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.

Personalising intravenous to oral antibiotic switch decision making through fair interpretable machine learning.

Antimicrobial resistance (AMR) and healthcare associated infections pose a significant threat globally. One key prevention strategy is to follow antimicrobial stewardship practices, in particular, to maximise targeted oral therapy and reduce the use of indwelling vascular devices for intravenous (IV) administration. Appreciating when an individual patient can switch from IV to oral antibiotic treatment is often non-trivial and not standardised. To tackle this problem we created a machine learning model to predict when a patient could switch based on routinely collected clinical parameters. 10,362 unique intensive care unit stays were extracted and two informative feature sets identified. Our best model achieved a mean AUROC of 0.80 (SD 0.01) on the hold-out set while not being biased to individuals protected characteristics. Interpretability methodologies were employed to create clinically useful visual explanations. In summary, our model provides individualised, fair, and interpretable predictions for when a patient could switch from IV-to-oral antibiotic treatment. Prospectively evaluation of safety and efficacy is needed before such technology can be applied clinically.

A low-volume LC/MS method for highly sensitive monitoring of phenoxymethylpenicillin, benzylpenicillin, and probenecid in human serum.

Background: The optimization of antimicrobial dosing plays a crucial role in improving the likelihood of achieving therapeutic success while reducing the risks associated with toxicity and antimicrobial resistance. Probenecid has shown significant potential in enhancing the serum exposure of phenoxymethylpenicillin, thereby allowing for lower doses of phenoxymethylpenicillin to achieve similar pharmacokinetic/pharmacodynamic (PK/PD) targets. We developed a triple quadrupole liquid chromatography mass spectrometry (TQ LC/MS) analysis of, phenoxymethylpenicillin, benzylpenicillin and probenecid using benzylpenicillin-d7 and probenecid-d14 as IS in single low-volumes of human serum, with improved limit of quantification to support therapeutic drug monitoring. Methods: Sample clean-up was performed by protein precipitation using acetonitrile. Reverse phase chromatography was performed using TQ LC/MS. The mobile phase consisted of 55% methanol in water + 0.1% formic acid, with a flow rate of 0.4 mL min-1. Antibiotic stability was assessed at different temperatures. Results: Chromatographic separation was achieved within 2 minutes, allowing simultaneous measurement of phenoxymethylpenicillin, benzylpenicillin and probenecid in a single 15 µL blood sample. Validation indicated linearity over the range 0.0015-10 mg L-1, with accuracy of 96-102% and a LLOQ of 0.01 mg L-1. All drugs demonstrated good stability under different storage conditions. Conclusion: The developed method is simple, rapid, accurate and clinically applicable for the quantification of phenoxymethylpenicillin, benzylpenicillin and probenecid in tandem.

Coronavirus disease 2019 (COVID-19) impact on central-line-associated bloodstream infections (CLABSI): a systematic review.

Introduction: Central line-associated bloodstream infections (CLABSI) are an important clinical and public health issue, impacted by the purported increase in healthcare-associated infections (including CLABSI) during the COVID-19 pandemic. This review evaluates the impact of COVID-19 on CLABSI at a global level, to determine risk factors, effective preventive measures and microbiological epidemiology. Methods: A systematic literature review was performed using a PECO framework, with COVID-19 infection as the exposure measure and CLABSI rates as the main outcome of interest, pre- and during the pandemic. Results: Overall, most studies (17 of N=21) found a significant increase in CLABSI incidence/rates during the pandemic. Four studies showed a reduction (N=1) or no increase (N=3). High workload, redeployment, and 'overwhelmed' healthcare staff were recurrent risk-factor themes, likely to have negatively influenced basic infection control practices, including compliance with hand hygiene and line care bundles. Microbiological epidemiology was also impacted, with an increase in enterococcal infections and other pathogens. Conclusion: The COVID-19 pandemic significantly impacted CLABSI incidence/rates. Observations from the different studies highlight significant gaps in healthcare associated infections (HCAI) knowledge and practice during the pandemic, and the importance of identifying preventive measures effective in reducing CLABSI, essential to health system resilience for future pandemics. Central to this are changes to CLABSI surveillance, as reporting is not mandatory in many healthcare systems. An audit tool combined with regular assessments of the compliance with infection control measures and line care bundles also remains an essential step in the prevention of CLABSI.

Rapid Diagnostic Test Value and Implementation in Antimicrobial Stewardship Across Low-to-Middle and High-Income Countries: A Mixed-Methods Review.

