics that have resistance rates > 10%. Further studies are warranted to show the option that maximizes patient comfort and safety.

Efficacy of Random Forest Models in Predicting Multidrug-Resistant Gram-Negative Bacterial Nosocomial Infections Compared to Traditional Logistic Regression Models

This study evaluates whether random forest (RF) models are as effective as traditional Logistic Regression (LR) models in predicting multidrug-resistant Gram-negative bacterial nosocomial infections. Data were collected from 541 patients with hospital-acquired Gram-negative bacterial infections at two tertiary-level hospitals in Urumqi, Xinjiang, China, from August 2022 to November 2023. Relevant literature informed the selection of significant predictors based on patients' pre-infection clinical information and medication history. The data were split into a training set of 379 cases and a validation set of 162 cases, adhering to a 7:3 ratio. Both RF and LR models were developed using the training set and subsequently evaluated on the validation set. The LR model achieved an accuracy of 84.57%, sensitivity of 82.89%, specificity of 80.10%, positive predictive value of 84%, negative predictive value of 85.06%, and a Yoden index of 0.69. In contrast, the RF model demonstrated superior performance with an accuracy of 89.51%, sensitivity of 90.79%, specificity of 88.37%, positive predictive value of 87.34%, negative predictive value of 91.57%, and a Yoden index of 0.79. Receiver operating characteristic curve analysis revealed an area under the curve of 0.91 for the LR model and 0.94 for the RF model. These findings indicate that the RF model surpasses the LR model in specificity, sensitivity, and accuracy in predicting hospital-acquired multidrug-resistant Gram-negative infections, showcasing its greater potential for clinical application.

Tackling the Antimicrobial Resistance "Pandemic" with Machine Learning Tools: A Summary of Available Evidence

Antimicrobial resistance is recognised as one of the top threats healthcare is bound to face in the future. There have been various attempts to preserve the efficacy of existing antimicrobials, develop new and efficient antimicrobials, manage infections with multi-drug resistant strains, and improve patient outcomes, resulting in a growing mass of routinely available data, including electronic health records and microbiological information that can be employed to develop individualised antimicrobial stewardship. Machine learning methods have been developed to predict antimicrobial resistance from whole-genome sequencing data, forecast medication susceptibility, recognise epidemic patterns for surveillance purposes, or propose new antibacterial treatments and accelerate scientific discovery. Unfortunately, there is an evident gap between the number of machine learning applications in science and the effective implementation of these systems. This narrative review highlights some of the outstanding opportunities that machine learning offers when applied in research related to antimicrobial resistance. In the future, machine learning tools may prove to be superbugs' kryptonite. This review aims to provide an overview of available publications to aid researchers that are looking to expand their work with new approaches and to acquaint them with the current application of machine learning techniques in this field.

Machine learning to predict antimicrobial resistance: future applications in clinical practice?

INTRODUCTION

Machine learning (ML) is increasingly being used to predict antimicrobial resistance (AMR). This review aims to provide physicians with an overview of the literature on ML as a means of AMR prediction.

METHODS

References for this review were identified through searches of MEDLINE/PubMed, EMBASE, Google Scholar, ACM Digital Library, and IEEE Xplore Digital Library up to December 2023.

RESULTS

Thirty-six studies were included in this review. Thirty-two studies (32/36, 89 %) were based on hospital data and four (4/36, 11 %) on outpatient data. The vast majority of them were conducted in high-resource settings (33/36, 92 %). Twenty-four (24/36, 67 %) studies developed systems to predict drug resistance in infected patients, eight (8/36, 22 %) tested the performances of ML-assisted antibiotic prescription, two (2/36, 6 %) assessed ML performances in predicting colonization with carbapenem-resistant bacteria and, finally, two assessed national and international AMR trends. The most common inputs were demographic characteristics (25/36, 70 %), previous antibiotic susceptibility testing (19/36, 53 %) and prior antibiotic exposure (15/36, 42 %). Thirty-three (92 %) studies targeted prediction of Gram-negative bacteria (GNB) resistance as an output (92 %). The studies included showed moderate to high performances, with AUROC ranging from 0.56 to 0.93.

CONCLUSION

ML can potentially provide valuable assistance in AMR prediction. Although the literature on this topic is growing, future studies are needed to design, implement, and evaluate the use and impact of ML decision support systems.

Using machine learning algorithms to predict the prognosis of advanced nasopharyngeal carcinoma after intensity-modulated radiotherapy

BACKGROUND

The prognosis of advanced nasopharyngeal carcinoma (NPC) patients after intensity-modulated radiotherapy (IMRT) has not been well studied. We aimed to construct prognostic models for advanced NPC patients with stage III-IV after their first treatment with IMRT by using machine learning algorithms and to identify the most important predictors.

METHODS

A total of 427 patients treated in Meizhou People's Hospital in Guangdong province, China from January 1, 2013 to December 12, 2018 were enrolled in this study, with an average follow-up period of 7.16 years from July 2020 to March 2021. Candidate predictors were selected from demographics, clinical features, medical examinations and test results. Three machine learning algorithms were applied to construct advanced NPC prognostic models: logistic regression (LR), decision tree (DT), and random forest (RF). Area under the receiver operating characteristic curve (AUC) was used to evaluate the model performance. The important predictors of the optimal model for unfavourable prognosis were identified and ranked.

RESULTS

There were 50 (11.7%) NPC-related deaths observed in this study. The mean age of all participants was 49.39±11.29 years, of whom 299 (70.0%) were males. In general, RF showed the best predictive performance with the highest AUC (0.753, 95% CI: 0.609, 0.896), compared to LR (0.736, 95% confidence interval (CI): 0.590, 0.881), and DT (0.720, 95% CI: 0.520, 0.921). The six most important predictors identified by RF were Epstein-Barr virus deoxyribonucleic acid, aspartate aminotransferase, body mass index, age, blood glucose level, and alanine aminotransferase.

CONCLUSIONS

We proposed RF as a simple and accurate tool for the evaluation of the prognosis of advanced NPC patients after the treatment with IMRT in clinical settings.

Identifying the Optimal Look-back Period for Prior Antimicrobial Resistance Clinical Decision Support

BACKGROUND

Lack of consensus on the appropriate look-back period for multi-drug resistance (MDR) complicates antimicrobial clinical decision support. We compared the predictive performance of different MDR look-back periods for five common MDR mechanisms (MRSA, VRE, ESBL, AmpC, CRE).

METHODS

We mapped microbiological cultures to MDR mechanisms and labeled them at different look-back periods. We compared predictive performance for each look-back period-MDR combination using precision, recall, F1 scores, and odds ratios.

RESULTS

Longer look-back periods resulted in lower odds ratios, lower precisions, higher recalls, and lower delta changes in precision and recall compared to shorter periods. We observed higher precision with more information available to clinicians.

CONCLUSION

A previously positive MDR culture may have significant enough precision depending on the mechanism of resistance and varying information available. One year is a clinically relevant and statistically sound look-back period for empiric antimicrobial decision-making at varying points of care for the studied population.

The clinical impact of early detection of ESBL-producing Enterobacterales with PCR-based blood culture assays

BACKGROUND

Starting January 4, 2021, our health system core microbiology laboratory changed blood culture identification (BCID) platforms to ePlex BCID from BioFire BCID1 with the additional capability to detect the blaCTX-M-Type gene of ESBL-producing organisms. Clinical outcomes of ESBL bloodstream infections (BSI) after implementing ePlex BCID were unknown.