Despite technological advancements in infectious disease rapid diagnostic tests (RDTs) and use to direct therapy at the per-patient level, RDT utilisation in antimicrobial stewardship programmes (ASPs) is variable across low-to-middle income and high-income countries. Key insights from a panel of seven infectious disease experts from Colombia, Japan, Nigeria, Thailand, the UK, and the USA, combined with evidence from a literature review, were used to assess the value of RDTs in ASPs. From this, a value framework is proposed which aims to define the benefits of RDT use in ASPs, separate from per-patient benefits. Expert insights highlight that, to realise the value of RDTs within ASPs, effective implementation is key; actionable advice for choosing an RDT is proposed. Experts advocate the inclusion of RDTs in the World Health Organization Model List of essential in vitro diagnostics and in iterative development of national action plans.

Management of Bacterial and Fungal Infections in the ICU: Diagnosis, Treatment, and Prevention Recommendations.

Bacterial and fungal infections are common issues for patients in the intensive care unit (ICU). Large, multinational point prevalence surveys have identified that up to 50% of ICU patients have a diagnosis of bacterial or fungal infection at any one time. Infection in the ICU is associated with its own challenges. Causative organisms often harbour intrinsic and acquired mechanisms of drug-resistance, making empiric and targeted antimicrobial selection challenging. Infection in the ICU is associated with worse clinical outcomes for patients. We review the epidemiology of bacterial and fungal infection in the ICU. We discuss risk factors for acquisition, approaches to diagnosis and management, and common strategies for the prevention of infection.

Towards optimizing carbapenem selection in stewardship strategies: a prospective propensity score-matched study of ertapenem versus class 2 carbapenems for empirical treatment of third-generation cephalosporin-resistant Enterobacterales bacteraemia.

BACKGROUND: Third-generation cephalosporin-resistant Enterobacterales (3GCRE) are increasing in prevalence, leading to greater carbapenem consumption. Selecting ertapenem has been proposed as a strategy to reduce carbapenem resistance development. However, there are limited data for the efficacy of empirical ertapenem for 3GCRE bacteraemia. OBJECTIVES: To compare the efficacy of empirical ertapenem and class 2 carbapenems for the treatment of 3GCRE bacteraemia. METHODS: A prospective non-inferiority observational cohort study was performed from May 2019 to December 2021. Adult patients with monomicrobial 3GCRE bacteraemia receiving carbapenems within 24 h were included at two hospitals in Thailand. Propensity scores were used to control for confounding, and sensitivity analyses were performed in several subgroups. The primary outcome was 30 day mortality. This study is registered with clinicaltrials.gov (NCT03925402). RESULTS: Empirical carbapenems were prescribed in 427/1032 (41%) patients with 3GCRE bacteraemia, of whom 221 received ertapenem and 206 received class 2 carbapenems. One-to-one propensity score matching resulted in 94 pairs. Escherichia coli was identified in 151 (80%) of cases. All patients had underlying comorbidities. Septic shock and respiratory failure were the presenting syndromes in 46 (24%) and 33 (18%) patients, respectively. The overall 30 day mortality rate was 26/188 (13.8%). Ertapenem was non-inferior to class 2 carbapenems in 30 day mortality (12.8% versus 14.9%; mean difference -0.02; 95% CI: -0.12 to 0.08). Sensitivity analyses were consistent regardless of aetiological pathogens, septic shock, source of infection, nosocomial acquisition, lactate levels or albumin levels. CONCLUSIONS: Ertapenem may be of comparable efficacy to class 2 carbapenems in the empirical treatment of 3GCRE bacteraemia.

Can precision antibiotic prescribing help prevent the spread of carbapenem-resistant organisms in the hospital setting?

The emergence of carbapenem-resistant organisms (CROs) is a significant global threat. Reduction of carbapenem consumption can decrease CROs. In the global endemic era of ESBL-producing bacteria, carbapenems are considered the treatment of choice, leading to challenge in limiting carbapenem use. This review describes the role of precision prescribing for prevention of CROs. This involves improving antibiotic selection, dosing and shortening duration. The effect of different antibiotics, dosing and duration on CRO development are explored. Available options for precision prescribing, gaps in the scientific evidence, and areas for future research are also presented.

A rapid, simple, high-performance liquid chromatography method for the clinical measurement of beta-lactam antibiotics in serum and interstitial fluid.