METHODS

Patients with ESBL BSI were compared pre and postimplementation of ePlex BCID in this 11-hospital retrospective analysis (BioFire BCID1 in 2019 vs ePlex BCID in 2021). The primary outcome was time from the Gram stain result to escalation to a carbapenem. Secondary outcomes included in-hospital mortality, 30-day readmission rate, length of stay (LOS), and the duration of antimicrobial therapy.

RESULTS

A total of 275 patients were analyzed. The median time of Gram stain result to escalation to carbapenem was reduced from 44.5 hours with BioFire BCID1 to 7.9 hours with ePlex BCID (P < .001). There were no significant differences in mortality, 30-day readmission, or LOS. The duration of antimicrobial therapy for ESBL BSI was lower in the ePlex BCID group (from 14.4 days to 12.7 days, P = .014).

CONCLUSIONS

Timely detection of the blaCTX-M-Type gene by BCID provides valuable information for the early initiation of appropriate and effective antimicrobial therapy. Although it was not associated with lower mortality, 30-day readmission, or LOS, it may have benefits such as decreasing antimicrobial exposure to patients.

Machine learning-based antibiotic resistance prediction models: An updated systematic review and meta-analysis

BACKGROUND

The widespread use of antibiotics has led to a gradual adaptation of bacteria to these drugs, diminishing the effectiveness of treatments.

OBJECTIVE

To comprehensively assess the research progress of antibiotic resistance prediction models based on machine learning (ML) algorithms, providing the latest quantitative analysis and methodological evaluation.

METHODS

Relevant literature was systematically retrieved from databases, including PubMed, Embase and the Cochrane Library, from inception up to December 2023. Studies meeting predefined criteria were selected for inclusion. The prediction model risk of bias assessment tool was employed for methodological quality assessment, and a random-effects model was utilised for meta-analysis.

RESULTS

The systematic review included a total of 22 studies with a combined sample size of 43,628; 10 studies were ultimately included in the meta-analysis. Commonly used ML algorithms included random forest, decision trees and neural networks. Frequently utilised predictive variables encompassed demographics, drug use history and underlying diseases. The overall sensitivity was 0.57 (95% CI: 0.42-0.70; p< 0.001; I2= 99.7%), the specificity was 0.95 (95% CI: 0.79-0.99; p< 0.001; I2 = 99.9%), the positive likelihood ratio was 10.7 (95% CI: 2.9-39.5), the negative likelihood ratio was 0.46 (95% CI: 0.34-0.61), the diagnostic odds ratio was 23 (95% CI: 7-81) and the area under the receiver operating characteristic curve was 0.78 (95% CI: 0.74-0.81; p< 0.001), indicating a good discriminative ability of ML models for antibiotic resistance. However, methodological assessment and funnel plots suggested a high risk of bias and publication bias in the included studies.

CONCLUSION

This meta-analysis provides a current and comprehensive evaluation of ML models for predicting antibiotic resistance, emphasising their potential application in clinical practice. Nevertheless, stringent research design and reporting are warranted to enhance the quality and credibility of future studies. Future research should focus on methodological innovation and incorporate more high-quality studies to further advance this field.

From Epidemiology of Community-Onset Bloodstream Infections to the Development of Empirical Antimicrobial Treatment-Decision Algorithm in a Region with High Burden of Antimicrobial Resistance

Antimicrobial-resistant (AMR) infections have increased in community settings. Our objectives were to study the epidemiology of community-onset bloodstream infections (BSIs), identify risk factors for AMR-BSI and mortality-related factors, and develop the empirical antimicrobial treatment-decision algorithm. All adult, positive blood cultures at the emergency room and outpatient clinics were evaluated from 08/2021 to 04/2022. AMR was defined as the resistance of organisms to an antimicrobial to which they were previously sensitive. A total of 1151 positive blood cultures were identified. There were 450 initial episodes of bacterial BSI, and 114 BSIs (25%) were AMR-BSI. Non-susceptibility to ceftriaxone was detected in 40.9% of 195 E. coli isolates and 16.4% among 67 K. pneumoniae isolates. A treatment-decision algorithm was developed using the independent risk factors for AMR-BSI: presence of multidrug-resistant organisms (MDROs) within 90 days (aOR 3.63), prior antimicrobial exposure within 90 days (aOR 1.94), and urinary source (aOR 1.79). The positive and negative predictive values were 53.3% and 83.2%, respectively. The C-statistic was 0.73. Factors significantly associated with 30-day all-cause mortality were Pitt bacteremia score (aHR 1.39), solid malignancy (aHR 2.61), and urinary source (aHR 0.30). In conclusion, one-fourth of community-onset BSI were antimicrobial-resistant, and one-third of Enterobacteriaceae were non-susceptible to ceftriaxone. Treatment-decision algorithms may reduce overly broad antimicrobial treatment.

Antimicrobial Multidrug Resistance: Clinical Implications for Infection Management in Critically Ill Patients

The increasing incidence of antimicrobial resistance (AMR) worldwide represents a serious threat in the management of sepsis. Due to resistance to the most common antimicrobials prescribed, multidrug-resistant (MDR) pathogens have been associated with delays in adequate antimicrobial therapy leading to significant increases in mortality, along with prolonged hospital length of stay (LOS) and increases in healthcare costs. In response to MDR infections and the delay of microbiological results, broad-spectrum antibiotics are frequently used in empirical antimicrobial therapy. This can contribute to the overuse and misuse of antibiotics, further promoting the development of resistance. Multiple measures have been suggested to combat AMR. This review will focus on describing the epidemiology and trends concerning MDR pathogens. Additionally, it will explore the crucial aspects of identifying patients susceptible to MDR infections and optimizing antimicrobial drug dosing, which are both pivotal considerations in the fight against AMR. Expert commentary: The increasing AMR in ICUs worldwide makes the empirical antibiotic therapy challenging in septic patients. An AMR surveillance program together with improvements in MDR identification based on patient risk stratification and molecular rapid diagnostic tools may further help tailoring antimicrobial therapies and avoid unnecessary broad-spectrum antibiotics. Continuous infusions of antibiotics, therapeutic drug monitoring (TDM)-based dosing regimens and combination therapy may contribute to optimizing antimicrobial therapy and limiting the emergence of resistance.

Predicting Extended-Spectrum Beta-Lactamase and Carbapenem Resistance in Enterobacteriaceae Bacteremia: A Diagnostic Model Systematic Review and Meta-Analysis

Enterobacteriaceae bacteremia, particularly when associated with antimicrobial resistance, can result in increased mortality, emphasizing the need for timely effective therapy. Clinical risk prediction models are promising tools, stratifying patients based on their risk of resistance due to ESBL and carbapenemase-producing Enterobacteriaceae in bloodstream infections (BSIs) and, thereby, improving therapeutic decisions. This systematic review and meta-analysis synthesized the literature on the performance of these models. Searches of PubMed and EMBASE led to the identification of 10 relevant studies with 6106 unique patient encounters. Nine studies concerned ESBL prediction, and one focused on the prediction of carbapenemases. For the two ESBL model derivation studies, the discrimination performance showed sensitivities of 53-85% and specificities of 93-95%. Among the four ESBL model derivation and validation studies, the sensitivities were 43-88%, and the specificities were 77-99%. The sensitivity and specificity for the subsequent external validation studies were 7-37% and 88-96%, respectively. For the three external validation studies, only two models were evaluated across multiple studies, with a pooled AUROC of 65-71%, with one study omitting the sensitivity/specificity. Only two studies measured clinical utility through hypothetical therapy assessments. Given the limited evidence on their interventional application, it would be beneficial to further assess these or future models, to better understand their clinical utility and ensure their safe and impactful implementation.

A Comparative Study of Cr(VI) Sorption by Aureobasidium pullulans AKW Biomass and Its Extracellular Melanin: Complementary Modeling with Equilibrium Isotherms, Kinetic Studies, and Decision Tree Modeling

Melanin as a natural polymer is found in all living organisms, and plays an important role in protecting the body from harmful UV rays from the sun. The efficiency of fungal biomass (Aureobasidium pullulans) and its extracellular melanin as Cr(VI) biosorbents was comparatively considered. The efficiency of Cr(VI) biosorption by the two sorbents used was augmented up to 240 min. The maximum sorption capacities were 485.747 (fungus biomass) and 595.974 (melanin) mg/g. The practical data were merely fitted to both Langmuir and Freundlich isotherms. The kinetics of the biosorption process obeyed the pseudo-first-order. Melanin was superior in Cr(VI) sorption than fungal biomass. Furthermore, four independent variables (contact time, initial concentration of Cr(VI), biosorbent dosage, and pH,) were modeled by the two decision trees (DTs). Conversely, to equilibrium isotherms and kinetic studies, DT of fungal biomass had lower errors compared to DT of melanin. Lately, the DTs improved the efficacy of the Cr(VI) removal process, thus introducing complementary and alternative solutions to equilibrium isotherms and kinetic studies. The Cr(VI) biosorption onto the biosorbents was confirmed and elucidated through FTIR, SEM, and EDX investigations. Conclusively, this is the first report study attaining the biosorption of Cr(VI) by biomass of A. pullulans and its extracellular melanin among equilibrium isotherms, kinetic study, and algorithmic decision tree modeling.

Empiric antimicrobial therapy in the intensive care unit based on the risk of multidrug-resistant bacterial infection: a single-centre case‒control study of blood culture results in Japan

BACKGROUND

Infections and sepsis are the leading causes of death in intensive care units (ICUs). Antimicrobial agent selection is challenging because the intervention is directly related to the outcome, and the problem of antimicrobial resistance (AMR) must be considered. Therefore, in this study, we aimed to clarify the epidemiological data and examine whether the detection rate of multidrug-resistant (MDR) bacteria differed depending on the presence or absence of the risk of MDR bacterial infections to establish guidance regarding the choice of antimicrobial therapy for ICU patients.

METHODS

This retrospective case‒control study was performed in a single ICU in Japan. Patients admitted to the ICU who underwent blood culture (BC) analysis were considered for inclusion in this study; patients were at risk of MDR bacterial infections, and controls were not. The primary outcome measure was the detection rate of MDR bacteria in BCs collected from patients and controls. The secondary outcome measure was the selection rate of anti-Pseudomonas and anti-methicillin-resistant Staphylococcus aureus (MRSA) drugs for patients and controls.

RESULTS

Among the 1,730 patients admitted to the ICU during the study period, BCs were obtained from 186 patients, and 173 samples were finally included in the analysis (n = 129 cases; n = 44 controls). No MDR bacteria or Pseudomonas aeruginosa were detected in the controls (14 (11%) vs. 0 (0%)) (P = 0.014) However, there was no difference in empiric antimicrobials, including anti-MRSA (30 (23%) vs. 12 (27%)) (P = 0.592) and anti-Pseudomonas aeruginosa (61 (47%) vs. 16 (36%)) (P = 0.208) drugs, that were administered to the two groups.

CONCLUSIONS

Even in critically ill patients in the ICU, MDR bacteria are unlikely to be detected in patients without the risk of MDR bacterial infections. Therefore, for such patients, a strategy of starting empiric narrow-spectrum antimicrobial therapy rather than empiric broad-spectrum therapy should be considered. This strategy, in conjunction with daily updates of clinical and epidemiological data at each facility, will promote the appropriate use of antimicrobials and reduce the emergence of MDR bacteria in the ICU.

TRIAL REGISTRATION

None.

Community carriage of ESBL-producing Escherichia coli and Klebsiella pneumoniae: a cross-sectional study of risk factors and comparative genomics of carriage and clinical isolates

The global prevalence of infections caused by extended-spectrum β-lactamase-producing Enterobacterales (ESBL-E) is increasing, and for Escherichia coli, observations indicate that this is partly driven by community-onset cases. The ESBL-E population structure in the community is scarcely described, and data on risk factors for carriage are conflicting. Here, we report the prevalence and population structure of fecal ESBL-producing E. coli and Klebsiella pneumoniae (ESBL-Ec/Kp) in a general adult population, examine risk factors, and compare carriage isolates with contemporary clinical isolates. Fecal samples obtained from 4,999 participants (54% women) ≥40 years in the seventh survey of the population-based Tromsø Study, Norway (2015, 2016), were screened for ESBL-Ec/Kp. In addition, we included 118 ESBL-Ec clinical isolates from the Norwegian surveillance program in 2014. All isolates were whole-genome sequenced. Risk factors associated with carriage were analyzed using multivariable logistic regression. ESBL-Ec gastrointestinal carriage prevalence was 3.3% [95% confidence interval (CI) 2.8%-3.9%, no sex difference] and 0.08% (0.02%-0.20%) for ESBL-Kp. For ESBL-Ec, travel to Asia was the only independent risk factor (adjusted odds ratio 3.46, 95% CI 2.18-5.49). E. coli ST131 was most prevalent in both collections. However, the ST131 proportion was significantly lower in carriage (24%) versus clinical isolates (58%, P < 0.001). Carriage isolates were genetically more diverse with a higher proportion of phylogroup A (26%) than clinical isolates (5%, P < 0.001), indicating that ESBL gene acquisition occurs in a variety of E. coli lineages colonizing the gut. STs commonly related to extraintestinal infections were more frequent in clinical isolates also carrying a higher prevalence of antimicrobial resistance, which could indicate clone-associated pathogenicity.IMPORTANCEESBL-Ec and ESBL-Kp are major pathogens in the global burden of antimicrobial resistance. However, there is a gap in knowledge concerning the bacterial population structure of human ESBL-Ec/Kp carriage isolates in the community. We have examined ESBL-Ec/Kp isolates from a population-based study and compared these to contemporary clinical isolates. The large genetic diversity of carriage isolates indicates frequent ESBL gene acquisition, while those causing invasive infections are more clone dependent and associated with a higher prevalence of antibiotic resistance. The knowledge of factors associated with ESBL carriage helps to identify patients at risk to combat the spread of resistant bacteria within the healthcare system. Particularly, previous travel to Asia stands out as a major risk factor for carriage and should be considered in selecting empirical antibiotic treatment in critically ill patients.

Bloodstream Infections: Comparison of Diagnostic Methods and Therapeutic Consequences between a Hospital in a Resource-Limited Setting and Two French Hospitals

In recent years, the diagnosis of bloodstream infections has been complemented by rapid microbiological methods, unattainable to most clinical laboratories in resource-limited settings. We evaluated the impact of their shortage on antibiotic therapy adequacy. We conducted a prospective multicenter cohort study including 150 adult Gram-negative bacilli bacteremia episodes, evenly distributed across three university hospitals: one in Lebanon, a resource-limited setting, and two in France, a resource-rich setting. Previous colonization by multidrug-resistant organisms (MDRO) was significantly more prevalent among the Lebanese than the French group of patients (16/50 vs. 5/100; p < 0.01). Bloodstream infections by carbapenemase-producing Enterobacterales and other MDRO were higher among the Lebanese than the French group of patients (25/50 vs. 12/100; p < 0.01). For the French group, rapid identification of species and mechanisms of resistance significantly shortened turnaround time for definitive laboratory diagnosis and increased antibiotic therapy adequacy. No statistically significant differences were noted in targeted antibiotic therapy between the two groups. This study suggests that, in settings where bacterial resistance is prevalent, rapid microbiological methods have not provided any additional value. The clinical and economic impact of rapid microbiological methods will likely depend on local CPE, VRE, and other MDRO epidemiology and are areas for future research.