Background: enhanced methods of therapeutic drug monitoring are required to support the individualisation of antibiotic dosing based on pharmacokinetics (PK) parameters. PK studies can be hampered by limited total serum volume, especially in neonates, or by sensitivity in the case of critically ill patients. We aimed to develop a liquid chromatography-mass spectrometry (LC/MS) analysis of benzylpenicillin, phenoxymethylpenicillin and amoxicillin in single low volumes of human serum and interstitial fluid (ISF) samples, with an improved limit of detection (LOD) and limit of quantification (LOQ), compared with previously published assays. Methods: sample clean-up was performed by protein precipitation using acetonitrile. Reverse phase chromatography was performed using triple quadrupole LC/MS. The mobile phase consisted of 55% methanol in water + 0.1% formic acid, with a flow rate of 0.4 mL min-1. Antibiotics stability was assessed at different temperatures. Results: chromatographic separation was achieved within 3 minutes for all analytes. Three common penicillins can now be measured in a single low-volume blood and ISF sample (15 µL) for the first time. Validation has demonstrated the method to be linear over the range 0.0015-10 mg L-1, with an accuracy of 93-104% and high sensitivity, with LOD ≈ 0.003 mg L-1 and LOQ ≈ 0.01 mg L-1 for all three analytes, which is critical for use in dose optimisation/individualisation. All evaluated penicillins indicated good stability at room temperature over 4 h, at (4 °C) over 24 h and at -80 °C for 6 months. Conclusion: the developed method is simple, rapid, accurate and clinically applicable for the quantification of three penicillin classes.

Triple quadrupole LC/MS method for the simultaneous quantitative measurement of cefiderocol and meropenem in serum.

Background: therapeutic drug monitoring is a crucial aspect of the management of hospitalized patients. The correct dosage of antibiotics is imperative to ensure their adequate exposure specially in critically ill patients. The aim of this study is to establish and validate a robust and fast liquid chromatography-tandem mass spectrometry (LC/MS) method for the simultaneous quantification of two important antibiotics in critically ill patients, cefiderocol and meropenem in human plasma. Methods: sample clean-up was performed by protein precipitation using acetonitrile. Reverse phase chromatography was performed using triple quadrupole LC/MS. The mobile phase was consisted of 55% methanol in water +0.1% formic acid, with flow rate of 0.4 ml min-1. Antibiotics stability was assessed at different temperatures. Serum protein binding was assessed using ultrafiltration devices. Results: chromatographic separation was achieved within 1.5 minutes for all analytes. Validation has demonstrated the method to be linear over the range 0.0025-50 mg L-1 for cefiderocol and 0.00028-50 mg L-1 for meropenem, with accuracy of 94-101% and highly sensitive, with LLOQ ≈ 0.02 mg L-1 and 0.003 mg L-1 for cefiderocol and meropenem, respectively. Both cefiderocol and meropenem showed a good stability at room temperature over 6 h, and at (4 °C) over 24 h. Cefiderocol and meropenem demonstrated a protein binding of 49-60% and 98%, respectively in human plasma. Conclusion: the developed method is simple, rapid, accurate and clinically applicable for the quantification of cefiderocol and meropenem.

Machine learning and synthetic outcome estimation for individualised antimicrobial cessation.

The decision on when it is appropriate to stop antimicrobial treatment in an individual patient is complex and under-researched. Ceasing too early can drive treatment failure, while excessive treatment risks adverse events. Under- and over-treatment can promote the development of antimicrobial resistance (AMR). We extracted routinely collected electronic health record data from the MIMIC-IV database for 18,988 patients (22,845 unique stays) who received intravenous antibiotic treatment during an intensive care unit (ICU) admission. A model was developed that utilises a recurrent neural network autoencoder and a synthetic control-based approach to estimate patients' ICU length of stay (LOS) and mortality outcomes for any given day, under the alternative scenarios of if they were to stop vs. continue antibiotic treatment. Control days where our model should reproduce labels demonstrated minimal difference for both stopping and continuing scenarios indicating estimations are reliable (LOS results of 0.24 and 0.42 days mean delta, 1.93 and 3.76 root mean squared error, respectively). Meanwhile, impact days where we assess the potential effect of the unobserved scenario showed that stopping antibiotic therapy earlier had a statistically significant shorter LOS (mean reduction 2.71 days, p -value <0.01). No impact on mortality was observed. In summary, we have developed a model to reliably estimate patient outcomes under the contrasting scenarios of stopping or continuing antibiotic treatment. Retrospective results are in line with previous clinical studies that demonstrate shorter antibiotic treatment durations are often non-inferior. With additional development into a clinical decision support system, this could be used to support individualised antimicrobial cessation decision-making, reduce the excessive use of antibiotics, and address the problem of AMR.