Clinical Risk Scores to Predict Nonsusceptibility to Trimethoprim-Sulfamethoxazole, Fluoroquinolone, Nitrofurantoin, and Third-Generation Cephalosporin Among Adult Outpatient Episodes of Complicated Urinary Tract Infection

Background

Clinical risk scores were developed to estimate the risk of adult outpatients having a complicated urinary tract infection (cUTI) that was nonsusceptible to trimethoprim-sulfamethoxazole (TMP-SMX), fluoroquinolone, nitrofurantoin, or third-generation cephalosporin (3-GC) based on variables available on clinical presentation.

Methods

A retrospective cohort study (1 December 2017-31 December 2020) was performed among adult members of Kaiser Permanente Southern California with an outpatient cUTI. Separate risk scores were developed for TMP-SMX, fluoroquinolone, nitrofurantoin, and 3-GC. The models were translated into risk scores to quantify the likelihood of nonsusceptibility based on the presence of final model covariates in a given cUTI outpatient.

Results

A total of 30 450 cUTIs (26 326 patients) met the study criteria. Rates of nonsusceptibility to TMP-SMX, fluoroquinolone, nitrofurantoin, and 3-GC were 37%, 20%, 27%, and 24%, respectively. Receipt of prior antibiotics was the most important predictor across all models. The risk of nonsusceptibility in the TMP-SMX model exceeded 20% in the absence of any risk factors, suggesting that empiric use of TMP-SMX may not be advisable. For fluoroquinolone, nitrofurantoin, and 3-GC, clinical risk scores of 10, 7, and 11 predicted a ≥20% estimated probability of nonsusceptibility in the models that included cumulative number of prior antibiotics at model entry. This finding suggests that caution should be used when considering these agents empirically in patients who have several risk factors present in a given model at presentation.

Conclusions

We developed high-performing parsimonious risk scores to facilitate empiric treatment selection for adult outpatients with cUTIs in the critical period between infection presentation and availability of susceptibility results.

Risk factors for predicting extended-spectrum β-lactamase-producing Enterobacterales (ESBLE) infections in non-urinary isolates

Background

With increases in antimicrobial resistance, it is crucial that patients receive appropriate antimicrobial therapy in a timely manner. Advancements in rapid diagnostics offer the ability to identify resistant organisms quickly. However, this technology is not always accessible and relies on correct specimen collection. While awaiting new microbiology methods, it may be beneficial to identify risk factors associated with common types of resistance. Specifically, extended-spectrum β-lactamase-producing Enterobacterales (ESBLE) are a rising threat globally.

Objective

The primary objective of this retrospective case-control analysis was to identify factors associated with non-urinary ESBLE versus non-ESBLE infections.

Design/Methods

Patient cultures were randomly selected based on type of culture (blood, bacterial, or exudate) and organism (E. coli, K. pneumoniae, or K. oxytoca) to provide a 1:1 ratio of ESBLE to non-ESBLE infections. Baseline demographics and potential risk factors (malignancy, cirrhosis, acute kidney injury (AKI), and diabetes) were collected for each patient encounter.

Results

In the univariate analysis, risk factors that achieved a significant difference included cirrhosis, AKI, presence of urinary catheter, presence of center venous catheter, history of an ESBLE infection, hospital-acquired infection, and recent fluoroquinolone, cephalosporin, or beta-lactam use. The multivariate analysis showed that four factors were independently associated with an ESBLE infection: cirrhosis, urinary catheter, central venous catheter, and history of ESBLE. Having a history of an ESBLE had the highest adjusted odds ratio (aOR 12.49; 95% CI 4.71-33.15, P < .001) of the four factors.

Conclusions

These results demonstrate that there may be benefit in incorporating select risk factors into clinical decision support tools to identify patients at highest risk of ESBLE infection.

Safety of megadose meropenem in the empirical treatment of nosocomial sepsis: a pilot randomized clinical trial

Objective: To evaluate the safety of megadose meropenem as empirical treatment of nosocomial sepsis. Materials & methods: Critically ill patients diagnosed with sepsis received either high-dose (2 g every 8 h) or megadose (4 g every 8 h) meropenem as an intravenous infusion over 3 h. Results: A total of 23 patients with nosocomial sepsis were eligible and included in the megadose (n = 11) or high-dose (n = 12) group. No treatment-related adverse events were observed during a 14-day follow-up. Clinical response was also comparable between the groups. Conclusion: Megadose meropenem may be considered for empirical treatment of nosocomial sepsis without serious concern regarding its safety.

A Clinical Decision Tree to Support Keratoconus Patients Considering Corneal Cross-Linking Combined with Refractive Treatment

BACKGROUND

To develop a fast and frugal decision tree to identify keratoconus patients most likely to benefit visually from the combination of corneal cross-linking (CXL) with topography-guided photorefractive keratectomy ("CXL plus").

PATIENTS AND METHODS

The outcome of interest was an improvement in uncorrected distance visual acuity (UDVA) by at least two lines at the 12-month follow-up. Preoperative and 12-month follow-up data from patients who received CXL plus (n = 96) and CXL only (n = 96) were used in a recursive partitioning approach to construct a frugal tree with three variables (corneal thickness [>/< 430 um], patient interest in CXL plus [yes/no], and tomographic cylinder [</> 3 D]). In addition, we estimated the probability of the outcome from a multivariate logistic regression model for each combination of variables used in the decision tree.

RESULTS

In the complete sample, 101/192 (52.6%) patients improved by at least two lines at the 12-month follow-up. Patients affirmative in all three answers had a 75.6% (34/45) probability of gaining at least two lines of improvement in UDVA by CXL plus. The statistical model estimated a 66.0% probability for a successful outcome.

CONCLUSION

A fast and frugal tree consisting of three variables can be used to select a patient group with a high likelihood to benefit from CXL plus. The tree is useful in the preoperative counseling of keratoconus patients contemplating the CXL plus option, an intervention that is not fully covered by many health insurances.

Exclusive Biosynthesis of Pullulan Using Taguchi’s Approach and Decision Tree Learning Algorithm by a Novel Endophytic Aureobasidium pullulans Strain

Pullulan is a biodegradable, renewable, and environmentally friendly hydrogel biopolymer, with potential uses in food, medicine, and cosmetics. New endophytic Aureobasidium pullulans (accession number; OP924554) was used for the biosynthesis of pullulan. Innovatively, the fermentation process was optimized using both Taguchi’s approach and the decision tree learning algorithm for the determination of important variables for pullulan biosynthesis. The relative importance of the seven tested variables that were obtained by Taguchi and the decision tree model was accurate and followed each other’s, confirming the accuracy of the experimental design. The decision tree model was more economical by reducing the quantity of medium sucrose content by 33% without a negative reduction in the biosynthesis of pullulan. The optimum nutritional conditions (g/L) were sucrose (60 or 40), K2HPO4 (6.0), NaCl (1.5), MgSO4 (0.3), and yeast extract (1.0) at pH 5.5, and short incubation time (48 h), yielding 7.23% pullulan. The spectroscopic characterization (FT-IR and 1H-NMR spectroscopy) confirmed the structure of the obtained pullulan. This is the first report on using Taguchi and the decision tree for pullulan production by a new endophyte. Further research is encouraged for additional studies on using artificial intelligence to maximize fermentation conditions.

Using Machine Learning to Predict Antimicrobial Resistance-A Literature Review

Machine learning (ML) algorithms are increasingly applied in medical research and in healthcare, gradually improving clinical practice. Among various applications of these novel methods, their usage in the combat against antimicrobial resistance (AMR) is one of the most crucial areas of interest, as increasing resistance to antibiotics and management of difficult-to-treat multidrug-resistant infections are significant challenges for most countries worldwide, with life-threatening consequences. As antibiotic efficacy and treatment options decrease, the need for implementation of multimodal antibiotic stewardship programs is of utmost importance in order to restrict antibiotic misuse and prevent further aggravation of the AMR problem. Both supervised and unsupervised machine learning tools have been successfully used to predict early antibiotic resistance, and thus support clinicians in selecting appropriate therapy. In this paper, we reviewed the existing literature on machine learning and artificial intelligence (AI) in general in conjunction with antimicrobial resistance prediction. This is a narrative review, where we discuss the applications of ML methods in the field of AMR and their value as a complementary tool in the antibiotic stewardship practice, mainly from the clinician's point of view.

Epidemiology and outcomes of hospital-acquired bloodstream infections in intensive care unit patients: the EUROBACT-2 international cohort study

PURPOSE

In the critically ill, hospital-acquired bloodstream infections (HA-BSI) are associated with significant mortality. Granular data are required for optimizing management, and developing guidelines and clinical trials.

METHODS

We carried out a prospective international cohort study of adult patients (≥ 18 years of age) with HA-BSI treated in intensive care units (ICUs) between June 2019 and February 2021.

RESULTS

2600 patients from 333 ICUs in 52 countries were included. 78% HA-BSI were ICU-acquired. Median Sequential Organ Failure Assessment (SOFA) score was 8 [IQR 5; 11] at HA-BSI diagnosis. Most frequent sources of infection included pneumonia (26.7%) and intravascular catheters (26.4%). Most frequent pathogens were Gram-negative bacteria (59.0%), predominantly Klebsiella spp. (27.9%), Acinetobacter spp. (20.3%), Escherichia coli (15.8%), and Pseudomonas spp. (14.3%). Carbapenem resistance was present in 37.8%, 84.6%, 7.4%, and 33.2%, respectively. Difficult-to-treat resistance (DTR) was present in 23.5% and pan-drug resistance in 1.5%. Antimicrobial therapy was deemed adequate within 24 h for 51.5%. Antimicrobial resistance was associated with longer delays to adequate antimicrobial therapy. Source control was needed in 52.5% but not achieved in 18.2%. Mortality was 37.1%, and only 16.1% had been discharged alive from hospital by day-28.

CONCLUSIONS

HA-BSI was frequently caused by Gram-negative, carbapenem-resistant and DTR pathogens. Antimicrobial resistance led to delays in adequate antimicrobial therapy. Mortality was high, and at day-28 only a minority of the patients were discharged alive from the hospital. Prevention of antimicrobial resistance and focusing on adequate antimicrobial therapy and source control are important to optimize patient management and outcomes.

Know your Microbe Foes: The Role of Surveillance in Combatting Antimicrobial Resistance

Antibiotic-resistant organisms (AROs) are difficult and costly to treat, associated with high mortality rates, and are on the rise. In the United States, there is limited tracking of AROs, which can contribute to transmission and inhibit infection prevention interventions. Surveillance is limited by a lack of standardized methods for colonization screening and limited communication regarding patient ARO-status between healthcare settings. Some regional surveillance and reporting efforts are in place for extensively-resistant AROs such as carbapenem-resistant Enterobacterales (CRE), but need to be further expanded nationwide and to include other AROs such as extended-spectrum β-lactamase (ESBL) producing organisms. Increased surveillance of ARO infections and colonization will inform future targeted intervention and infection prevention strategies.

Machine learning in predicting antimicrobial resistance: a systematic review and meta-analysis

INTRODUCTION

Antimicrobial resistance (AMR) is a global health threat; rapid and timely identification of AMR improves patient prognosis and reduces inappropriate antibiotic use.

METHODS

Relevant literature in PubMed, Web of Science, Embase and Institute of Electrical and Electronics Engineers prior to 28 September 2021 was searched. Any study that deployed machine learning (ML) or a risk score as a tool to predict AMR was included in the final review; there were 25 studies that employed the ML algorithm to predict AMR.

RESULTS

Extended spectrum β-lactamases, methicillin-resistant Staphylococcus aureus (MRSA) and carbapenem resistance were the most common outcomes in studies with a specific resistance pattern. The most common algorithms in ML prediction were logistic regression (n = 14 studies), decision tree (n = 14) and random forest (n = 7). The area under the curve (AUC) range for ML prediction was 0.48-0.93. The pooled AUC for ML prediction was 0.82 (0.78-0.85). Compared with risk score, higher specificity [87% (82-91) vs. 37% (25-51)] was indicated for ML prediction, but not sensitivity [67% (62-72) vs. 73% (67-79)].

CONCLUSIONS

Machine learning might be a potential technology for AMR prediction; however, retrospective methodology for model development, nonstandard data processing and scarcity of validation in a randomised controlled trial or real-world study limit the application of these models in clinical practice.

Risk Factors of Infections Due to Multidrug-Resistant Gram-Negative Bacteria in a Community Hospital in Rural Thailand

Antimicrobial resistance is a major public health concern globally. The most serious antimicrobial resistance problem among pathogenic bacteria is multidrug resistance (MDR). The objectives of this study were to investigate the risk factors of MDR infections and to develop a risk assessment tool for MDR Gram-negative bacteria (MDR-GNB) infections at a community hospital in rural Thailand. The study revealed 30.77% MDR-GNB among GNB strains. The most common MDR-GNB strains were 63.02% for Escherichia coli and 11.46% for Klebsiella pneumoniae. A case-control study was applied to collect clinical data between January 2016 and December 2020. Univariate logistic regression and multivariate logistic regression were used to analyze the risk factors for MDR-GNB and a risk assessment score for each factor was determined based on its regression coefficient. The risk factors for MDR-GNB infections were as follows: the presence of Enterobacteriaceae that produce extended-spectrum beta-lactamase (ESBL) (OR_Adj. 23.53, 95% CI 7.00-79.09), infections occurring within the urinary tract (OR_Adj. 2.25, 95% CI 1.44-3.53), and patients with a history of steroid usage (OR_Adj. 1.91, 95% CI 1.15-3.19). Based on the assigned risk scores for each associated factor, the newly developed risk assessment tool for MDR-GNB infections achieved 64.54% prediction accuracy (AUC-ROC 0.65, 95% CI 0.61-0.68), demonstrating that the tool could be used to assess bacterial infection cases in community hospitals. Its use should provide practical guidance on MDR evaluation and prevention. This study was part of an antibiotic stewardship program; the study surveyed antibiotic-resistant situations in a hospital and implemented an effective risk assessment tool using key risk factors of MDR-GNB infections.