Haematological and hepatic adverse effects of ceftriaxone in ambulatory care: a dual-centre retrospective observational analysis of standard vs high dose.

BACKGROUND: European Committee on Antimicrobial Susceptibility Testing (EUCAST) breakpoint criteria for methicillin-susceptible Staphylococcus aureus (MSSA) treatment with ceftriaxone are based upon high dose (4 g/day) rather than standard dose (2 g/day) posology. This is particularly relevant for invasive infections, and for patients managed via Outpatient Parenteral Antimicrobial Therapy (OPAT), but may result in increased drug toxicity. We quantified the incidence of neutropenia, thrombocytopenia and raised liver enzymes between standard and high dose ceftriaxone in adult patients. METHOD: Adult outpatients prescribed ≥ 7 days of ceftriaxone therapy were identified, and clinical, pharmacological, and laboratory parameters extracted from electronic health records between May 2021 and December 2021. Incidence and median time to haematological and hepto-toxicity were analysed. Univariate odds ratios were calculated for neutrophil count and ALT levels with 95% confidence level and Chi squared/Fisher's exact test used to identify statistical significance. RESULTS: Incidence of neutropenia was comparable between both groups; 8/47 (17%) in the 2 g group vs 6/39 (15.4%) in the 4 g group (OR 0.89 (95% CI 0.26-2.63), p > 0.999). Median time to neutropenia was 12 and 17 days in the 2 g and 4 g groups respectively. Thrombocytopenia was observed in 0/47 in the 2 g group compared with 3/39 (7.7%) in the 4 g group (p 0.089). Median time to thrombocytopenia was 7 days in the 4 g group. Elevated liver enzymes did not clearly correlate with ceftriaxone dosing; present in 5/47 (10.6%) and 2/39 (5.1%) for 2 g and 4 g respectively (OR 0.45 (95% CI 0.87-2.36), p 0.448). Treatment cessation due to any adverse effect was similar between both groups 2/47 (4.3%) for 2 g and 3/39 (7.7%) for 4 g (OR 1.86 (95% CI 0.36-10.92), p 0.655). CONCLUSIONS: Increased adverse effects with 4 g (over 2 g) daily dosing of ceftriaxone was not observed in an OPAT population. However absolute development of haematological and liver dyscrasias was appreciable-monitoring of liver function and full blood count in patients receiving prolonged ceftriaxone is indicated irrespective of dosing.

Closed-loop control of continuous piperacillin delivery: An in silico study.

Background and objective: Sub-therapeutic dosing of piperacillin-tazobactam in critically-ill patients is associated with poor clinical outcomes and may promote the emergence of drug-resistant infections. In this paper, an in silico investigation of whether closed-loop control can improve pharmacokinetic-pharmacodynamic (PK-PD) target attainment is described. Method: An in silico platform was developed using PK data from 20 critically-ill patients receiving piperacillin-tazobactam where serum and tissue interstitial fluid (ISF) PK were defined. Intra-day variability on renal clearance, ISF sensor error, and infusion constraints were taken into account. Proportional-integral-derivative (PID) control was selected for drug delivery modulation. Dose adjustment was made based on ISF sensor data with a 30-min sampling period, targeting a serum piperacillin concentration between 32 and 64 mg/L. A single tuning parameter set was employed across the virtual population. The PID controller was compared to standard therapy, including bolus and continuous infusion of piperacillin-tazobactam. Results: Despite significant inter-subject and simulated intra-day PK variability and sensor error, PID demonstrated a significant improvement in target attainment compared to traditional bolus and continuous infusion approaches. Conclusion: A PID controller driven by ISF drug concentration measurements has the potential to precisely deliver piperacillin-tazobactam in critically-ill patients undergoing treatment for sepsis.

Application of therapeutic drug monitoring to the treatment of bacterial central nervous system infection: a scoping review.

BACKGROUND: Bacterial central nervous system (CNS) infection is challenging to treat and carries high risk of recurrence, morbidity, and mortality. Low CNS penetration of antibiotics may contribute to poor clinical outcomes from bacterial CNS infections. The current application of therapeutic drug monitoring (TDM) to management of bacterial CNS infection was reviewed. METHODS: Studies were included if they described adults treated for a suspected/confirmed bacterial CNS infection and had antibiotic drug concentration(s) determined that affected individual treatment. RESULTS: One-hundred-and-thirty-six citations were retrieved. Seventeen manuscripts were included describing management of 68 patients. TDM for vancomycin (58/68) and the beta-lactams (29/68) was most common. Timing of clinical sampling varied widely between studies and across different antibiotics. Methods for setting individual PK-PD targets, determining parameters and making treatment changes varied widely and were sometimes unclear. DISCUSSION: Despite increasing observational data showing low CNS penetration of various antibiotics, there are few clinical studies describing practical implementation of TDM in management of CNS infection. Lack of consensus around clinically relevant CSF PK-PD targets and protocols for dose-adjustment may contribute. Standardised investigation of TDM as a tool to improve treatment is required, especially as innovative drug concentration-sensing and PK-PD modelling technologies are emerging. Data generated at different centres offering TDM should be open access and aggregated to enrich understanding and optimize application.

Addition of probenecid to oral ß-lactam antibiotics: a systematic review and meta-analysis.

OBJECTIVES: To explore the literature comparing the pharmacokinetic and clinical outcomes from adding probenecid to oral ß-lactams. METHODS: Medline and EMBASE were searched from inception to December 2021 for all English language studies comparing the addition of probenecid (intervention) with an oral ß-lactam [flucloxacillin, penicillin V, amoxicillin (±â€Šclavulanate), cefalexin, cefuroxime axetil] alone (comparator). ROBINS-I and ROB-2 tools were used. Data on antibiotic therapy, infection diagnosis, primary and secondary outcomes relating to pharmacokinetics and clinical outcomes, plus adverse events were extracted and reported descriptively. For a subset of studies comparing treatment failure between probenecid and control groups, meta-analysis was performed. RESULTS: Overall, 18/295 (6%) screened abstracts were included. Populations, methodology and outcome data were heterogeneous. Common populations included healthy volunteers (9/18; 50%) and those with gonococcal infection (6/18; 33%). Most studies were crossover trials (11/18; 61%) or parallel-arm randomized trials (4/18; 22%). Where pharmacokinetic analyses were performed, addition of probenecid to oral ß-lactams increased total AUC (7/7; 100%), Cmax (5/8; 63%) and serum t½ (6/8; 75%). Probenecid improved PTA (2/2; 100%). Meta-analysis of 3105 (2258 intervention, 847 control) patients treated for gonococcal disease demonstrated a relative risk of treatment failure in the random-effects model of 0.33 (95% CI 0.20-0.55; I2 = 7%), favouring probenecid. CONCLUSIONS: Probenecid-boosted ß-lactam therapy is associated with improved outcomes in gonococcal disease. Pharmacokinetic data suggest that probenecid-boosted oral ß-lactam therapy may have a broader application, but appropriately powered mechanistic and efficacy studies are required.

Antiviral metabolite 3'-deoxy-3',4'-didehydro-cytidine is detectable in serum and identifies acute viral infections including COVID-19.

BACKGROUND: There is a critical need for rapid viral infection diagnostics to enable prompt case identification in pandemic settings and support targeted antimicrobial prescribing. METHODS: Using untargeted high-resolution liquid chromatography coupled with mass spectrometry, we compared the admission serum metabolome of emergency department patients with viral infections (including COVID-19), bacterial infections, inflammatory conditions, and healthy controls. Sera from an independent cohort of emergency department patients admitted with viral or bacterial infections underwent profiling to validate findings. Associations between whole-blood gene expression and the identified metabolite of interest were examined. FINDINGS: 3'-Deoxy-3',4'-didehydro-cytidine (ddhC), a free base of the only known human antiviral small molecule ddhC-triphosphate (ddhCTP), was detected for the first time in serum. When comparing 60 viral with 101 non-viral cases in the discovery cohort, ddhC was the most significantly differentially abundant metabolite, generating an area under the receiver operating characteristic curve (AUC) of 0.954 (95% CI: 0.923-0.986). In the validation cohort, ddhC was again the most significantly differentially abundant metabolite when comparing 40 viral with 40 bacterial cases, generating an AUC of 0.81 (95% CI 0.708-0.915). Transcripts of viperin and CMPK2, enzymes responsible for ddhCTP synthesis, were among the five genes most highly correlated with ddhC abundance. CONCLUSIONS: The antiviral precursor molecule ddhC is detectable in serum and an accurate marker for acute viral infection. Interferon-inducible genes viperin and CMPK2 are implicated in ddhC production in vivo. These findings highlight a future diagnostic role for ddhC in viral diagnosis, pandemic preparedness, and acute infection management. FUNDING: NIHR Imperial BRC; UKRI.