Prediction of extubation failure in the paediatric cardiac ICU using machine learning and high-frequency physiologic data

BACKGROUND

Cardiac intensivists frequently assess patient readiness to wean off mechanical ventilation with an extubation readiness trial despite it being no more effective than clinician judgement alone. We evaluated the utility of high-frequency physiologic data and machine learning for improving the prediction of extubation failure in children with cardiovascular disease.

METHODS

This was a retrospective analysis of clinical registry data and streamed physiologic extubation readiness trial data from one paediatric cardiac ICU (12/2016-3/2018). We analysed patients' final extubation readiness trial. Machine learning methods (classification and regression tree, Boosting, Random Forest) were performed using clinical/demographic data, physiologic data, and both datasets. Extubation failure was defined as reintubation within 48 hrs. Classifier performance was assessed on prediction accuracy and area under the receiver operating characteristic curve.

RESULTS

Of 178 episodes, 11.2% (N = 20) failed extubation. Using clinical/demographic data, our machine learning methods identified variables such as age, weight, height, and ventilation duration as being important in predicting extubation failure. Best classifier performance with this data was Boosting (prediction accuracy: 0.88; area under the receiver operating characteristic curve: 0.74). Using physiologic data, our machine learning methods found oxygen saturation extremes and descriptors of dynamic compliance, central venous pressure, and heart/respiratory rate to be of importance. The best classifier in this setting was Random Forest (prediction accuracy: 0.89; area under the receiver operating characteristic curve: 0.75). Combining both datasets produced classifiers highlighting the importance of physiologic variables in determining extubation failure, though predictive performance was not improved.

CONCLUSION

Physiologic variables not routinely scrutinised during extubation readiness trials were identified as potential extubation failure predictors. Larger analyses are necessary to investigate whether these markers can improve clinical decision-making.

Short- and long-term mortality in patients with urosepsis caused by Escherichia coli susceptible and resistant to 3rd generation cephalosporins

Background

The aim of this study was to compare short- and long-term mortality among patients with urosepsis caused by Escherichia coli susceptibile (EC-SC) and resistant (EC-RC) to 3rd generation cephalosporins.

Methods

A retrospective cohort study that included all patients with E. coli urosepsis admitted to a 700-bed hospital from January 2014 until December 2019. Mortality up to 30 days, 6 months and 1 year was assessed using logistic multivariate regression analysis and Cox regression analysis.

Results

A total of 313 adult were included, 195 with EC-SC and 118 patients with EC-RC. 205 were females (74%), mean age was 79 (SD 12) years. Mean Charlson score was 4.93 (SD 2.18) in the EC-SC group and 5.74 (SD 1.92) in the EC-RC group. Appropriate empiric antibiotic therapy was initiated in 245 (78.3%) patients, 100% in the EC-SC group but only 42.5% in the EC-RC group. 30-day mortality occurred in 12 (6.3%) of EC-SC group and 15 (12.7%) in the EC-RC group. Factors independently associated with 30-day mortality were Charlson score, Pitt bacteremia score, fever upon admission and infection with a EC-RC. Appropriate antibiotic therapy was not independently associated with 30-day mortality. Differences in mortality between groups remained significant one year after the infection and were significantly associated with the Charlson co-morbidity score.

Conclusions

Mortality in patients with urosepsis due to E. coli is highly affected by age and comorbidities. Although mortality was higher in the EC-RC group, we could not demonstrate an association with inappropriate empirical antibiotic treatment. Mortality remained higher at 6 months and 1 year long after the infection resolved but was associated mainly with co-morbidity.

A Data-Driven Framework for Identifying Intensive Care Unit Admissions Colonized With Multidrug-Resistant Organisms

Background

The rising prevalence of multi-drug resistant organisms (MDROs), such as Methicillin-resistant Staphylococcus aureus (MRSA), Vancomycin-resistant Enterococci (VRE), and Carbapenem-resistant Enterobacteriaceae (CRE), is an increasing concern in healthcare settings.

Materials and Methods

Leveraging data from electronic healthcare records and a unique MDRO universal screening program, we developed a data-driven modeling framework to predict MRSA, VRE, and CRE colonization upon intensive care unit (ICU) admission, and identified the associated socio-demographic and clinical factors using logistic regression (LR), random forest (RF), and XGBoost algorithms. We performed threshold optimization for converting predicted probabilities into binary predictions and identified the cut-off maximizing the sum of sensitivity and specificity.

Results

Four thousand six hundred seventy ICU admissions (3,958 patients) were examined. MDRO colonization rate was 17.59% (13.03% VRE, 1.45% CRE, and 7.47% MRSA). Our study achieved the following sensitivity and specificity values with the best performing models, respectively: 80% and 66% for VRE with LR, 73% and 77% for CRE with XGBoost, 76% and 59% for MRSA with RF, and 82% and 83% for MDRO (i.e., VRE or CRE or MRSA) with RF. Further, we identified several predictors of MDRO colonization, including long-term care facility stay, current diagnosis of skin/subcutaneous tissue or infectious/parasitic disease, and recent isolation precaution procedures before ICU admission.

Conclusion

Our data-driven modeling framework can be used as a clinical decision support tool for timely predictions, characterization and identification of high-risk patients, and selective and timely use of infection control measures in ICUs.

Pan-genome and resistome analysis of extended-spectrum ß-lactamase-producing Escherichia coli: A multi-setting epidemiological surveillance study from Malaysia

Objectives

This study profiled the prevalence of extended-spectrum ß-lactamase-producing Escherichia coli (ESBL-EC) in the community and compared their resistome and genomic profiles with isolates from clinical patients through whole-genome sequencing.

Methods

Fecal samples from 233 community dwellers from Segamat, a town in southern Malaysia, were obtained between May through August 2018. Putative ESBL strains were screened and tested using antibiotic susceptibility tests. Additionally, eight clinical ESBL-EC were obtained from a hospital in the same district between June through October 2020. Whole-genome sequencing was then conducted on selected ESBL-EC from both settings (n = 40) for pan-genome comparison, cluster analysis, and resistome profiling.

Results

A mean ESBL-EC carriage rate of 17.82% (95% CI: 10.48%– 24.11%) was observed in the community and was consistent across demographic factors. Whole-genome sequences of the ESBL-EC (n = 40) enabled the detection of multiple plasmid replicon groups (n = 28), resistance genes (n = 34) and virulence factors (n = 335), with no significant difference in the number of genes carried between the community and clinical isolates (plasmid replicon groups, p = 0.13; resistance genes, p = 0.47; virulence factors, p = 0.94). Virulence gene marker analysis detected the presence of extraintestinal pathogenic E. coli (ExPEC), uropathogenic E. coli (UPEC), and enteroaggregative E. coli (EAEC) in both the community and clinical isolates. Multiple blaCTX-M variants were observed, dominated by blaCTX-M-27 (n = 12), blaCTX-M-65 (n = 10), and blaCTX-M-15 (n = 9). The clinical and community isolates did not cluster together based on the pan-genome comparison, suggesting isolates from the two settings were clonally unrelated. However, cluster analysis based on carried plasmids, resistance genes and phenotypic susceptibility profiles identified four distinct clusters, with similar patterns between the community and clinical isolates.

Conclusion

ESBL-EC from the clinical and community settings shared similar resistome profiles, suggesting the frequent exchange of genetic materials through horizontal gene transfer.

Antibiotic Decision-Making in the ICU

It is well established that Intensive Care Units (ICUs) are a focal point in antimicrobial consumption with a major influence on the ecological consequences of antibiotic use. With the high prevalence and mortality of infections in critically ill patients, and the clinical challenges of treating patients with septic shock, the impact of real life clinical decisions made by intensivists becomes more significant. Both under- and over-treatment with unnecessarily broad spectrum antibiotics can lead to detrimental outcomes. Even though substantial progress has been made in developing rapid diagnostic tests that can help guide antibiotic use, there is still a time window when clinicians must decide the empiric antibiotic treatment with insufficient clinical data. The continuous streams of data available in the ICU environment make antimicrobial optimization an ongoing challenge for clinicians but at the same time can serve as the input for sophisticated models. In this review, we summarize the evidence to help guide antibiotic decision-making in the ICU. We focus on 1) deciding IF: to start antibiotics, 2) choosing the spectrum of the empiric agents to use, and 3) de-escalating the chosen empiric antibiotics. We provide a perspective on the role of machine learning and artificial intelligence models for clinical decision support systems that can be incorporated seamlessly into clinical practice in order to improve the antibiotic selection process and, more importantly, current and future patients' outcomes.

Comparison of the performance of a clinical classification tree versus clinical gestalt in predicting sepsis with extended-spectrum beta-lactamase–producing gram-negative rods

A clinical decision tree was developed using point-of-care characteristics to identify patients with culture-proven sepsis due to extended-spectrum β-lactamase–producing Enterobacterales (ESBL-PE). We compared its performance with the clinical gestalt of emergency department (ED) clinicians and hospital-based clinicians. The developed tree outperformed ED-based clinicians but was comparable to inpatient-based clinicians.

Impact of extended-spectrum beta-lactamase-producing Enterobacterales (ESBL-E) rectal carriage in cancer patients admitted to the intensive care unit

Little data is available on extended-spectrum beta-lactamase-producing Enterobacterales (ESBL-E) rectal colonization in cancer patients admitted to the intensive care unit (ICU). We aimed to describe the epidemiology of ESBL-E in cancer patients hospitalized in the ICU compared with non-cancer patients. ESBL-E colonization was detected in 6.6% of 1,013 cancer patients and 6.4% of 1,625 non-cancer patients. At admission, among the 172 colonized patients: 48/67 cancer patients and 78/105 non-cancer patients developed an infection, documented with an ESBL-E for 21% and 24% of them, respectively. The in-hospital mortality rate among colonized patients was 33% in cancer patients and 12% in non-cancer patients. In cancer patients, ESBL-E infections are rare but systematic rectal screening identifies high-risk population and guides empirical antibiotic therapy. It also contributes to being aware of the ICU microbiological ecology.

Ceftriaxone versus antistaphylococcal antibiotics for definitive treatment of methicillin-susceptible Staphylococcus aureus infections: a systematic review and meta-analysis

Optimal therapy for methicillin-susceptible Staphylococcus aureus (MSSA) infections is unclear. Current standard of care consists of antistaphylococcal antibiotics (ASAs) such as nafcillin, oxacillin and cefazolin. Ceftriaxone has been evaluated due to its advantage as a once-daily outpatient regimen. However, questions remain regarding its efficacy compared with ASAs. We aimed to conduct a review and synthesis of available literature for outcomes of patients treated with ceftriaxone or ASAs for MSSA infections. We searched Cochrane Central Register of Controlled Trials, Embase Ovid, MEDLINE Ovid, Scopus and Web of Science (1990 to June 2021). Risk of bias for cohort studies was assessed by the Newcastle-Ottawa scale. We pooled risk ratios (RRs) using the DerSimonian-Laird random-effects model for outcomes of those receiving ceftriaxone versus ASAs. Heterogeneity was assessed by the I² index. From 459 identified studies, 7 were included in the quantitative synthesis totalling 1640 patients. Definitive therapy with ceftriaxone was associated with a lower risk of toxicity requiring therapy alteration (RR 0.49, 95% CI 0.27-0.88; I² = 0%). There was no difference in terms of 90-day all-cause mortality (RR 0.93, 95% CI 0.46-1.88; I² = 9%), hospital readmission (RR 0.96, 95% CI 0.57-1.64; I² = 0%) or infection recurrence (RR 1.04, 95% CI 0.63-1.72; I² =0%). Current evidence suggests there is no difference in efficacy between ceftriaxone and ASAs for MSSA infection, with a lower risk of toxicity with ceftriaxone. Within the limitations of available retrospective studies, ceftriaxone is a consideration for definitive therapy of MSSA infection. [Trial registration: PROSPERO ID: CRD42021259086].

Artificial intelligence and machine learning assisted drug delivery for effective treatment of infectious diseases

In the era of antimicrobial resistance, the prevalence of multidrug-resistant microorganisms that resist conventional antibiotic treatment has steadily increased. Thus, it is now unquestionable that infectious diseases are significant global burdens that urgently require innovative treatment strategies. Emerging studies have demonstrated that artificial intelligence (AI) can transform drug delivery to promote effective treatment of infectious diseases. In this review, we propose to evaluate the significance, essential principles, and popular tools of AI in drug delivery for infectious disease treatment. Specifically, we will focus on the achievements and key findings of current research, as well as the applications of AI on drug delivery throughout the whole antimicrobial treatment process, with an emphasis on drug development, treatment regimen optimization, drug delivery system and administration route design, and drug delivery outcome prediction. To that end, the challenges of AI in drug delivery for infectious disease treatments and their current solutions and future perspective will be presented and discussed.

A decision tree prediction model for a short-term outcome of delirium in patients with advanced cancer receiving pharmacological interventions: A secondary analysis of a multicenter and prospective observational study (Phase-R)

Objective

There is no widely used prognostic model for delirium in patients with advanced cancer. The present study aimed to develop a decision tree prediction model for a short-term outcome.

Method

This is a secondary analysis of a multicenter and prospective observational study conducted at 9 psycho-oncology consultation services and 14 inpatient palliative care units in Japan. We used records of patients with advanced cancer receiving pharmacological interventions with a baseline Delirium Rating Scale Revised-98 (DRS-R98) severity score of ≥10. A DRS-R98 severity score of <10 on day 3 was defined as the study outcome. The dataset was randomly split into the training and test dataset. A decision tree model was developed using the training dataset and potential predictors. The area under the curve (AUC) of the receiver operating characteristic curve was measured both in 5-fold cross-validation and in the independent test dataset. Finally, the model was visualized using the whole dataset.

Results

Altogether, 668 records were included, of which 141 had a DRS-R98 severity score of <10 on day 3. The model achieved an average AUC of 0.698 in 5-fold cross-validation and 0.718 (95% confidence interval, 0.627–0.810) in the test dataset. The baseline DRS-R98 severity score (cutoff of 15), hypoxia, and dehydration were the important predictors, in this order.

Significance of results

We developed an easy-to-use prediction model for the short-term outcome of delirium in patients with advanced cancer receiving pharmacological interventions. The baseline severity of delirium and precipitating factors of delirium were important for prediction.

Splitting Choice and Computational Complexity Analysis of Decision Trees

Some theories are explored in this research about decision trees which give theoretical support to the applications based on decision trees. The first is that there are many splitting criteria to choose in the tree growing process. The splitting bias that influences the criterion chosen due to missing values and variables with many possible values has been studied. Results show that the Gini index is superior to entropy information as it has less bias regarding influences. The second is that noise variables with more missing values have a better chance to be chosen while informative variables do not. The third is that when there are many noise variables involved in the tree building process, it influences the corresponding computational complexity. Results show that the computational complexity increase is linear to the number of noise variables. So methods that decompose more information from the original data but increase the variable dimension can also be considered in real applications.

The Japanese Clinical Practice Guidelines for Management of Sepsis and Septic Shock 2020 (J-SSCG 2020)

The Japanese Clinical Practice Guidelines for Management of Sepsis and Septic Shock 2020 (J-SSCG 2020), a Japanese-specific set of clinical practice guidelines for sepsis and septic shock created as revised from J-SSCG 2016 jointly by the Japanese Society of Intensive Care Medicine and the Japanese Association for Acute Medicine, was first released in September 2020 and published in February 2021. An English-language version of these guidelines was created based on the contents of the original Japanese-language version. The purpose of this guideline is to assist medical staff in making appropriate decisions to improve the prognosis of patients undergoing treatment for sepsis and septic shock. We aimed to provide high-quality guidelines that are easy to use and understand for specialists, general clinicians, and multidisciplinary medical professionals. J-SSCG 2016 took up new subjects that were not present in SSCG 2016 (e.g., ICU-acquired weakness [ICU-AW], post-intensive care syndrome [PICS], and body temperature management). The J-SSCG 2020 covered a total of 22 areas with four additional new areas (patient- and family-centered care, sepsis treatment system, neuro-intensive treatment, and stress ulcers). A total of 118 important clinical issues (clinical questions, CQs) were extracted regardless of the presence or absence of evidence. These CQs also include those that have been given particular focus within Japan. This is a large-scale guideline covering multiple fields; thus, in addition to the 25 committee members, we had the participation and support of a total of 226 members who are professionals (physicians, nurses, physiotherapists, clinical engineers, and pharmacists) and medical workers with a history of sepsis or critical illness. The GRADE method was adopted for making recommendations, and the modified Delphi method was used to determine recommendations by voting from all committee members.As a result, 79 GRADE-based recommendations, 5 Good Practice Statements (GPS), 18 expert consensuses, 27 answers to background questions (BQs), and summaries of definitions and diagnosis of sepsis were created as responses to 118 CQs. We also incorporated visual information for each CQ according to the time course of treatment, and we will also distribute this as an app. The J-SSCG 2020 is expected to be widely used as a useful bedside guideline in the field of sepsis treatment both in Japan and overseas involving multiple disciplines.

Clinical Decision Support Trees Can Help Optimize Utilization of Anaplasma phagocytophilum Nucleic Acid Amplification Testing

Anaplasmosis, a tick-borne illness caused by Anaplasma phagocytophilum (AP), presents with nonspecific clinical symptoms, including fever and headache, and is often accompanied by laboratory abnormalities of leukopenia, thrombocytopenia, and mildly elevated liver function tests (LFTs). Laboratory confirmation of acute infection occurs with nucleic acid amplification testing (NAAT). This retrospective cohort study aimed to develop a clinical decision support algorithm to aid in decision-making about test ordering. A data set was constructed with AP NAAT results and time-adjacent complete blood count and LFT results for adult patients tested for AP in a 12.5-year period. A second, smaller data set matched each patient with a positive AP NAAT to two patients with negative tests. Chart review for clinical symptoms was performed on this smaller data set. A decision tree algorithm was deployed to identify patient clusters with negative AP NAAT results. A total of 137/1,204 (11%) patients tested positive by NAAT for AP. In the larger, laboratory-only data set (n = 1,204), patients with a platelet count of >177 × 10³/μl and age of <48 years had a negative AP NAAT (204/1,204, 17%, P < 0.05). In the smaller, cohorted data set with chart review (n = 402), patients with a platelet count of >188 × 10³/μl and no fever or chills also did not have positive AP NAAT (58/402, 14%, P < 0.05). We generated two decision trees that can help determine the utility of AP NAAT using readily available clinical and laboratory data. These have the potential to significantly reduce unnecessary AP testing.

LightGBM: an efficient and accurate method for predicting pregnancy diseases

As machine learning is becoming the fashion in disease prediction while no prediction model has performed very efficiently and accurately on predicting pregnancy diseases up to now, it's necessary to compare several common machine learning methods' performance on pregnancy diseases prediction and select out the best one. The data of two common pregnancy complications, pregnancy-induced hypertension (PIH) and Intrahepatic cholestasis of pregnancy (ICP), based on various maternal characteristics measured in patients' routine blood examination in 10-19 weeks of gestation are considered to be suitable to be learned. This is a retrospective study of 320 healthy pregnancies in 10-19 weeks, with 149 patients who subsequently developed PIH and 250 patients who subsequently developed ICP. Nine machine learning methods were used to predict PIH and ICP and their performance was compared via 8 evaluation indexes. Finally, the light Gradient Boosting Machine (lightGBM) is considered to be the best method to predict gestational diseases.Impact statementWhat is already known on this subject? As a kind of commonly used method in disease prediction, machine learning could be applied to clinical data for developing robust risk models and many achievements have been made. Also, machine learning can be used to predict pregnancy diseases. Although some machine learning methods have been used for screening gestational diseases, methods based on simple theories, such as logistic regression and decision tree, are frequently used. They don't always have a very satisfactory prediction results. Besides, only a few types of pregnancy diseases can be predicted.What do the results of this study add? LightGBM has the best prediction results of PIH and ICP among 9 machine learning methods in this study. It can predict PIH (AUC = 81.72%) with a sensitivity of 70.59%, and ICP (AUC = 95.91%) with a sensitivity of 97.91%.What are the implications of these findings for clinical practice and/or further research? A new model has been developed for effective first-trimester screening for two common pregnancy diseases, PIH and ICP. This lightGBM model can be used in relative hospitals and population of the research, and provide references for doctors' diagnosis and treatment of pregnant women. In further research, the predicted effect of lightGBM on daily practice and other pregnancy diseases such as pregnancy diabetes, will be verified.

Improving Decision Making in Empiric Antibiotic Selection (IDEAS) for Gram-negative Bacteremia: A Prospective Clinical Implementation Study

BACKGROUND

Timely selection of adequate empiric antibiotics has become increasingly difficult due to rising resistance rates and the competing desire to apply antimicrobial stewardship (AMS) principles. Individualized clinical prediction models offer the promise of reducing broad-spectrum antibiotic use and preserving/improving adequacy of treatment, but few have been validated in the clinical setting.

METHODS

Multivariable models were used to predict the probability of susceptibility for gram-negative (GN) bacteria in bloodstream infections (bacteremia) to ceftriaxone, ciprofloxacin, ceftazidime, piperacillin-tazobactam, and meropenem. The models were combined with existing resistance-prediction methods to generate optimized and individualized suggestions for empiric therapy that were provided to prescribers by an AMS pharmacist. De-escalation of empiric antibiotics and adequacy of therapy were analyzed using a quasi-experimental design comparing two 9-month periods (pre- and postintervention) at a large academic tertiary care institution.

RESULTS

Episodes of bacteremia (n = 182) were identified in the preintervention and postintervention (n = 201) periods. Patients who received the intervention were more likely to have their therapy de-escalated (29 vs 21%; aOR = 1.77; 95% CI, 1.09-2.87; P = .02). The intervention also increased the proportion of patients who were on the narrowest adequate therapy at the time of culture finalization (44% in the control and 55% in the intervention group; aOR = 2.04; 95% CI, 1.27-3.27; P = .003). Time to adequate therapy was similar in the intervention and control groups (5 vs 4 hours; P = .95).

CONCLUSIONS

An AMS intervention, based on individualized predictive models for resistance, can influence empiric antibiotic selections for GN bacteremia to facilitate early de-escalation of therapy without compromising adequacy of antibiotic coverage.